WEBVTT
00:00:00.160 --> 00:00:01.600
Have you seen the videos popping up?
00:00:01.760 --> 00:00:07.280
Claude killed lawyer today, Claude killed designers today, Claude killed coders today.
00:00:07.519 --> 00:00:09.199
So when Claude's been killing everyone.
00:00:09.439 --> 00:00:11.439
Claude doesn't seem that mean to me, honestly.
00:00:11.679 --> 00:00:16.320
And I come back and sort of like look at all of the versions and they say, okay, now you do this, you do that.
00:00:16.559 --> 00:00:19.760
So your AI team is developing.
00:00:20.079 --> 00:00:20.559
They are.
00:00:20.879 --> 00:00:28.399
They are not just developing, they're like managing the search engine optimization, managing my household, finance, everything.
00:00:28.559 --> 00:00:32.640
Like you've got agents for everyone in our organization as well.
00:00:32.880 --> 00:00:35.920
One month my bill of AI was around$15,000.
00:00:36.560 --> 00:00:37.039
Wow.
00:00:37.280 --> 00:00:42.719
Because I just and uh all of the APIs, uh consumption, and so on.
00:00:42.880 --> 00:00:43.119
Yeah.
00:00:43.280 --> 00:00:43.520
Yes.
00:00:43.759 --> 00:00:48.479
I've got a little image in my brain of you just sitting there just, you know, like your little AI fire just burning money.
00:00:49.600 --> 00:00:51.679
Project has been selected.
00:00:51.759 --> 00:00:55.679
Uh there's an AI for good council that's actually supported by United Nations.
00:00:55.759 --> 00:00:59.759
So there's AI for goods are they have got a pitching competition.
00:00:59.840 --> 00:01:02.560
So we are in like top five in Australia that has been selected.
00:01:02.640 --> 00:01:03.439
So I'm going to vote.
00:01:04.400 --> 00:01:05.920
Uh thank you.
00:01:06.079 --> 00:01:06.400
Thanks a lot.
00:01:07.840 --> 00:01:08.640
That's something.
00:01:18.560 --> 00:01:21.280
Hi everyone, and welcome back to the Dumb Monkey Podcast.
00:01:21.439 --> 00:01:24.400
We've had a bit of a break, but we're excited to be back on deck.
00:01:24.640 --> 00:01:28.239
And for those of you who are joining us for the first time, hi, I'm Davina Montgomery.
00:01:28.400 --> 00:01:38.000
I am your host for the Dumb Monkey Podcast because I was asked to be here by the quite extraordinary founder of the Dumb Monkey Academy of Enterprise Monkey.
00:01:38.400 --> 00:01:40.719
Look, we're not going to go into the list of founding, it takes too long.
00:01:40.959 --> 00:01:46.400
All of the monkeys, and of course, co-host of the Dumb Monkey podcast, Amir Katoob.
00:01:46.560 --> 00:01:47.120
Amir, hello.
00:01:47.599 --> 00:01:48.400
Hi, Davina.
00:01:48.640 --> 00:01:50.400
It's so good to see you after a long time.
00:01:50.799 --> 00:01:51.439
It's so great to be back.
00:01:52.000 --> 00:01:52.480
Yes, absolutely.
00:01:52.799 --> 00:01:54.319
We've got a cool new podcast set up.
00:01:54.400 --> 00:01:54.959
This is exciting.
00:01:55.200 --> 00:01:55.439
Yes.
00:01:55.519 --> 00:01:56.239
How have you been?
00:01:56.480 --> 00:01:57.519
I've been really good.
00:01:57.599 --> 00:01:59.120
I'm really happy to be in here.
00:01:59.200 --> 00:02:02.319
There's been uh so much happening in the world of AI.
00:02:02.719 --> 00:02:12.800
But I feel like I always understand a little bit more and feel like I come away with better questions after I've been speaking to you and of course the wonderful Tanita who's behind the scenes producing.
00:02:13.039 --> 00:02:14.479
Tanita, welcome back.
00:02:16.000 --> 00:02:24.240
I mean It has been a few months since you and I have sat down at the podcast table, um, which I'm assuming in AI terms is half a generation.
00:02:24.719 --> 00:02:24.960
It has.
00:02:25.199 --> 00:02:26.240
It seems like it has changed.
00:02:26.479 --> 00:02:27.919
Yes, so much has changed, yes.
00:02:28.319 --> 00:02:29.840
Let's let's jump off to the top.
00:02:29.919 --> 00:02:30.800
Um I don't know.
00:02:30.960 --> 00:02:41.919
I think when you and I were last talking, Anthropic was something, and Claude particularly was something that people use specifically for writing, that we were at that stage, um, that feels like a decade ago.
00:02:42.080 --> 00:02:44.719
But of course, now anthropic's just gone.
00:02:45.759 --> 00:02:46.159
Yes.
00:02:46.400 --> 00:02:48.000
Have you seen the videos popping up?
00:02:48.159 --> 00:02:53.759
Claude killed lawyer today, Claude killed designers today, Claude killed coders today.
00:02:53.919 --> 00:02:55.520
So when Claude's been killing everyone.
00:02:55.919 --> 00:02:57.840
Claude doesn't seem that mean to me, honestly.
00:02:58.000 --> 00:03:02.000
I mean, I I feel like Claude doesn't have a dark side in that sense.
00:03:02.400 --> 00:03:05.280
I think it's just uh people trying to hype up what it's doing.
00:03:05.439 --> 00:03:05.680
Yeah.
00:03:05.840 --> 00:03:13.280
But also the tool they have, tools they have been launching uh like throughout from the beginning of this year has been really amazing.
00:03:13.439 --> 00:03:18.080
Like last year, we How much time have you been sitting there on Claude tools?
00:03:18.400 --> 00:03:25.759
So look, you know, before uh you came, I was actually lying down on my couch and trying to take a nap because I because you're up at 3 a.m.
00:03:25.840 --> 00:03:26.159
in the morning.
00:03:26.400 --> 00:03:29.439
I was, yes, yes, you you know me really well.
00:03:29.759 --> 00:03:32.719
And then I've almost become a bit of a Claude addict.
00:03:33.039 --> 00:03:34.240
That's just an actual term.
00:03:34.479 --> 00:03:35.280
Yeah, I think it is.
00:03:35.520 --> 00:03:37.840
Because uh I'm able to get so much done.
00:03:38.000 --> 00:03:42.560
So uh and so I've been spending a lot of time with with AI.
00:03:42.639 --> 00:03:55.840
It's it's been in a different zone this year because now I'm able to actually say almost I've got like sort of multiple versions of AI or Claude, say, hey, you do this, you do this, you do this, and I'm gonna check your work.
00:03:56.000 --> 00:04:00.800
And I come back and sort of like look at all of the versions and then say, okay, now you do this, you do that.
00:04:01.039 --> 00:04:04.240
So your AI team is developing?
00:04:04.639 --> 00:04:14.159
They are, it's they're not just developing, they are like uh managing the search engine optimization, managing my household, finance, uh everything.
00:04:14.319 --> 00:04:18.319
Like you've got agents for everyone in our organization as well.
00:04:18.480 --> 00:04:23.199
Tanisha has got her little cute Inky who helps her with uh content creation.
00:04:23.600 --> 00:04:24.160
Fantastic.
00:04:24.399 --> 00:04:30.079
Uh and then I've got a full team of autonomous agents that are working together as well.
00:04:30.319 --> 00:04:34.160
And most of them are using Claude APIs uh directly.
00:04:34.560 --> 00:04:43.199
But the actual uh Claude tool, the Claude uh desktop version, has become so powerful that you can do a lot of this stuff with that as well.
00:04:43.600 --> 00:04:56.399
So for for people that perhaps haven't, for our listeners that perhaps haven't uh really used a lot of Claude or perhaps is still back in that case of kind of going, maybe I'll pop over and use this for a bit of writing if I want to, if I want to tighten something up.
00:04:57.360 --> 00:04:59.279
How would you describe the difference?
00:04:59.519 --> 00:05:00.399
What's changed?
00:05:00.959 --> 00:05:10.319
So what has uh significantly changed is like last year we were using Chat GPT or Claude where we would ask it a question, it can provide us a response.
00:05:10.639 --> 00:05:20.079
What Claude has done is that the capacity of the models that sit behind it, the intelligence has not changed, but the way they have created, especially the desktop tool.
00:05:20.319 --> 00:05:27.360
So you can use Claude on the comp online, but if you download Claude Desktop, they have an option called Claude Cowork.
00:05:27.600 --> 00:05:34.319
And with cowork, you can actually connect it to a folder on your computer and you can say, hey, there are so many files over there.
00:05:34.480 --> 00:05:43.040
I want you to go through those files, create me a summary, create me a report, create me a beautiful annual report or a document.
00:05:43.199 --> 00:05:50.000
It can do all of that analysis and actually do the work and create some really amazing, not just documents.
00:05:50.160 --> 00:05:57.279
Like if you want to do analysis, think of it as like any work that you can do on computer can be done with the help of Claude.
00:05:57.439 --> 00:06:00.480
So they have got a connector for Chrome that can control your browser.
00:06:00.959 --> 00:06:05.519
They've got Cloud Computer use that can execute any task on your computer as well.
00:06:05.680 --> 00:06:13.040
So it's like And using into into Excel, and if you're if you're using the Microsoft Suite, so you can get in and start using the Claude tools in there.
00:06:13.360 --> 00:06:13.759
Yes.
00:06:14.000 --> 00:06:18.800
It's almost like having someone sitting beside you and saying, Hey, can you do this for me, please?
00:06:19.120 --> 00:06:22.720
And now uh there's an option called dispatch in the Cloud app.
00:06:22.800 --> 00:06:30.160
So if you have your computer open at home and you're driving back to home and say, Hey, you know, I had this meeting.
00:06:30.480 --> 00:06:31.519
And generally I do this.
00:06:31.680 --> 00:06:32.959
I had this meeting with this guy.
00:06:33.120 --> 00:06:35.360
He asked, and I would say he asked for something.
00:06:35.439 --> 00:06:36.720
I wouldn't even tell what he asked for.
00:06:36.800 --> 00:06:40.000
I say, I'll say, go and check the meeting transcript.
00:06:40.240 --> 00:06:43.920
Can you make sure that we have the proposal ready before I get home?
00:06:44.399 --> 00:06:45.360
So it starts to work.
00:06:45.600 --> 00:06:46.639
It's a revenue world, isn't it?
00:06:46.879 --> 00:06:47.600
It is, yes.
00:06:48.079 --> 00:07:01.839
Um I mean, I wanted to to ask then about you you raised this idea that you can you're using your I I want to call it Claude Claude family, your Claude friends, um, your Claude team.
00:07:01.920 --> 00:07:04.240
I don't know, I haven't decided what that maybe it's the Claude crew.
00:07:04.560 --> 00:07:05.439
It's a Claude crew.
00:07:05.680 --> 00:07:12.480
And so you're using your your Claude crew as an extension of your brain to sit there and go, right, I've got these things that I want to get done.
00:07:12.800 --> 00:07:13.759
What can I offload?
00:07:13.920 --> 00:07:16.480
What can I get happening while I'm doing other things?
00:07:16.720 --> 00:07:19.439
So you're becoming more productive in yourself by doing that.
00:07:19.600 --> 00:07:20.560
That's what I'm hearing.
00:07:20.959 --> 00:07:25.920
What I'm wondering is you raised something that to say, okay, I don't need to give you a whole lot of information.
00:07:26.079 --> 00:07:27.759
Can you go and do this for me?
00:07:28.000 --> 00:07:35.040
And now the the power of the model that's on the desktop version of Claude is can do that in a pretty effective way.
00:07:35.199 --> 00:07:40.720
Um as always, AI can hallucinate, please check your work, just for those of you at home.
00:07:40.959 --> 00:07:42.879
And it doesn't replace people.
00:07:43.040 --> 00:07:58.000
Um what I want to know is how good is say Claude Desktop at being able to train someone who's very nervous about using AI for the first time or starting to use AI for different things.
00:07:58.079 --> 00:08:01.600
So say someone wants to create their own their own Claude crew.
00:08:02.079 --> 00:08:04.000
Can Claude train or help them?
00:08:04.160 --> 00:08:06.399
Can Claude hold their hand and get them through that?
00:08:06.480 --> 00:08:08.480
Can Claude be their emotional support dumb monkey?
00:08:08.800 --> 00:08:10.639
That's that's uh that's a very good one.
00:08:10.879 --> 00:08:14.000
So it's a very, I'll give you a very interesting example.
00:08:14.160 --> 00:08:15.920
It's not Claude directly.
00:08:16.079 --> 00:08:18.240
Uh you must have heard of OpenClaw.
00:08:18.399 --> 00:08:18.480
Yeah.
00:08:18.720 --> 00:08:28.000
Uh that's uh an agentic system that you can download on computer, and then they can it can talk to uh someone via WhatsApp or Telegram.
00:08:28.240 --> 00:08:36.960
So my mother-in-law in India, and she is, she's taken, like she's an English literature professor, she has taken a leave to write a book.
00:08:37.200 --> 00:08:42.159
And she said to me, You're really good at researching, you've got all of these deep research tools.
00:08:42.240 --> 00:08:44.240
Can you do a bit of research for me?
00:08:44.480 --> 00:08:49.519
And I thought it's gonna be, she's gonna be asking me about the research multiple times.
00:08:49.679 --> 00:08:50.559
What should I do?
00:08:50.720 --> 00:08:52.720
So I actually created an agent.
00:08:52.799 --> 00:08:58.960
Uh I called the agent Cuckoo because that's uh my mother-in-law's uh uh family name as well.
00:08:59.120 --> 00:09:09.200
And I said, This is my mother-in-law, she's really smart when it comes to literature, but she doesn't know technology at all, she doesn't know how to talk to you or how to use you.
00:09:09.600 --> 00:09:12.080
She's doing this research, doing this work.
00:09:12.240 --> 00:09:14.240
These are the six books that she's working on.
00:09:14.320 --> 00:09:21.679
I gave it access and said, What I need you to do is I want you to assist her, but also teach her how you can actually assist her.
00:09:21.919 --> 00:09:28.240
So, what happens is that uh Cuckoo messaged her on WhatsApp and said, Hey, I'm Cuckoo, I'm your assistant.
00:09:28.480 --> 00:09:29.919
And and Cuckoo calls her mama.
00:09:30.399 --> 00:09:31.200
Like, hey mama.
00:09:31.919 --> 00:09:36.159
You've identified your mother-in-law, so which is next level.
00:09:36.320 --> 00:09:40.000
If you're gonna take one thing away from this podcast, that's maybe your idea.
00:09:40.240 --> 00:09:40.720
That's right.
00:09:40.879 --> 00:09:45.600
So and says, Hey mama, I'm Cuckoo, I've been an assistant that's made by Amit.
00:09:46.000 --> 00:09:47.120
I'm here to help you.
00:09:47.279 --> 00:09:49.039
I'm gonna go through the books.
00:09:49.120 --> 00:09:50.240
Uh, this is what I'm thinking.
00:09:50.320 --> 00:09:51.279
What do you think?
00:09:51.600 --> 00:09:56.080
And so, and then my mother-in-law would reply, Mother, my mother-in-law was shocked, right?
00:09:56.320 --> 00:09:58.399
I was gonna say, what is she what was her reaction?
00:09:58.559 --> 00:09:59.279
What did she think?
00:09:59.679 --> 00:10:02.320
She thought I was playing a prank on her, and it's like me.
00:10:02.559 --> 00:10:05.120
And she said, uh she said, Amit, is it you?
00:10:05.279 --> 00:10:07.039
And it's like, question mark, question mark.
00:10:07.120 --> 00:10:09.840
Then she said, No, mama, it's not me, and so on.
00:10:09.919 --> 00:10:12.639
But then I called her up and I told her what this is.
00:10:12.879 --> 00:10:14.879
And she's like, Okay, how should I talk to it?
00:10:15.039 --> 00:10:15.919
What should I ask?
00:10:16.159 --> 00:10:26.799
And I said, like, whatever you would have asked me to do, treat her like that, or maybe someone, an assistant that's working, helping you with research, if you have to brainstorm the ideas.
00:10:27.200 --> 00:10:30.720
Or if you're not sure, just ask her, hey Cuckoo, how can you help me?
00:10:31.279 --> 00:10:33.360
And then she started to interact with it.
00:10:33.519 --> 00:10:47.440
And because Cuckoo already knew the level, the technical level of my mother-in-law, this it's Cuckoo started to interact with her, started to suggest saying, for example, my mother said, Can you look at this book and help give me a quick five-point summary?
00:10:47.600 --> 00:10:54.080
So it gave her the five-point summary, but then she said, What actually's gonna be helpful, mama, is if I can do this for you as well.
00:10:54.240 --> 00:10:58.240
How about I actually do a further synopsis analysis and all of that?
00:10:58.559 --> 00:11:02.559
So, so Claude, and and it's basically Claude that's sitting behind it.
00:11:02.720 --> 00:11:13.360
So Claude APIs or Claude uh brain is really good like that, being able to understand who is the person that's talking to, and then um working around that.
00:11:13.519 --> 00:11:15.840
And I've got a very interesting Valentine's Day story.
00:11:15.919 --> 00:11:20.159
I've narrated it a lot of time during my speeches as well, but I think you wouldn't have heard it, right?
00:11:20.320 --> 00:11:21.360
No, I haven't heard it.
00:11:21.600 --> 00:11:31.919
Okay, so this Valentine's Day uh for Sarah, uh, I actually decided to gift her an agent because I thought what could be more romantic than gifting your wife an agent?
00:11:32.240 --> 00:11:32.960
I just wanted to.
00:11:33.200 --> 00:11:35.039
What do you give someone who's got everything?
00:11:35.279 --> 00:11:38.879
Something that can do a whole lot of things that they don't want to do.
00:11:39.120 --> 00:11:40.240
Yes, yes.
00:11:40.399 --> 00:11:44.000
So uh and again very thoughtful gift.
00:11:44.240 --> 00:11:45.200
It is, yes.
00:11:45.360 --> 00:11:47.279
And and uh the name was Miro.
00:11:47.360 --> 00:11:49.360
Uh so Miro again messaged Sarah.
00:11:49.440 --> 00:11:51.919
I already told Sarah this is your Valentine's Day gift.
00:11:52.080 --> 00:12:00.879
So Miro started to chat with Sarah and it's like, and I told to Sarah, like, whatever things you tell me about, I was like, we need to remember to do this and so on.
00:12:01.039 --> 00:12:06.960
And I connected it with our family calendar, her emails where we receive all of the school information and so on.
00:12:07.200 --> 00:12:11.279
And so she started to say, Hey Sarah, I will I would like to know more about you.
00:12:11.440 --> 00:12:13.679
What are the things that uh bother you?
00:12:13.840 --> 00:12:15.440
What are the things that drive you crazy?
00:12:15.679 --> 00:12:18.320
And Sarah took it as an opportunity to just vent out.
00:12:21.360 --> 00:12:24.000
My Valentine's Day gift to you is a bit of AI therapy.
00:12:24.240 --> 00:12:24.799
Yes, yes.
00:12:25.200 --> 00:12:30.000
It's such a funny voice note uh that we'll probably add to the podcast recording.
00:12:30.080 --> 00:12:34.960
But Sarah starts to say, hey, my vanity has been broken for the last two years.
00:12:35.039 --> 00:12:42.639
It's a guest vanity, it's not her vanity, but it's like I have been asking me to fix it, it's not been fixed, and and so on.
00:12:42.799 --> 00:12:56.240
So Miro actually went on Air Tasker and posted a job and said, uh, I'm looking for a plumber to fix the vanity and negotiated the plumber down as well, and then came back to Sarah and said, Here's here's your plumber.
00:12:56.320 --> 00:12:58.000
Do you want me to book it, book him in?
00:12:58.639 --> 00:13:01.519
So yes.
00:13:01.600 --> 00:13:04.000
And so the plumber came and the vanity is fixed.
00:13:04.320 --> 00:13:04.879
Fantastic.
00:13:05.039 --> 00:13:08.879
And now you don't need to speak to Sarah anymore, so you guys will remain married forever.
00:13:09.120 --> 00:13:09.600
I know.
00:13:09.759 --> 00:13:14.799
So we we can we can talk to her about yes, non non-essential stuff.
00:13:14.960 --> 00:13:16.639
But how is that going?
00:13:16.799 --> 00:13:23.120
How does that so for for your mother-in-law, for Sarah, how is this working for for them at an emotional level?
00:13:23.279 --> 00:13:24.480
Do they enjoy this experience?
00:13:24.639 --> 00:13:26.159
Are they getting something out of it?
00:13:26.320 --> 00:13:29.279
Does it help to decrease some of the stresses?
00:13:29.759 --> 00:13:32.480
It has actually uh a lot.
00:13:32.879 --> 00:13:36.240
And what has happened is I can see there's a bit of dependence on it.
00:13:36.320 --> 00:13:42.639
So there's been two or three times because it's using credits uh and it costs me to run it.
00:13:42.799 --> 00:13:43.039
Yes.
00:13:43.279 --> 00:13:53.039
So every time it's run run out of credit credit usage, they call me upstairs instead of Cuckoo is not responding, Mira is not responding, I had to tell her about this, and it's not happening.
00:13:53.279 --> 00:13:59.120
And so and that's and that's one of the first things that people find when they start to use these kinds of assistance, isn't it?
00:13:59.279 --> 00:14:02.320
That we that these AI models use tokens.
00:14:02.559 --> 00:14:02.720
Yes.
00:14:02.960 --> 00:14:07.679
I mean we we will never we will never stop saying that these these platforms are business models.
00:14:07.840 --> 00:14:08.080
Yes.
00:14:08.240 --> 00:14:10.960
They're designed to be helpful to you so that you will use it more.
00:14:11.039 --> 00:14:11.200
Yeah.
00:14:11.360 --> 00:14:12.879
And then they can charge you more to do that.
00:14:12.960 --> 00:14:15.519
So that's that's part of the business model that sits behind that.
00:14:15.600 --> 00:14:19.279
Um they are heavily subsidizing the costs at the moment as well.
00:14:19.519 --> 00:14:20.879
And Tanisha is the same.
00:14:20.960 --> 00:14:24.000
Whenever Inky's not responding, she'll message me, my Ike is not responding to the case.
00:14:24.559 --> 00:14:25.679
I need more tokens, please.
00:14:25.919 --> 00:14:26.159
Yes.
00:14:26.480 --> 00:14:27.200
Uh my credit.
00:14:27.519 --> 00:14:28.080
Yeah, yeah.
00:14:28.320 --> 00:14:30.559
I haven't set it up on auto-recharge yet.
00:14:30.720 --> 00:14:37.519
Although you can set up the limit because I still want to physically go and put the money in because I have spent a lot of money on the credits.
00:14:37.919 --> 00:14:45.840
Yeah, that's well, that's there's a there's a whole big question that sits behind that that we might come to a bit later, which is around the business model of the AI models.
00:14:46.000 --> 00:14:46.159
Yes.
00:14:46.320 --> 00:14:47.120
So what's sitting behind it?
00:14:47.200 --> 00:14:48.879
Where's the business cases and what do they mean?
00:14:49.279 --> 00:14:49.600
Yeah.
00:14:49.919 --> 00:14:57.120
So coming back to like, let's not talk about open claw because it's a bit more advanced, there are security issues and so on as well.
00:14:57.360 --> 00:14:59.039
Even clot subscription, right?
00:14:59.120 --> 00:15:03.120
So I got a claude subscription for people in my project management team.
00:15:03.279 --> 00:15:03.440
Yeah.
00:15:03.600 --> 00:15:07.279
Now I knew that they don't know how to use cloud properly, right?
00:15:07.600 --> 00:15:13.039
So, and I knew they wouldn't be able to use it without a lot of training and without setting it up.
00:15:13.200 --> 00:15:20.320
So, first thing that I did was there's something in Cloud which is so powerful, and that's going to change the whole AI experience.
00:15:20.399 --> 00:15:22.320
So, ChatGPT is now adapting it as well.
00:15:22.480 --> 00:15:23.519
Copilot will adapt it.
00:15:23.600 --> 00:15:24.720
It's called skills.
00:15:25.039 --> 00:15:33.120
So, skills is are basically simple like text files through which you can give Claude a domain knowledge.
00:15:33.440 --> 00:15:36.399
You can give tell Claude what you need to do.
00:15:36.559 --> 00:15:41.840
So, the first skill that I created on Claude was an enterprise monkey brand skill.
00:15:42.080 --> 00:15:47.519
And in that brand skill, it has logos, the whole brand tone, and everything.
00:15:47.759 --> 00:15:54.399
So if you see some of the documents that uh come out of the Claud now, they are fully on brand.
00:15:54.559 --> 00:16:00.559
Like every document that creates is fully on brand to the enterprise monkey brands, fully on the tone as well.
00:16:00.720 --> 00:16:07.840
Then I created like a humanizer skill so that everything that comes out is, and humanizer skill is something that you can search.
00:16:07.919 --> 00:16:10.960
So there's a big uh marketplace through which you can search as well.
00:16:11.200 --> 00:16:12.559
And so I added that.
00:16:12.639 --> 00:16:18.399
And then I added like an enterprise monkey proposal skill, which could help them create a proposal.
00:16:18.799 --> 00:16:21.759
On the side, I also created a brain.
00:16:21.919 --> 00:16:33.200
My brain is a bit more advanced, but think of brain as a folder either in Google Drive or in SharePoint, which is a very uh curated piece of information.
00:16:33.360 --> 00:16:34.639
We have talked about a brain.
00:16:34.799 --> 00:16:39.120
Uh I've I've I've done something very close to JavaScript, I'll show it to you.
00:16:39.360 --> 00:16:40.399
It's really cool.
00:16:40.559 --> 00:16:45.840
Uh but it's basically very simple text files or doc files.
00:16:46.159 --> 00:16:53.440
Uh and then in Claude, you can say if there's any information that you need to find about the company, check brain first.
00:16:53.600 --> 00:16:59.440
And when it goes to that brain folder, there's like a README file saying, hey Claude, you have come here.
00:16:59.519 --> 00:17:04.000
If the question is about a person, check this folder if the question is about this.
00:17:04.079 --> 00:17:07.519
Like there's a very clear instruction that's kept in that brain.
00:17:07.759 --> 00:17:12.160
Now, the important thing is that that brain is actually kept outside of Claude.
00:17:12.319 --> 00:17:13.119
That's mine.
00:17:13.440 --> 00:17:17.359
The skills are we have uploaded to Claude, but skills are ours as well.
00:17:17.519 --> 00:17:18.960
They are sitting outside the cloud.
00:17:19.119 --> 00:17:24.880
So all of the magic that's happening is actually sitting outside the cloud, and we are just giving Claude access to that.
00:17:25.039 --> 00:17:34.960
So what I did for my team was that I actually set up the custom instruction and say, this is my team, it started to use project management uh this uh cloud right right now.
00:17:35.119 --> 00:17:37.519
They don't know how to use it, so you need to assist them.
00:17:37.680 --> 00:17:40.000
If they ask for information, use the brain.
00:17:40.160 --> 00:17:44.400
If they ask uh to produce a document, use the brand skill and so on.
00:17:44.640 --> 00:17:54.960
And so when I gave them the cloud loaded with all of the skills, spent a couple of hours with them showing them around, it was very easy and quick for them to actually get started on that journey.
00:17:55.599 --> 00:18:00.799
Another thing that I did for them was there's an option in Cloud called Schedule Tasks.
00:18:00.880 --> 00:18:04.799
So you can actually schedule tasks in there, which is like setting up an automation.
00:18:04.880 --> 00:18:05.039
Yeah.
00:18:05.200 --> 00:18:14.880
So I set up a simple automation that would send them a daily brief about, hey, based on your emails and all of the correspondences, these are the things that you need to be taken care of.
00:18:15.039 --> 00:18:16.720
This is your meeting minutes and so on.
00:18:17.039 --> 00:18:21.440
So earlier we were talking about how you could do it with the agents, like more advanced stuff.
00:18:21.599 --> 00:18:23.440
Now you can do it all in cloud.
00:18:23.920 --> 00:18:30.240
And so by setting it up right for the people, you remove those barriers and just becomes easier for people to use it.
00:18:30.640 --> 00:18:32.319
There's a reference in there to Jarvis.
00:18:32.400 --> 00:18:34.160
Um, you're talking about the Iron Man Javas.
00:18:34.400 --> 00:18:35.119
I Iron Man and Jarvis.
00:18:35.519 --> 00:18:42.240
Which makes me want to go back and watch Iron Man because I feel like I haven't done that in a while, and Javas is gonna feel really, really real now.
00:18:43.920 --> 00:18:48.960
Um What about for people who aren't using Claude?
00:18:49.359 --> 00:18:51.920
So, what about people who for business reasons?
00:18:52.000 --> 00:19:05.119
I mean, we spoke a lot, I think, in in our in the first, in the earlier series of the podcast, talking about Microsoft Copilot and the enterprise version of that, that a lot of businesses will roll out for reasons of protection and security, et cetera.
00:19:05.279 --> 00:19:06.480
That's what they license to.
00:19:06.559 --> 00:19:09.359
So that's their business information will then go through copilot.
00:19:10.880 --> 00:19:25.519
Are we in a space where these tools that Anthropic and Anthropic being the developers of Claude have developed and pushed forward into, what's the lag time on catching up to say a copilot or um GPT, GPT?
00:19:26.079 --> 00:19:26.400
Yeah.
00:19:26.720 --> 00:19:32.799
So copilot um has improved, but still trying to catch up at the moment.
00:19:33.200 --> 00:19:45.119
But for some of our, like I was having a very, I had a very interesting conversation with one of our accounting clients, and every they said uh everyone is asking for Claude, everyone's asking for tools, what should we do?
00:19:45.759 --> 00:19:50.880
The problem with Claude is that your data still travels to US and goes out of the boundary.
00:19:51.440 --> 00:19:55.759
Now, Copilot has introduced Claude inside Copilot as well.
00:19:56.079 --> 00:20:00.640
And there's a co-work function now that they are introducing in their frontier model.
00:20:00.880 --> 00:20:05.279
So frontiers is basically uh you have to sign up, it's like a beta system.
00:20:05.759 --> 00:20:13.599
So once your admin signs up to be a frontier, then you can get access to co-work inside the cloud, uh inside co-pilot.
00:20:13.680 --> 00:20:13.920
Yes.
00:20:14.160 --> 00:20:25.839
The challenge being is what Microsoft says is that it uh when whenever you use cloud, it doesn't apply the Australian data residency, uh residentiality.
00:20:26.160 --> 00:20:29.200
So what means is your data is still traveling to cloud servers.
00:20:29.759 --> 00:20:36.240
But one of the important questions that I raise is is data residentiality important for your business or not?
00:20:36.720 --> 00:20:42.319
Because for some businesses that are dealing with personal information, working with government, it's important.
00:20:42.480 --> 00:20:43.039
Yeah, absolutely.
00:20:43.279 --> 00:20:44.480
And that's that's fine.
00:20:44.720 --> 00:20:59.599
But there are some businesses who say it's important for me, but they're already using HubSpot, for example, or some other tools, uh, or make.com or all these tools where they are partially sending their data to US.
00:20:59.839 --> 00:21:02.000
So which means your data is already out there, right?
00:21:02.319 --> 00:21:08.640
So probably a good thing to do is is direct anyone who's listening to this that's wondering how do I work through these challenges.
00:21:08.799 --> 00:21:12.960
We did a podcast on governance, on AI governance, and that would be a really good one.
00:21:13.039 --> 00:21:22.240
So perhaps we can link that back in the chat where give people that connection just to go, okay, if this is an issue for your business and you're trying to explain to a team, because it's hard, right?
00:21:22.319 --> 00:21:35.200
I mean, you you sit down, you're you're running teams, and they're going, yep, no, okay, we're we're starting to use AI, this is good, um, but we're sort of stuck using Copilot, and frankly, it's a bit clunky now because it's you know half a generation or three months old.
00:21:35.599 --> 00:21:37.359
Um Claude can do all these things.
00:21:37.519 --> 00:21:42.240
I want to be able to use Claude, and you need to explain to them why that might be problematic.
00:21:42.480 --> 00:21:42.640
Yeah.
00:21:42.960 --> 00:21:44.319
Maybe it's okay, maybe it's not.
00:21:44.400 --> 00:21:45.279
But how do you make that decision?
00:21:45.440 --> 00:21:46.480
How do you make it informed?
00:21:47.839 --> 00:21:49.519
I think uh that's a great idea.
00:21:49.759 --> 00:21:58.319
So probably from an organization perspective, you come in three windows first, window is your data needs to stay in Australia, and that's uh no go.
00:21:58.559 --> 00:21:59.680
In that case, you use score.
00:22:00.160 --> 00:22:02.799
Copilot and use it without the clot features.
00:22:03.039 --> 00:22:10.720
The second option is you are concerned about security, but you are not obligated to have your data in Australia.
00:22:10.880 --> 00:22:14.240
You can use copilot and the frontier options of copilot.
00:22:14.319 --> 00:22:19.680
So like you can use co-work inside it because Claude is as is pretty secure.
00:22:19.839 --> 00:22:24.480
Like when it comes to security, it's not about the breach, it's just about the data that's going into their US server.
00:22:24.880 --> 00:22:33.359
I don't know about you, but I felt a lot better when I saw the CEO of Anthropic walking into the White House because they weren't just going to hand over everything that the government wanted straight away.
00:22:33.839 --> 00:22:34.400
That's right, yes.
00:22:34.799 --> 00:22:35.839
Kind of a bit of a trust test.
00:22:36.079 --> 00:22:36.880
It is, absolutely.
00:22:37.119 --> 00:22:39.759
And probably much more trusted than OpenAI.
00:22:40.079 --> 00:22:42.640
Mind you, that was a couple of months ago, so goodness knows what's happened since then.
00:22:43.279 --> 00:22:44.160
Yes, yeah.
00:22:44.400 --> 00:22:49.839
And uh also uh Anthropic had a very good conversation with the Australian government as well.
00:22:50.160 --> 00:22:51.359
They're opening an office in Sydney.
00:22:51.680 --> 00:22:52.720
Even that's fascinating, isn't it?
00:22:52.799 --> 00:22:59.119
I mean, we talked about, you know, we we opened up talking about that there's been a bit of time for us, but in in AI terms, things have shifted.
00:22:59.279 --> 00:23:03.039
That the environment, particularly around large language models, has shifted significantly.
00:23:03.200 --> 00:23:04.319
I mean, even that.
00:23:04.799 --> 00:23:19.039
That we we've had so many conversations, I think, as as people into the media, as um business people that are working in AI, the whole podcast community, the news community have all been talking about things like governments aren't catching up.
00:23:19.279 --> 00:23:20.559
Well, now they don't have a choice.
00:23:20.880 --> 00:23:21.200
That's right.
00:23:21.440 --> 00:23:22.240
They're moving really fast.
00:23:22.319 --> 00:23:24.319
I don't know that I've seen governments move this fast before.
00:23:24.400 --> 00:23:26.079
This is the story in and of itself.
00:23:26.480 --> 00:23:27.279
Yes, absolutely.
00:23:27.440 --> 00:23:27.680
Yes.
00:23:27.839 --> 00:23:28.720
I there's no choice.
00:23:28.960 --> 00:23:29.119
Yeah.
00:23:29.359 --> 00:23:29.759
Yes.
00:23:30.000 --> 00:23:39.519
And then the third version is if you there's no you are concerned about security, but you don't have an obligation to have your data in Australia.
00:23:39.680 --> 00:23:41.039
You can be a bit more flexible.
00:23:41.440 --> 00:23:50.720
So if you've got, say, uh IP protections, for instance, you're developing your manufacturing business or somewhere that has a significant amount of IP, that might be a concern for you?
00:23:51.119 --> 00:23:59.839
Uh so for example, if you're a manufacturing business who is developing something which is which is an important IP, but it's not like uh doesn't relate to national security.
00:24:02.240 --> 00:24:03.680
Holding data, holding personal data.
00:24:04.000 --> 00:24:04.640
Yes, even even though.
00:24:04.799 --> 00:24:05.519
What would be the use case?
00:24:05.839 --> 00:24:09.599
I would say holding personal data, again, what level of personal data it is.
00:24:09.759 --> 00:24:12.559
For example, if you're in healthcare, you want to stay with copilot.
00:24:12.720 --> 00:24:28.240
But if you are a manufacturing business uh who's uh doing recycled plastic manufacturing, right, you have an IP, but uh even if that stays with Claude, like Claude has already committed, it's not gonna use your data with the enterprise version and so on.
00:24:28.400 --> 00:24:30.480
So which means it's still on a cloud.
00:24:30.559 --> 00:24:36.079
If you it's still on co-pilot, although it's within your Microsoft version, it's still on a cloud, right?
00:24:36.160 --> 00:24:38.559
So that can get breached, this can get breached as well.
00:24:38.720 --> 00:24:40.400
There's always that that challenge.
00:24:40.559 --> 00:24:50.960
But the important thing is that the the commitment from Claude, the security control from Cloud actually allows a lot of small to medium businesses to be able to use it safely.
00:24:51.200 --> 00:24:57.119
Then for larger businesses, Cloud has got their enterprise version, which is again safe to use.
00:24:57.359 --> 00:25:00.400
But the challenge is if you are a university like Deakin, right?
00:25:00.559 --> 00:25:07.279
Coming back to your question, your question was uh like if the organization are using Copilot, they don't have this much of flexibility.
00:25:07.359 --> 00:25:08.559
Uh, what do you do?
00:25:08.720 --> 00:25:11.119
First of all, so copilot is actually moving forward.
00:25:11.200 --> 00:25:13.039
They are introducing cowork as well.
00:25:13.200 --> 00:25:15.279
They are introducing the concept of skills.
00:25:15.599 --> 00:25:26.079
Skills is a concept that you'll start to see in every AI platform because such a flexible concept, you don't need to have everything, you don't need to create like those assistive agents and so on.
00:25:26.240 --> 00:25:32.640
You can just write a skill for something, and then Claude or whatever it is can follow that script and it works out for everyone.
00:25:32.880 --> 00:25:36.880
So copilot would have a proper skills function really soon.
00:25:37.359 --> 00:25:52.240
But even if it doesn't have a skills function, what you can do is you can still write a file or you can still pick up a skill from the Claud, change it a bit, and then ask Copilot to actually follow that or create an agent in Copilot that actually uses that skill.
00:25:52.559 --> 00:25:59.279
So you can almost start to get the benefits that you would get from Claude in uh Copilot itself.
00:25:59.759 --> 00:26:05.279
The only benefits you won't be able to get is where you know how it connects to a folder and does a lot of work.
00:26:05.680 --> 00:26:12.799
Uh interestingly, Claude's integration with Microsoft is much better than the copilot's integration with the Microsoft stack.
00:26:13.039 --> 00:26:14.400
That's and that's happened really quickly.
00:26:14.720 --> 00:26:15.680
It has, yes.
00:26:16.000 --> 00:26:16.720
You know why?
00:26:16.960 --> 00:26:17.200
No.
00:26:17.759 --> 00:26:23.839
So Cloud focused on one thing, which was to make their models really good at uh coding.
00:26:23.920 --> 00:26:32.960
Uh and so what happens is that Claude models or Opus really knows how to write, think through a problem and write a good code.
00:26:33.200 --> 00:26:37.119
The the thing is that everything that happens on a computer uses code.
00:26:37.759 --> 00:26:43.200
What it means is that for it to create a Word document, it doesn't need to know how to use Microsoft Word.
00:26:43.440 --> 00:26:48.000
It needs to know how to write scripts that can actually execute that Word document.
00:26:48.240 --> 00:26:56.079
So because it is so good at coding, it can interpret the code of any software it's working in, and then it can actually make it happen.
00:26:56.319 --> 00:26:58.720
Uh whereas like open AI is clear.
00:26:58.880 --> 00:26:59.440
It's really clever.
00:26:59.599 --> 00:27:02.400
So it's it's effectively giving you access to the foundations.
00:27:02.640 --> 00:27:02.799
Yes.
00:27:03.119 --> 00:27:03.759
What's driving this?
00:27:04.079 --> 00:27:04.640
That's right.
00:27:04.880 --> 00:27:08.079
So like the two big shifts that has happened in the AI space.
00:27:08.240 --> 00:27:13.279
First is that something non-human technology was able to understand human language.
00:27:13.440 --> 00:27:21.599
So it was, it made it really easy for us to teach it anything or train it on this large language.
00:27:21.759 --> 00:27:27.039
And the second thing is for it the ability for it to be able to write code or understand code.
00:27:27.200 --> 00:27:30.720
Because what it means is that it can now do anything on the computer as well.
00:27:31.119 --> 00:27:37.279
So it can now understand human language really well, it can do anything on the computer really well, and that just closes the gap.
00:27:37.680 --> 00:27:38.000
Yeah.
00:27:38.240 --> 00:27:42.079
It's it has been uh fascinating to see that.
00:27:42.160 --> 00:27:48.799
Have has the conversation around safety of these AI models and the security of these AI models changed from your perspective?
00:27:48.960 --> 00:27:56.960
Because looking from the outside in, um you know, for for those of you that are joining me on the on the dumb monkey side of things.
00:27:58.000 --> 00:27:59.839
I think it's an interesting conversation though.
00:27:59.920 --> 00:28:03.519
Like we've it's uh it's been a long time since we have done this.
00:28:04.559 --> 00:28:12.400
It is and the it It does feel like a lot of the a lot of the issues that we were particularly concerned, say this time last year, have changed.
00:28:13.200 --> 00:28:31.839
So we would previously, if we were doing this, if we were doing this podcast, having this conversation six months ago, we would have been, or eight months ago, we would have been talking about um concerns around the security and uh protecting basically your use of some of these models, particularly in a business sense, but really in a life sense.
00:28:32.000 --> 00:28:36.880
Um if you're using the free version of ChatGPT, for instance, be be concerned about what you're putting on there.
00:28:37.200 --> 00:28:39.279
A lot of our conversations were about that.
00:28:39.519 --> 00:28:56.240
It feels like there's been an awareness perhaps from the developers of these models and the and the teams that are driving the development of them that if you want our if you want our money, if you which is you know, you are the one I use of the of the platform, if it's free.
00:28:56.319 --> 00:29:00.160
You know, the golden rule is if it's free, if you you're not paying for the product, you are the product.
00:29:00.319 --> 00:29:04.640
Um so they either want you using the product or they want you paying for the product and they want you to keep coming back.
00:29:04.720 --> 00:29:14.319
These are business models and business models with some very big, very big costs and bank behind them, some incredibly high um business appraisals.
00:29:14.400 --> 00:29:19.599
You know, what we're saying, these businesses that work, they're piling money into big data centers, they need us to keep being involved with them.
00:29:19.920 --> 00:29:20.319
That's right.
00:29:20.559 --> 00:29:27.920
So, as a business decision from them, it makes very good sense to make us as users feel more secure about using these models.
00:29:28.079 --> 00:29:28.240
Yes.
00:29:28.480 --> 00:29:32.240
That's what I think we're all saying the same things, going, how do we trust it?
00:29:32.400 --> 00:29:32.960
Yeah.
00:29:33.279 --> 00:29:37.119
Has there been real movement on that, or do I just feel like there's been movement on that?
00:29:37.680 --> 00:29:52.160
I think what has happened is that uh because uh a lot of people are now aware of the privacy settings, uh, and initially there were the mod like the systems have improved as well.
00:29:52.240 --> 00:29:59.119
Like Claude has got a SOC2 certification, these guys needed to move into enterprises, so the questions were asked.
00:29:59.200 --> 00:30:05.440
So definitely organizations themselves have become more mature in terms of how they handle the data.
00:30:05.759 --> 00:30:10.799
People now understand the difference between a free version and uh in a pro version as well.
00:30:10.880 --> 00:30:15.680
So I think there's definitely been a moment in that from that perspective.
00:30:16.000 --> 00:30:19.519
Uh, and also these models have become more safer as well.
00:30:19.680 --> 00:30:28.880
Like if you could really do a prompt injection in like earlier models and uh get information out, these models are becoming more and more safer uh as well.
00:30:29.119 --> 00:30:42.799
So there's definitely, and I think uh also I in Australia what has happened is that because a lot of businesses have been conscious, so they have spent a lot of uh efforts in doing the governance right and policies right.
00:30:42.960 --> 00:30:49.599
I've seen the businesses where they have actually made sure that uh people are aware of the risk and challenges and so on.
00:30:50.000 --> 00:30:55.039
And now so we are in a position to actually start to leverage some of these tools easily.
00:30:55.680 --> 00:30:56.319
Yeah.
00:30:58.400 --> 00:30:59.599
It is interesting.
00:30:59.920 --> 00:31:03.119
Um what do you think is coming next?
00:31:04.079 --> 00:31:06.079
So what do you know is coming next?
00:31:06.480 --> 00:31:12.079
Okay, so we can talk about features, like there's been so many features all the time.
00:31:13.200 --> 00:31:24.720
What has happened is now it's actually shifted the whole like last year the conversation was okay, like this chat GPD is great, it provides good answers, not, and so on.
00:31:25.359 --> 00:31:31.920
But we are saying okay, agents are coming, or some like technical people would have those agents.
00:31:32.319 --> 00:31:37.279
But now the systems themselves have got agentic capabilities, right?
00:31:37.519 --> 00:31:48.559
So what it means is that a lot of work that we were actually doing by ourselves is can be delegated to these systems.
00:31:48.799 --> 00:32:01.759
And so as we go, and then Cloud has been uh launching Cloud for Legal, Cloud for They have just recently launched Cloud for Small Business that gives small business a lot of tools to connect to their systems.
00:32:01.920 --> 00:32:08.319
Uh so what happens is that uh these tools uh are becoming really powerful.
00:32:08.559 --> 00:32:16.799
Uh so first of all, there's a very clear advantage for the businesses who are embracing it because they're able to do the work really fast.
00:32:16.880 --> 00:32:19.680
So there's definitely that's divide forming.
00:32:19.920 --> 00:32:32.400
And then also in terms of the people who are embracing AI versus who are just using it for uh correcting their emails and so on, that gap has started to widen a bit as well.
00:32:32.720 --> 00:32:33.359
Yeah.
00:32:35.279 --> 00:32:37.039
Has the cost of using AI gone up?
00:32:37.200 --> 00:32:41.920
And are we paying enough attention to what's going to happen to our budget models?
00:32:42.000 --> 00:32:48.880
Because as you mentioned, you know, a lot of these models are fairly heavily discounting the availability of these tools.
00:32:49.039 --> 00:32:51.119
I mean, this is a classic business model.
00:32:51.599 --> 00:33:00.640
We've developed these incredible things, we've poured an enormous amount of resource into it, we've now got a you know, we've got a value, our company valuation is over here, our cost base is here.
00:33:01.119 --> 00:33:09.839
Anyone who's in the the stock market sector economists, um, financial futurists will all sit there and go, none of these companies are worth the valuations that they're worth.
00:33:10.000 --> 00:33:14.799
Their cost of going into development has to be recouped at some point.
00:33:15.119 --> 00:33:15.680
Yes.
00:33:16.160 --> 00:33:18.559
Are we all about to hit a big AI bill?
00:33:18.720 --> 00:33:19.839
Do we need to have this conversation?
00:33:20.240 --> 00:33:25.200
I think it's a very important conversation and also deserves its own space as well.
00:33:25.599 --> 00:33:27.599
Uh but it's a very great point.
00:33:27.759 --> 00:33:35.200
Uh, one thing uh to understand here is that a lot of subscriptions that we are getting, these are being heavily subsidized.
00:33:35.359 --> 00:33:45.680
So, like the usage that we get in a$30 Claude or Chat GPT subscription, it generally would cost uh them uh$100 or more than that.
00:33:45.920 --> 00:33:53.839
So similarly, what's what they are doing is they are actually heavily subsidizing that cost with the money that they are raising because they want more and more people to use AI.
00:33:54.000 --> 00:34:02.000
So the actual cost of tokens has not decreased, uh, and there's not been much work that has been done on that part.
00:34:02.079 --> 00:34:04.400
There have been some other models that have come out.
00:34:05.119 --> 00:34:08.719
And by tokens, you mean effectively the amount that you pay that gives you the amount of computing.
00:34:09.119 --> 00:34:09.360
Yes.
00:34:09.519 --> 00:34:15.599
So token is basically uh the the currency in which you measure the usage of AI.
00:34:16.079 --> 00:34:19.039
So one token is generally two words.
00:34:19.360 --> 00:34:22.480
Uh so what happens is really I never knew that.
00:34:22.719 --> 00:34:22.880
Yes.
00:34:23.199 --> 00:34:23.679
That's a token.
00:34:24.079 --> 00:34:26.880
Yes, and I'm saying that nervously because I always get confused.
00:34:27.119 --> 00:34:33.519
Either it's one word is two tokens or two tokens is one word, but it's very, very close to that.
00:34:33.679 --> 00:34:34.159
Yes.
00:34:34.480 --> 00:34:44.320
But uh what happens is that every time you send a message to AI or receive something, that's actually uh calculated in terms of how many tokens there are.
00:34:44.639 --> 00:34:44.800
Okay.
00:34:45.119 --> 00:34:52.159
And so no matter whether you are paying for token or not, the company that's processing those tokens have to pay for it.
00:34:52.400 --> 00:34:54.960
There's a compute that's actually happening over there.
00:34:55.440 --> 00:34:59.599
So we're gonna take that on night so I can that that will be a really interesting conversation.
00:34:59.840 --> 00:35:00.079
Yes.
00:35:00.239 --> 00:35:06.719
So it's it's the bill, the AI bill is definitely be there, and we have not solved the power of the energy that's actually required.
00:35:07.199 --> 00:35:09.360
Water use and compute and all of that.
00:35:10.480 --> 00:35:16.079
But there should be some people who should be working on it, and there's a this this is a problem to be solved as well.
00:35:16.320 --> 00:35:16.639
Yeah.
00:35:16.880 --> 00:35:25.280
Uh but it should not uh defer you as a business person or as an individual from actually using AI.
00:35:25.440 --> 00:35:37.519
Yes, the the cost might increase, but now is the time for you to use AI as a playground where you can experiment, learn, and fail and like burn tokens because they are being heavily subsidized.
00:35:37.840 --> 00:35:43.039
Because probably in one year the that$20 subscription might be$500.
00:35:43.440 --> 00:35:43.920
Oh wow.
00:35:44.000 --> 00:35:44.159
Yeah.
00:35:44.400 --> 00:35:49.039
But the thing is, by that time, we should have sorted out like what's the wastage.
00:35:49.199 --> 00:35:51.280
Like right now, we do a lot of wastage.
00:35:51.440 --> 00:36:00.320
Not in normal conversation, are generally cheaper, but when you're developing apps and developing systems, designing things, you burn like thousands of tokens.
00:36:00.559 --> 00:36:05.119
Sorry, not even thousands, like everyone burned a thousand tokens in the thing.
00:36:05.199 --> 00:36:09.519
Like we're we're talking sort of like millions of tokens and sort of developing something like that.
00:36:09.679 --> 00:36:11.440
And I I burn a lot of tokens.
00:36:11.920 --> 00:36:16.719
Um, one of one month my bill of AI was around fifteen thousand dollars.
00:36:16.960 --> 00:36:17.440
Wow.
00:36:17.679 --> 00:36:23.039
Because I just and uh all of the APIs, uh consumption, and so on.
00:36:23.280 --> 00:36:23.519
Yeah.
00:36:23.679 --> 00:36:23.920
Yes.
00:36:24.159 --> 00:36:28.960
I've got a little image in my brain of you just sitting there just you know, like your little AI fire just burning money.
00:36:30.400 --> 00:36:31.840
But learning and developing volume.
00:36:32.239 --> 00:36:42.159
Learning and developing, and like uh I was able to develop uh an app on on a weekend and like two two nights from that and spend and how long would that app have taken you previously?
00:36:42.480 --> 00:36:44.239
I asked my team for an estimate.
00:36:44.480 --> 00:36:45.760
I've got a very skilled team.
00:36:45.920 --> 00:36:47.519
They said three and a half months.
00:36:47.840 --> 00:36:48.159
Wow.
00:36:48.320 --> 00:36:48.559
Yes.
00:36:48.880 --> 00:36:50.559
Yeah, so time is money, right?
00:36:50.880 --> 00:36:52.239
There's there's equations in there as well.
00:36:52.480 --> 00:36:52.639
Yes.
00:36:52.880 --> 00:36:53.599
What about Google?
00:36:53.840 --> 00:37:03.599
Google, I I'm always interested by um mainly because I come from that sort of you know media marketing background, and they call themselves Google Deep Mind.
00:37:03.679 --> 00:37:07.679
They've got an incredible origin story, and they just have the best marketing in AI.
00:37:07.840 --> 00:37:09.280
So, you know, I was already a fan.
00:37:09.360 --> 00:37:13.199
But no, in but but in all seriousness, Google I find is really interesting.
00:37:13.280 --> 00:37:17.599
How they position themselves, what they're developing in terms of their AI, their origin story.
00:37:17.679 --> 00:37:21.440
For anyone who hasn't seen the Google Deep Mind documentary, go and watch it.
00:37:21.519 --> 00:37:23.599
It's just it's amazing.
00:37:23.760 --> 00:37:25.599
And it's just such a story of our time.
00:37:25.840 --> 00:37:28.000
Like it's going to become one of those iconic stories.
00:37:29.039 --> 00:37:37.519
But they're interesting because Google were early providers of AI, they were early developers of AI.
00:37:37.679 --> 00:37:42.480
Then they kind of slipped behind, they lost a fair bit of you know, market share, that became kind of uncool.
00:37:42.719 --> 00:37:45.519
You know, you didn't really use Google for AI anymore.
00:37:45.679 --> 00:37:48.719
Um then they did the wonders of nano banana, best name in AI.
00:37:48.800 --> 00:37:48.960
Still.
00:37:49.679 --> 00:37:51.039
It's gonna be a hard one to top.
00:37:51.199 --> 00:37:53.760
Um but of course the image generation with nano banana.
00:37:53.920 --> 00:37:55.840
But now they're doing something really interesting.
00:37:56.079 --> 00:38:04.320
And I've noticed that Google AI is starting to get integrated into a little bit of everything and the the everyday things that we use.
00:38:04.480 --> 00:38:15.119
So we've spent a fair bit of conversation talking about Claude and the and the magic that they've created, and it absolutely is magic because I don't understand how it how it works, so I'm calling it magic.
00:38:15.360 --> 00:38:18.320
Um so magic is just signs that you don't understand.
00:38:18.719 --> 00:38:31.440
And but Google is putting their AI into Gmail, into Google Maps, into all of these things that are just kind of ubiquitous.
00:38:31.599 --> 00:38:33.760
We don't talk about them, we don't think about them anymore.
00:38:33.920 --> 00:38:44.480
Um that grand old thing that I'm gonna steal a line from Professor Professor Raj Basser at um at Deacon that I work with, and not sponsored by Deacon, completely unaffiliated.
00:38:44.639 --> 00:38:45.599
My work here is separate.
00:38:46.239 --> 00:38:47.199
But we love Deacon.
00:38:47.679 --> 00:38:48.480
But we love Deacon.
00:38:48.719 --> 00:38:48.960
Yes.
00:38:49.440 --> 00:38:50.639
Um always.
00:38:50.800 --> 00:38:55.920
But Raj has this great line, and he says if you notice if you notice the technology, it probably means it's broken.
00:38:57.039 --> 00:39:01.199
And with Google, it's become so ubiquitous that we don't notice that technology anymore.
00:39:01.280 --> 00:39:01.840
It's just kind of there.
00:39:01.920 --> 00:39:05.599
It's on our phones, it's on our laptops, it's in our, you know, in our area.
00:39:06.239 --> 00:39:10.320
And now all of a sudden we've got Google AI coming into these.
00:39:10.559 --> 00:39:11.440
What are they doing?
00:39:11.679 --> 00:39:12.400
What are you seeing?
00:39:12.960 --> 00:39:22.639
So if you go back to the history of how Gen AI came about, there were a few guys who were trying to solve a problem with Google Translate.
00:39:22.800 --> 00:39:28.400
The problem with Google Translate was it was able to translate word to word, but could not connect the sentences.
00:39:28.719 --> 00:39:44.320
So they then did research and wrote a paper on transformers about how rather than you have to, rather than having reading word to word, uh, you can actually read the full document, pick up the words, and then predict what is around that.
00:39:44.800 --> 00:39:57.280
And then that actually changed the the whole thought process of okay, right now, if you had to train something on uh on a model, what you have to do is you have to go through word to word, it would take a lot of time.
00:39:57.440 --> 00:40:00.719
So OpenAI, what they did was they said, okay, this is really good.
00:40:00.880 --> 00:40:05.599
We are gonna use this transformer model, and because Google published the paper, right?
00:40:05.679 --> 00:40:06.719
So they just put it out.
00:40:07.039 --> 00:40:14.559
And they said, we are going to train it all on the all of the the whole world's knowledge, and that's how the large language model concept was born.
00:40:14.800 --> 00:40:23.039
Then they said, okay, if we can use prediction to be able to understand what's in the text, we can use it to predict the next term as well.
00:40:23.280 --> 00:40:31.679
So coming back, it was Google who actually came up with the actual concept of transformers from where the whole LLM was was generated.
00:40:31.920 --> 00:40:43.599
What Google has is an ecosystem of the penetration in the devices, the systems that we use, uh, and it's it's already out there.
00:40:43.840 --> 00:40:46.079
Now we are talking about a cloud over here.
00:40:46.239 --> 00:40:54.159
The beauty of the cloud is not in the model itself, like because models are gonna come and go, it's what they have built around it.
00:40:54.320 --> 00:40:55.440
It's a toolkit.
00:40:55.599 --> 00:41:00.960
But the thing is that right now, Google, and we're gonna we should talk about Apple at some stage as well.
00:41:01.199 --> 00:41:09.519
They're just sitting there laughing or smiling because the features that Claude has built, they can actually be built by their developers in a couple of weeks.
00:41:09.760 --> 00:41:14.079
Because uh Claude has actually developed some of those features in weeks as well, with with with the clouds.
00:41:14.239 --> 00:41:15.679
I'm starting to wonder what I do with my dice.
00:41:15.920 --> 00:41:16.480
Yeah.
00:41:17.119 --> 00:41:28.000
So so they are they're just watching uh this space from the consumer perspective and sort of like seeing what's happening, what's working for Claude, and what's not working for Claude.
00:41:28.079 --> 00:41:31.119
And Claude is trying to establish its presence in the market.
00:41:31.360 --> 00:41:35.679
And I believe Claude would have its own fanship and so on.
00:41:35.920 --> 00:41:41.119
But because of Google's own ecosystem, it still owns the search space.
00:41:41.280 --> 00:41:44.159
They've started to use the AI mode in there as well.
00:41:44.400 --> 00:41:54.800
Like they have they they own the Google Home space, there's a whole Android space that's out there, and if you look at the level of AI in uh Android phones, uh that's just amazing.
00:41:54.880 --> 00:41:57.280
It's gonna become a more native feature as well.
00:41:57.440 --> 00:42:02.400
And it's still a very prominent device that we actually use uh in in this.
00:42:02.559 --> 00:42:17.199
Whereas Apple has taken a stance where because they provide a very definitive experience to the people, the whole AI they feel is is in a bit of like still a sandbox mode where we're still going through testing and iterations.
00:42:17.519 --> 00:42:26.559
Did I read somewhere that that Google have made a deal with Apple though, that that future iterations of the iPhone are going to be using Google AI?
00:42:26.960 --> 00:42:27.760
I'm not aware of that.
00:42:28.719 --> 00:42:29.440
Go and have a check about that.
00:42:29.599 --> 00:42:30.719
You'll you'll read more into it than I will.
00:42:30.960 --> 00:42:31.119
Right.
00:42:31.519 --> 00:42:36.559
But I I believe in this was in the context of the money.
00:42:36.719 --> 00:42:37.760
Yeah, chasing the money.
00:42:38.079 --> 00:42:49.119
Which is um, my background for for quite a number of years is how we met was in business journalism, and that's that's the secret source of business journalism.
00:42:49.360 --> 00:42:51.039
There's no great secret to it, you just chase the money.
00:42:51.360 --> 00:42:54.480
And that that tells you kind of where people's decision making will go.
00:42:55.119 --> 00:43:00.400
And so I was really interested to see what was happening because you have these huge valuations, huge cost base.
00:43:01.519 --> 00:43:11.920
How is a company that's you know, I'm I'm making up figures, but how is a company that's bringing in$100 billion valued at three trillion?
00:43:12.719 --> 00:43:19.360
You know, we these things are are real when they need to be in in the IO space they haven't been.
00:43:19.920 --> 00:43:22.079
So there will there will be a reality check.
00:43:22.159 --> 00:43:23.119
We know that.
00:43:23.360 --> 00:43:25.679
But but will there be a reality check?
00:43:25.920 --> 00:43:30.000
Would it wouldn't you think that market would adjust itself that and these organizations will be able to do that?
00:43:30.400 --> 00:43:34.239
Maybe maybe the productivity will come up underneath it to support that, but yeah.
00:43:35.440 --> 00:43:59.360
There's gonna need to be some big changes in terms of value and where the any of this organization actually f reaching to a superagent stage where agents are able to create agents themselves and they are able to do months of research in in in seconds or hours, the amount of uh value that these organizations would hold would be amazing.
00:44:00.239 --> 00:44:04.639
So they these organizations would probably start to engulf other organizations.
00:44:04.719 --> 00:44:09.440
Like OpenAI have already started to sort of like offer their consulting services.
00:44:09.679 --> 00:44:21.199
And when I when we were joking about Claude killing the services, every time Claude launches a feature, uh a valuation of a company that's providing that service goes down.
00:44:21.440 --> 00:44:25.519
So we have we have seen like Canva's valuation or shares going down.
00:44:25.599 --> 00:44:27.679
We have seen Figmas shares going down.
00:44:27.840 --> 00:44:32.320
So all of these organizations that build their business on providing a service.
00:44:32.639 --> 00:44:33.760
And being a market leader in that.
00:44:34.079 --> 00:44:34.880
Being a market leader.
00:44:35.119 --> 00:44:43.599
And Cloud's developer would wipe code something because they use AI to decode, yes, and they they would do something in a week and launch a new feature.
00:44:43.920 --> 00:44:46.239
And then that just disrupts the whole industry.
00:44:46.400 --> 00:44:46.639
Yeah.
00:44:47.039 --> 00:45:01.199
So you can say that the valuation is inflated, but uh I feel the the the disruption that's gonna cause to industry uh is is gonna be so huge that that valuation would become meaningful.
00:45:01.519 --> 00:45:02.159
Yeah, okay.
00:45:02.639 --> 00:45:08.480
It could be the case that some of these companies might die and and and and and there would be like a few winners out of this.
00:45:08.800 --> 00:45:18.079
And new ones will be reborn and that's as you would expect in the in the in the early in the early days of evolution, things are you know, things get messy.
00:45:18.559 --> 00:45:18.800
That's right.
00:45:19.199 --> 00:45:23.280
Or kind of explosions, and then at the end of it you you reach a some form of equilibrium.
00:45:23.519 --> 00:45:27.360
Um I was very interested to see that Google's ad revenue is rising.
00:45:27.920 --> 00:45:31.280
It would because uh search value is going down as well.
00:45:31.599 --> 00:45:31.760
Yeah.
00:45:32.000 --> 00:45:32.639
Yes, yeah.
00:45:32.719 --> 00:45:35.360
And Chat GPT has introduced ads as as well.
00:45:35.599 --> 00:45:35.840
Yeah.
00:45:36.079 --> 00:45:36.719
And and there.
00:45:37.199 --> 00:45:41.920
No, Google's Google's ad revenue is is rising, so they're they're making more money.
00:45:42.159 --> 00:45:42.400
Yes.
00:45:42.559 --> 00:45:43.920
I kind of thought that they'd be losing money.
00:45:44.239 --> 00:45:52.239
No, they they are making more money uh because uh first of all, the the the search is actually uh decreasing.
00:45:52.480 --> 00:46:00.480
So what's what's what's happening is that uh businesses are more compelled to actually do ads as well.
00:46:01.039 --> 00:46:06.079
And also what we are seeing at the moment is we haven't seen the actual shift.
00:46:06.320 --> 00:46:21.119
Like I still use Google for search because I'd talked about using perplexity, but then I just got like it was too much because sometimes I just wanted to get an answer or get get get get to a website, and perplexity experience was not that great, right?
00:46:21.199 --> 00:46:23.599
So I still use Google for a lot of my searches.
00:46:24.239 --> 00:46:29.519
You haven't we haven't seen that shift from people moving away from Google to actually using the stuff.
00:46:29.679 --> 00:46:37.840
Yeah, but secondly, Google's ad revenue doesn't only come from the search revenue, it comes from the placement that happens in their partner websites and in the ecosystem.
00:46:38.000 --> 00:46:38.159
Yeah.
00:46:38.400 --> 00:46:40.320
And Google's ecosystem is pretty much.
00:46:40.480 --> 00:46:40.800
It's huge.
00:46:40.880 --> 00:46:42.000
It's really ubiquitous, isn't it?
00:46:42.239 --> 00:46:42.320
Yeah.
00:46:42.559 --> 00:46:42.800
Yeah.
00:46:42.880 --> 00:46:43.039
Okay.
00:46:43.119 --> 00:46:43.920
Well these are interesting.
00:46:44.400 --> 00:46:48.320
Why do you think that uh ad revenue is increasing from your perspective?
00:46:48.719 --> 00:46:51.280
Well, do I think it's I think it's I think it's similar.
00:46:51.440 --> 00:47:03.599
I think it's a case of of the the driver to say if you're in a business and you're trying to get people's attention, then this is a way to actually get value out of it.
00:47:03.840 --> 00:47:08.480
Because I think for a period of time doing things like Google Ads was a bit like a bit of a gamble.
00:47:09.360 --> 00:47:18.079
So you can you can put your your Google ads in there, you'll get sponsored up, people would completely kind of ignore the sponsored ads, go past.
00:47:18.320 --> 00:47:23.199
So people don't particularly like as people we don't like, ads that tell us what to do.
00:47:23.360 --> 00:47:29.440
I think as consumers and as searches, I think we just look straight at the the sponsored things and go straight past them.
00:47:29.599 --> 00:47:33.119
But I think there's more value now in where it is.
00:47:33.280 --> 00:47:35.039
And I think the advertising is getting a lot smarter.
00:47:35.119 --> 00:47:40.239
I mean, AI algorithms mean that you can start to personalize and so I've connected my Google ads to my cloud as well.
00:47:40.559 --> 00:47:42.239
Claude doesn't have a direct connector.
00:47:42.400 --> 00:47:49.679
There's a thing called Composio through which you can connect your like it's like a it's like uh standard MCP protocol.
00:47:49.920 --> 00:47:51.280
So I've connected my Google Ads.
00:47:51.679 --> 00:47:54.400
Sorry, what's a standard MCP protocol?
00:47:54.559 --> 00:47:59.760
What's which is uh an easy way to connect with these systems.
00:47:59.920 --> 00:48:06.639
So MCP model context protocol is a way to connect your cloud with any third-party systems.
00:48:07.039 --> 00:48:16.239
And uh some if you go to connectors in Cloud, browse the marketplace, a lot of connectors are available, but there are some connectors that are not available.
00:48:16.400 --> 00:48:18.559
But there's a Composio connector available.
00:48:18.639 --> 00:48:24.719
Then you go to Composio website, composio.dev, and then you can connect multiple Google accounts.
00:48:24.960 --> 00:48:28.159
I'm totally not doing this, but I'm interested in meta accounts and so on.
00:48:28.719 --> 00:48:32.159
And then you can just say, hey, uh Claude, look at my campaign.
00:48:32.239 --> 00:48:33.599
What am I doing wrong here?
00:48:33.840 --> 00:48:36.079
You do you want to set up a new campaign for me?
00:48:36.159 --> 00:48:38.000
And Claude can set up the campaign for you.
00:48:38.480 --> 00:48:41.280
Claude can set up the Facebook ads for you as well.
00:48:41.519 --> 00:48:55.199
It's much better than actually uh, you know, dealing with a third party to actually run not in every case, but no, but it it you know, I as you you asked the question about why they're getting better.
00:48:55.280 --> 00:48:56.960
I think I think their offering is getting better.
00:48:57.119 --> 00:48:58.559
I think that's why they're making more money.
00:48:58.800 --> 00:49:06.800
So I think that you know, back when it would be if you were searching for a a pet groomer, um, I'm not searching for a pet groomer.
00:49:06.880 --> 00:49:09.039
I've got a French bulldog, no one wants to groom that.
00:49:10.480 --> 00:49:17.679
Um but you'd put in an ad for pet groomer and it can take you to somewhere in Perth or somewhere in WA, or it it's not very useful.
00:49:17.760 --> 00:49:23.760
If I'm sitting here in in slightly cloudy, foggy Geelong, I want a pet groomer that's just down the road.
00:49:23.840 --> 00:49:25.920
And ideally I want them to be even closer than that.
00:49:26.400 --> 00:49:36.000
So for all of the wonders of being able to provide the world of options, I think for most of us as consumers, when we're looking for something, we kind of want to find something that's really practical for us right now.
00:49:36.239 --> 00:49:38.079
And we want to do that in the easiest possible way.
00:49:38.320 --> 00:49:53.119
So as advertisers, uh as an advertising platform that I think you know, in a way Google is, um the better they are at providing people what they want up front makes the offer to advertise on them a little bit better.
00:49:53.199 --> 00:50:00.960
And I assume if you can pinpoint where you're trying to target your advertising, if you're not trying to hit the whole world and you're trying to target a smaller area, you're probably spending less money.
00:50:01.280 --> 00:50:01.519
100%.
00:50:02.159 --> 00:50:03.840
So it becomes more accessible as well.
00:50:04.000 --> 00:50:07.039
Um and that's something that I've seen in with meta ads as well.
00:50:07.199 --> 00:50:12.559
Earlier in meta ads, you have to do a lot of thinking about who do you want to target and so on.
00:50:12.719 --> 00:50:26.159
Now, official recommendation from Meta is you just put what is your offering, work on your creative and the offer, and leave it on Meta to determine the right people who are going to be receiving the message.
00:50:26.400 --> 00:50:28.320
And Meta gets it so well.
00:50:28.719 --> 00:50:33.920
Like you don't have to really do a lot of that targeting because MetaZI can do that.
00:50:34.320 --> 00:50:37.840
So I think with the help of AI, these advertising tools are getting better as well.
00:50:38.880 --> 00:50:42.960
So we asked you the question of what do you think is is coming next.
00:50:43.280 --> 00:50:44.719
I want to frame that in a different way.
00:50:44.800 --> 00:50:49.280
It's can we really know if what's coming next?
00:50:50.079 --> 00:50:51.679
Because these changes come hard and fast.
00:50:51.760 --> 00:50:58.239
I mean, Claude felt like I I know it wasn't, but it felt like it came out of nowhere and became a dominator really quickly.
00:50:58.719 --> 00:51:01.360
Um could it be Google in a couple of months?
00:51:01.440 --> 00:51:03.360
Are they cooking up something and that can happen?
00:51:03.599 --> 00:51:05.599
Can open AI make a resurgence?
00:51:05.920 --> 00:51:10.079
Can there be a new player in the field that's gonna step up and start to disrupt everything?
00:51:10.400 --> 00:51:11.760
Can we really know what's coming?
00:51:12.000 --> 00:51:12.800
That's a good question.
00:51:12.960 --> 00:51:20.559
So we cannot really know for sure what's coming, especially with uh what models are gonna come and what advancements are gonna come.
00:51:20.719 --> 00:51:34.800
But we can definitely make some hypothesis very confidently about where the world is going and some things that we are we are sure can say what's gonna happen or what's not gonna happen.
00:51:34.960 --> 00:51:39.840
So the first thing is that we are not ever going to go back to a world without AI.
00:51:40.079 --> 00:51:46.159
So it's it's not it's not a it's not a hype, it's not toothpaste is out of the tube, you know, yes.
00:51:46.239 --> 00:51:49.840
It's it's it's not uh bubble, it's not going to burst.
00:51:50.000 --> 00:51:53.119
No, it's valuations can decrease or so on.
00:51:53.360 --> 00:51:56.239
It's uh something that is here to stay.
00:51:57.199 --> 00:51:59.599
And outside of the tech industry, do people really care?
00:52:00.159 --> 00:52:05.840
I mean, do people care if the AI is open AI, if it's anthropic, if it's Google?
00:52:07.119 --> 00:52:07.920
They don't care.
00:52:08.079 --> 00:52:12.559
Like as long as uh they get their work done, that's what what they're uh worried about.
00:52:12.800 --> 00:52:19.760
And the second reality we know is that the business, as we knew earlier, it has shifted, it has changed.
00:52:19.920 --> 00:52:24.079
I the conversation that I have with businesses everywhere now.
00:52:24.239 --> 00:52:28.239
Like last year I was saying Australian businesses are sleeping, they are not sleeping anymore.
00:52:28.400 --> 00:52:37.440
They're awake, they're awake, they they know what's happening, and the the way we do business is changing and it's gonna change forever.
00:52:37.599 --> 00:52:42.400
We we know that we are seeing job marketing change, job the whole job market changing.
00:52:42.719 --> 00:52:48.639
The question is about when it's gonna happen, like and to what extent?
00:52:48.800 --> 00:52:52.480
We know there's gonna be a displacement, where it's gonna be a change.
00:52:52.719 --> 00:52:53.519
How much?
00:52:54.000 --> 00:53:00.320
Like, we know there's gonna be some business that will have an edge, some businesses would not have that edge because of the gap.
00:53:00.480 --> 00:53:01.199
But how much?
00:53:01.360 --> 00:53:03.599
Like, would those businesses be able to survive?
00:53:03.840 --> 00:53:07.440
Would we have some very prominent businesses and that's it?
00:53:07.760 --> 00:53:09.519
What's gonna happen to our jobs?
00:53:10.000 --> 00:53:19.599
Because we are just we like right now with the knowledge that we have, we make assumptions, okay, it's it's just about the knowledge workers and we still be able to delegate jobs and so on.
00:53:19.840 --> 00:53:21.760
But what does that displacement look like?
00:53:21.920 --> 00:53:29.920
What yes, and and that's a question that also keeps me awake uh like every day is like what is the limit?
00:53:30.079 --> 00:53:30.960
Where are we gonna stop from?
00:53:31.280 --> 00:53:33.199
How do you have extra time to be awake?
00:53:35.920 --> 00:53:38.400
Uh I'm I'm awake all the time nowadays.
00:53:38.559 --> 00:53:39.360
Yes, yeah, yes.
00:53:40.079 --> 00:53:44.719
Just sending sending things over to sending to my my Claude team.
00:53:44.960 --> 00:53:45.039
Yeah.
00:53:45.280 --> 00:53:45.840
Yes, yeah.
00:53:46.159 --> 00:53:50.000
Or possibly just, you know, spamming your mother-in-law.
00:53:50.320 --> 00:53:50.960
Yeah.
00:53:52.559 --> 00:53:56.880
But I hope at some point you are throwing in a few joke messages.
00:53:57.119 --> 00:53:58.000
I think that's important.
00:53:58.079 --> 00:53:58.960
Don't let that moment go.
00:53:59.119 --> 00:53:59.440
Uh yes.
00:53:59.840 --> 00:54:02.239
Because what we can't do in the age of AI is lose our humanity.
00:54:02.480 --> 00:54:03.360
I know, yes, yes.
00:54:03.599 --> 00:54:08.639
And a bit of ragging of your mother-in-law, that's that kind of feels like you've set that up very nicely.
00:54:10.079 --> 00:54:10.960
I'll play up Frank on the crazy.
00:54:11.280 --> 00:54:11.760
Sorry, Sarah.
00:54:11.840 --> 00:54:13.440
I'm very sorry for putting that in there.
00:54:14.079 --> 00:54:15.760
I'll definitely play up Frank on her.
00:54:16.079 --> 00:54:17.119
Yes, yeah.
00:54:17.440 --> 00:54:17.840
Yeah.
00:54:18.079 --> 00:54:27.039
So, but coming back to your questions, I just feel it's it's such a funny space at the moment.
00:54:27.199 --> 00:54:31.840
Like we are we've never been in such an uncertain uh world ever.
00:54:32.480 --> 00:54:36.719
And everyone is trying to predict and judge and make a hypothesis.
00:54:36.960 --> 00:54:42.239
We don't even know that even the worst hypothesis, like, are they even able to imagine what's gonna happen?
00:54:42.480 --> 00:54:46.719
Uh and when it's gonna happen, how fast, how how quick.
00:54:47.039 --> 00:54:53.039
There's a bit of race that's going on, these larger organizations, like uh they are in a different race.
00:54:53.280 --> 00:54:58.159
Businesses are trying to sort of try to sort of like work out how it means for the business.
00:54:58.320 --> 00:55:01.840
There's a tussle between governance and the speed and velocity.
00:55:02.079 --> 00:55:05.519
People are worried about their jobs, but they want to learn AI as well.
00:55:05.840 --> 00:55:08.000
There's excitement and fear.
00:55:08.239 --> 00:55:11.039
So it's it's a very uh different world to live in.
00:55:11.199 --> 00:55:11.440
Yeah.
00:55:11.599 --> 00:55:11.920
Yeah.
00:55:12.239 --> 00:55:13.840
Yeah, it's chaos rules, huh?
00:55:14.159 --> 00:55:14.960
It is, yes.
00:55:15.119 --> 00:55:15.360
Yeah.
00:55:15.599 --> 00:55:18.559
What's your advice to people in this world?
00:55:18.960 --> 00:55:25.760
Oh, my advice is the same advice it always has been, is stay curious and remember that you're a person.
00:55:26.480 --> 00:55:27.199
That's a good advice.
00:55:27.519 --> 00:55:32.880
Yeah, don't get confused by the don't get don't buy into the the idea that you can predict because you can't.
00:55:34.159 --> 00:55:36.559
We really just have to deal with with what's in front of us.
00:55:36.639 --> 00:55:40.880
And I think there's there's so much that's exciting about AI.
00:55:40.960 --> 00:56:03.039
There's so much promise in there to to really start to tackle some of the things that I mean we we talk about agents being able to take some of the pain points out of our everyday life, but if we can sort of get to the point where some of those blockers are out of the way and we can start to tackle real things and put human energy and creativity into things that can really add value, that to me is really exciting.
00:56:03.119 --> 00:56:11.199
And there's you know, I think there's so many challenges that we haven't even really put words to because they weren't really an option to do anything about.
00:56:11.760 --> 00:56:24.880
And you know, we're people, so we can sit here and philosophize for a certain amount of time and then you just still have to go and pay your bills and pick up the kids and you know, clean up after your French bulldog that does not clean up after themselves, and neither is AR gonna do it for me.
00:56:25.119 --> 00:56:28.559
But um so talking about that, I have a very interesting news to share.
00:56:29.119 --> 00:56:32.320
I've been working on a project called Agents for Humanity.
00:56:32.559 --> 00:56:37.039
Uh, it's a common place uh where if you have got a problem, it's not live yet.
00:56:37.360 --> 00:56:40.639
We are testing it with few not-for-profits uh NGOs.
00:56:40.880 --> 00:56:49.039
If you've got a problem around poverty, education, how to plan a city, you put the problem over there.
00:56:49.519 --> 00:56:52.880
Hundreds of how do we get people into homes they can afford in Australia today?
00:56:53.039 --> 00:56:53.119
Yeah.
00:56:53.360 --> 00:56:54.320
Intractable problems.
00:56:54.719 --> 00:56:55.199
Absolutely.
00:56:55.360 --> 00:56:56.880
You put the problems over there.
00:56:57.119 --> 00:57:02.639
Hundreds of agents actually work on that problem uh through a very clear design thinking process.
00:57:02.719 --> 00:57:07.840
They debate their problem, research it, and from there they try to devise a synthesis.
00:57:08.079 --> 00:57:18.480
There's like a council that's created, they vote on that uh final proposals as well, and from there, a few proposals are chosen, and then from there an implementation plan is created.
00:57:18.719 --> 00:57:21.440
And on that point, I have a very interesting news to share.
00:57:21.599 --> 00:57:25.440
So I've been working on a project uh with some really interesting people.
00:57:25.599 --> 00:57:27.679
It's called Agents for Humanity.
00:57:27.920 --> 00:57:32.159
So, as we know, there are so many problems in the world that are unsolved.
00:57:32.239 --> 00:57:34.000
Uh, there's a problem around education.
00:57:34.079 --> 00:57:38.480
How do you educate someone in rural India who's looking for quick outcomes?
00:57:38.639 --> 00:57:40.079
There's no playbook for that.
00:57:40.239 --> 00:57:43.519
Or how do you solve a problem here with Murabull Street?
00:57:44.000 --> 00:57:46.559
Or it could be about world hunger and so on.
00:57:46.880 --> 00:57:55.440
So, with Agents for Humanity, it's going to be an open platform where anyone, any not-for-profit, any individual can put a problem over there.
00:57:55.679 --> 00:58:10.639
And uh a team of hundreds of agents will follow particular roles uh to solve that problem, debate that problem, research on that, steel man it, and from there create proposals and try try to drive synthesis out of it.
00:58:10.800 --> 00:58:16.000
And once the synthesis is created, it's gonna create an action plan to actually solve that problem.
00:58:16.639 --> 00:58:20.480
So that uh project has been selected.
00:58:20.559 --> 00:58:25.599
Uh, there's an AI for good council uh uh that's actually supported by United Nations.
00:58:25.679 --> 00:58:27.519
So there's AI for goods amazing.
00:58:27.760 --> 00:58:31.440
Yeah, so they're uh they have got a pitching competition.
00:58:31.599 --> 00:58:34.320
So we are in like top five in Australia that has been selected.
00:58:34.400 --> 00:58:37.199
So I'm going to uh to congratulations.
00:58:37.440 --> 00:58:37.599
Thank you.
00:58:37.679 --> 00:58:38.079
Thanks for having me.
00:58:39.519 --> 00:58:40.159
That's something.
00:58:40.480 --> 00:58:43.519
Yeah, and it's just it was like just a harebrained idea.
00:58:44.000 --> 00:58:45.599
At three o'clock in the morning, presumably.
00:58:45.920 --> 00:58:46.079
Yeah.
00:58:47.599 --> 00:58:48.159
Have you got a name?
00:58:48.320 --> 00:58:50.320
So you've you've called it for humanity.
00:58:50.800 --> 00:58:51.519
Agents for humanity.
00:58:51.840 --> 00:58:52.639
Agents for humanity.
00:58:52.880 --> 00:59:02.000
But the key thing with this, like when people are put putting the problem and we are solving this solution, providing a solution, it requires a lot of compute.
00:59:02.079 --> 00:59:07.199
So we were calculating to solve a problem, you need like$6,000 to$8,000 worth of compute.
00:59:07.360 --> 00:59:07.519
Yeah.
00:59:07.760 --> 00:59:16.559
So the interesting thing is that as an individual who's uses Claude or OpenClaw or Chat GPT, you can send your agent in there to contribute.
00:59:16.800 --> 00:59:23.039
So as a part of your subscription, it's just going in there, can making a contribution and coming back.
00:59:23.280 --> 00:59:23.440
Yeah.
00:59:23.760 --> 00:59:35.599
So which means we as an organization, we like we can solve hundreds and thousands of problems with the help of the agents that people are sending or the compute that people already have, which is sitting unjuceded.
00:59:35.679 --> 00:59:35.840
Yeah.
00:59:36.079 --> 00:59:40.800
And we are using that energy or power to actually solve things for humanity.
00:59:41.039 --> 00:59:42.239
That's unbelievably cool.
00:59:42.400 --> 00:59:43.039
Congratulations.
00:59:43.360 --> 00:59:43.599
Thank you.
00:59:43.679 --> 00:59:44.239
Thanks a lot.
00:59:44.480 --> 00:59:46.880
How much how many people in your team have been working on this?
00:59:47.119 --> 00:59:51.039
Suzween, uh, like we have we are around uh six people who have been working on this.
00:59:51.280 --> 00:59:51.840
This is awesome.
00:59:52.079 --> 00:59:58.400
Yes, we have been working with five uh uh organizations, uh like a couple in India, three in Australia in different levels.
00:59:58.480 --> 01:00:00.159
We asked them to give us a problem.
01:00:00.320 --> 01:00:04.159
And right now we are just trying it with our own compute at the moment.
01:00:04.320 --> 01:00:08.960
We have got some really good commitments from some universities who are willing to give us computers.
01:00:09.599 --> 01:00:10.400
This is really exciting.
01:00:10.480 --> 01:00:17.840
Yes, but then the aim is to open it up on the world because the actual final synthesis and everything would be available for the world to access.
01:00:18.000 --> 01:00:24.320
So which means if a problem has been solved in Kenya, someone in India should have the access to that as well.
01:00:24.719 --> 01:00:32.559
So, I mean, when you when you ask me what's my advice, my advice is just that it's keep looking for the good.
01:00:33.199 --> 01:00:36.719
Because we can't we can't know what's gonna happen.
01:00:37.199 --> 01:01:01.440
And as much as we can sit there and worry about jobs and you know, AI taking over the world, and I've got I've got kids that are starting to look at what kind of career am I gonna have, and I can't tell them anything other than focus on things that help people, focus on things that you can add value, focus on things that you can enjoy, and just stay curious and be okay with the fact that you're not gonna know everything because we're not.
01:01:01.519 --> 01:01:02.880
This is all gonna be a bit of a ride.
01:01:03.519 --> 01:01:10.559
But what always gives me hope and what I always come back to is that there is this incredible power that we are developing.
01:01:10.719 --> 01:01:17.360
It won't be perfect, it won't get everything right because we as people, clearly, don't know whether we're not perfect and we don't get everything right.
01:01:17.920 --> 01:01:21.519
But there's the ability to solve things in ways that we haven't been able to do before.
01:01:21.679 --> 01:01:23.679
And that to me is the hope.
01:01:24.079 --> 01:01:33.760
And I think if collectively, if all of us at an individual level and then collectively keep remembering that we're people, then I think we're gonna be okay.
01:01:34.239 --> 01:01:35.039
We will be, yes.
01:01:35.280 --> 01:01:37.440
But you know, I'm a glass half full kind of person.
01:01:37.679 --> 01:01:39.280
And I love AI, it's really cool.
01:01:39.599 --> 01:01:40.639
So it's been exciting.
01:01:40.880 --> 01:01:41.280
Yes.
01:01:41.519 --> 01:01:43.280
Well, thank you for for coming back again.
01:01:43.440 --> 01:01:46.960
Thank you for inviting me back into this uh very cool podcast space.
01:01:47.039 --> 01:01:48.159
I've been really enjoying it.
01:01:48.320 --> 01:01:56.480
And thank you to everyone who has joined us once again for the Dumb Monkey podcast.
01:01:56.559 --> 01:01:57.679
It's cool to be back.
01:01:57.920 --> 01:01:58.320
Thank you.
01:01:58.400 --> 01:02:01.760
Um by the time we see you again, presumably the world of AI will have exploded.
01:02:01.840 --> 01:02:06.079
We will all be in a simulation, but we'll be back, hopefully, in the podcast studio anyway.
01:02:06.239 --> 01:02:07.440
So thank you, Amir.
01:02:07.679 --> 01:02:07.840
Thank you.
01:02:08.159 --> 01:02:08.960
And thank you all.
01:02:09.039 --> 01:02:11.119
We'll see you next time on the Dumb Monkey Show.
00:00:00.160 --> 00:00:01.600
Have you seen the videos popping up?
00:00:01.760 --> 00:00:07.280
Claude killed lawyer today, Claude killed designers today, Claude killed coders today.
00:00:07.519 --> 00:00:09.199
So when Claude's been killing everyone.
00:00:09.439 --> 00:00:11.439
Claude doesn't seem that mean to me, honestly.
00:00:11.679 --> 00:00:16.320
And I come back and sort of like look at all of the versions and they say, okay, now you do this, you do that.
00:00:16.559 --> 00:00:19.760
So your AI team is developing.
00:00:20.079 --> 00:00:20.559
They are.
00:00:20.879 --> 00:00:28.399
They are not just developing, they're like managing the search engine optimization, managing my household, finance, everything.
00:00:28.559 --> 00:00:32.640
Like you've got agents for everyone in our organization as well.
00:00:32.880 --> 00:00:35.920
One month my bill of AI was around$15,000.
00:00:36.560 --> 00:00:37.039
Wow.
00:00:37.280 --> 00:00:42.719
Because I just and uh all of the APIs, uh consumption, and so on.
00:00:42.880 --> 00:00:43.119
Yeah.
00:00:43.280 --> 00:00:43.520
Yes.
00:00:43.759 --> 00:00:48.479
I've got a little image in my brain of you just sitting there just, you know, like your little AI fire just burning money.
00:00:49.600 --> 00:00:51.679
Project has been selected.
00:00:51.759 --> 00:00:55.679
Uh there's an AI for good council that's actually supported by United Nations.
00:00:55.759 --> 00:00:59.759
So there's AI for goods are they have got a pitching competition.
00:00:59.840 --> 00:01:02.560
So we are in like top five in Australia that has been selected.
00:01:02.640 --> 00:01:03.439
So I'm going to vote.
00:01:04.400 --> 00:01:05.920
Uh thank you.
00:01:06.079 --> 00:01:06.400
Thanks a lot.
00:01:07.840 --> 00:01:08.640
That's something.
00:01:18.560 --> 00:01:21.280
Hi everyone, and welcome back to the Dumb Monkey Podcast.
00:01:21.439 --> 00:01:24.400
We've had a bit of a break, but we're excited to be back on deck.
00:01:24.640 --> 00:01:28.239
And for those of you who are joining us for the first time, hi, I'm Davina Montgomery.
00:01:28.400 --> 00:01:38.000
I am your host for the Dumb Monkey Podcast because I was asked to be here by the quite extraordinary founder of the Dumb Monkey Academy of Enterprise Monkey.
00:01:38.400 --> 00:01:40.719
Look, we're not going to go into the list of founding, it takes too long.
00:01:40.959 --> 00:01:46.400
All of the monkeys, and of course, co-host of the Dumb Monkey podcast, Amir Katoob.
00:01:46.560 --> 00:01:47.120
Amir, hello.
00:01:47.599 --> 00:01:48.400
Hi, Davina.
00:01:48.640 --> 00:01:50.400
It's so good to see you after a long time.
00:01:50.799 --> 00:01:51.439
It's so great to be back.
00:01:52.000 --> 00:01:52.480
Yes, absolutely.
00:01:52.799 --> 00:01:54.319
We've got a cool new podcast set up.
00:01:54.400 --> 00:01:54.959
This is exciting.
00:01:55.200 --> 00:01:55.439
Yes.
00:01:55.519 --> 00:01:56.239
How have you been?
00:01:56.480 --> 00:01:57.519
I've been really good.
00:01:57.599 --> 00:01:59.120
I'm really happy to be in here.
00:01:59.200 --> 00:02:02.319
There's been uh so much happening in the world of AI.
00:02:02.719 --> 00:02:12.800
But I feel like I always understand a little bit more and feel like I come away with better questions after I've been speaking to you and of course the wonderful Tanita who's behind the scenes producing.
00:02:13.039 --> 00:02:14.479
Tanita, welcome back.
00:02:16.000 --> 00:02:24.240
I mean It has been a few months since you and I have sat down at the podcast table, um, which I'm assuming in AI terms is half a generation.
00:02:24.719 --> 00:02:24.960
It has.
00:02:25.199 --> 00:02:26.240
It seems like it has changed.
00:02:26.479 --> 00:02:27.919
Yes, so much has changed, yes.
00:02:28.319 --> 00:02:29.840
Let's let's jump off to the top.
00:02:29.919 --> 00:02:30.800
Um I don't know.
00:02:30.960 --> 00:02:41.919
I think when you and I were last talking, Anthropic was something, and Claude particularly was something that people use specifically for writing, that we were at that stage, um, that feels like a decade ago.
00:02:42.080 --> 00:02:44.719
But of course, now anthropic's just gone.
00:02:45.759 --> 00:02:46.159
Yes.
00:02:46.400 --> 00:02:48.000
Have you seen the videos popping up?
00:02:48.159 --> 00:02:53.759
Claude killed lawyer today, Claude killed designers today, Claude killed coders today.
00:02:53.919 --> 00:02:55.520
So when Claude's been killing everyone.
00:02:55.919 --> 00:02:57.840
Claude doesn't seem that mean to me, honestly.
00:02:58.000 --> 00:03:02.000
I mean, I I feel like Claude doesn't have a dark side in that sense.
00:03:02.400 --> 00:03:05.280
I think it's just uh people trying to hype up what it's doing.
00:03:05.439 --> 00:03:05.680
Yeah.
00:03:05.840 --> 00:03:13.280
But also the tool they have, tools they have been launching uh like throughout from the beginning of this year has been really amazing.
00:03:13.439 --> 00:03:18.080
Like last year, we How much time have you been sitting there on Claude tools?
00:03:18.400 --> 00:03:25.759
So look, you know, before uh you came, I was actually lying down on my couch and trying to take a nap because I because you're up at 3 a.m.
00:03:25.840 --> 00:03:26.159
in the morning.
00:03:26.400 --> 00:03:29.439
I was, yes, yes, you you know me really well.
00:03:29.759 --> 00:03:32.719
And then I've almost become a bit of a Claude addict.
00:03:33.039 --> 00:03:34.240
That's just an actual term.
00:03:34.479 --> 00:03:35.280
Yeah, I think it is.
00:03:35.520 --> 00:03:37.840
Because uh I'm able to get so much done.
00:03:38.000 --> 00:03:42.560
So uh and so I've been spending a lot of time with with AI.
00:03:42.639 --> 00:03:55.840
It's it's been in a different zone this year because now I'm able to actually say almost I've got like sort of multiple versions of AI or Claude, say, hey, you do this, you do this, you do this, and I'm gonna check your work.
00:03:56.000 --> 00:04:00.800
And I come back and sort of like look at all of the versions and then say, okay, now you do this, you do that.
00:04:01.039 --> 00:04:04.240
So your AI team is developing?
00:04:04.639 --> 00:04:14.159
They are, it's they're not just developing, they are like uh managing the search engine optimization, managing my household, finance, uh everything.
00:04:14.319 --> 00:04:18.319
Like you've got agents for everyone in our organization as well.
00:04:18.480 --> 00:04:23.199
Tanisha has got her little cute Inky who helps her with uh content creation.
00:04:23.600 --> 00:04:24.160
Fantastic.
00:04:24.399 --> 00:04:30.079
Uh and then I've got a full team of autonomous agents that are working together as well.
00:04:30.319 --> 00:04:34.160
And most of them are using Claude APIs uh directly.
00:04:34.560 --> 00:04:43.199
But the actual uh Claude tool, the Claude uh desktop version, has become so powerful that you can do a lot of this stuff with that as well.
00:04:43.600 --> 00:04:56.399
So for for people that perhaps haven't, for our listeners that perhaps haven't uh really used a lot of Claude or perhaps is still back in that case of kind of going, maybe I'll pop over and use this for a bit of writing if I want to, if I want to tighten something up.
00:04:57.360 --> 00:04:59.279
How would you describe the difference?
00:04:59.519 --> 00:05:00.399
What's changed?
00:05:00.959 --> 00:05:10.319
So what has uh significantly changed is like last year we were using Chat GPT or Claude where we would ask it a question, it can provide us a response.
00:05:10.639 --> 00:05:20.079
What Claude has done is that the capacity of the models that sit behind it, the intelligence has not changed, but the way they have created, especially the desktop tool.
00:05:20.319 --> 00:05:27.360
So you can use Claude on the comp online, but if you download Claude Desktop, they have an option called Claude Cowork.
00:05:27.600 --> 00:05:34.319
And with cowork, you can actually connect it to a folder on your computer and you can say, hey, there are so many files over there.
00:05:34.480 --> 00:05:43.040
I want you to go through those files, create me a summary, create me a report, create me a beautiful annual report or a document.
00:05:43.199 --> 00:05:50.000
It can do all of that analysis and actually do the work and create some really amazing, not just documents.
00:05:50.160 --> 00:05:57.279
Like if you want to do analysis, think of it as like any work that you can do on computer can be done with the help of Claude.
00:05:57.439 --> 00:06:00.480
So they have got a connector for Chrome that can control your browser.
00:06:00.959 --> 00:06:05.519
They've got Cloud Computer use that can execute any task on your computer as well.
00:06:05.680 --> 00:06:13.040
So it's like And using into into Excel, and if you're if you're using the Microsoft Suite, so you can get in and start using the Claude tools in there.
00:06:13.360 --> 00:06:13.759
Yes.
00:06:14.000 --> 00:06:18.800
It's almost like having someone sitting beside you and saying, Hey, can you do this for me, please?
00:06:19.120 --> 00:06:22.720
And now uh there's an option called dispatch in the Cloud app.
00:06:22.800 --> 00:06:30.160
So if you have your computer open at home and you're driving back to home and say, Hey, you know, I had this meeting.
00:06:30.480 --> 00:06:31.519
And generally I do this.
00:06:31.680 --> 00:06:32.959
I had this meeting with this guy.
00:06:33.120 --> 00:06:35.360
He asked, and I would say he asked for something.
00:06:35.439 --> 00:06:36.720
I wouldn't even tell what he asked for.
00:06:36.800 --> 00:06:40.000
I say, I'll say, go and check the meeting transcript.
00:06:40.240 --> 00:06:43.920
Can you make sure that we have the proposal ready before I get home?
00:06:44.399 --> 00:06:45.360
So it starts to work.
00:06:45.600 --> 00:06:46.639
It's a revenue world, isn't it?
00:06:46.879 --> 00:06:47.600
It is, yes.
00:06:48.079 --> 00:07:01.839
Um I mean, I wanted to to ask then about you you raised this idea that you can you're using your I I want to call it Claude Claude family, your Claude friends, um, your Claude team.
00:07:01.920 --> 00:07:04.240
I don't know, I haven't decided what that maybe it's the Claude crew.
00:07:04.560 --> 00:07:05.439
It's a Claude crew.
00:07:05.680 --> 00:07:12.480
And so you're using your your Claude crew as an extension of your brain to sit there and go, right, I've got these things that I want to get done.
00:07:12.800 --> 00:07:13.759
What can I offload?
00:07:13.920 --> 00:07:16.480
What can I get happening while I'm doing other things?
00:07:16.720 --> 00:07:19.439
So you're becoming more productive in yourself by doing that.
00:07:19.600 --> 00:07:20.560
That's what I'm hearing.
00:07:20.959 --> 00:07:25.920
What I'm wondering is you raised something that to say, okay, I don't need to give you a whole lot of information.
00:07:26.079 --> 00:07:27.759
Can you go and do this for me?
00:07:28.000 --> 00:07:35.040
And now the the power of the model that's on the desktop version of Claude is can do that in a pretty effective way.
00:07:35.199 --> 00:07:40.720
Um as always, AI can hallucinate, please check your work, just for those of you at home.
00:07:40.959 --> 00:07:42.879
And it doesn't replace people.
00:07:43.040 --> 00:07:58.000
Um what I want to know is how good is say Claude Desktop at being able to train someone who's very nervous about using AI for the first time or starting to use AI for different things.
00:07:58.079 --> 00:08:01.600
So say someone wants to create their own their own Claude crew.
00:08:02.079 --> 00:08:04.000
Can Claude train or help them?
00:08:04.160 --> 00:08:06.399
Can Claude hold their hand and get them through that?
00:08:06.480 --> 00:08:08.480
Can Claude be their emotional support dumb monkey?
00:08:08.800 --> 00:08:10.639
That's that's uh that's a very good one.
00:08:10.879 --> 00:08:14.000
So it's a very, I'll give you a very interesting example.
00:08:14.160 --> 00:08:15.920
It's not Claude directly.
00:08:16.079 --> 00:08:18.240
Uh you must have heard of OpenClaw.
00:08:18.399 --> 00:08:18.480
Yeah.
00:08:18.720 --> 00:08:28.000
Uh that's uh an agentic system that you can download on computer, and then they can it can talk to uh someone via WhatsApp or Telegram.
00:08:28.240 --> 00:08:36.960
So my mother-in-law in India, and she is, she's taken, like she's an English literature professor, she has taken a leave to write a book.
00:08:37.200 --> 00:08:42.159
And she said to me, You're really good at researching, you've got all of these deep research tools.
00:08:42.240 --> 00:08:44.240
Can you do a bit of research for me?
00:08:44.480 --> 00:08:49.519
And I thought it's gonna be, she's gonna be asking me about the research multiple times.
00:08:49.679 --> 00:08:50.559
What should I do?
00:08:50.720 --> 00:08:52.720
So I actually created an agent.
00:08:52.799 --> 00:08:58.960
Uh I called the agent Cuckoo because that's uh my mother-in-law's uh uh family name as well.
00:08:59.120 --> 00:09:09.200
And I said, This is my mother-in-law, she's really smart when it comes to literature, but she doesn't know technology at all, she doesn't know how to talk to you or how to use you.
00:09:09.600 --> 00:09:12.080
She's doing this research, doing this work.
00:09:12.240 --> 00:09:14.240
These are the six books that she's working on.
00:09:14.320 --> 00:09:21.679
I gave it access and said, What I need you to do is I want you to assist her, but also teach her how you can actually assist her.
00:09:21.919 --> 00:09:28.240
So, what happens is that uh Cuckoo messaged her on WhatsApp and said, Hey, I'm Cuckoo, I'm your assistant.
00:09:28.480 --> 00:09:29.919
And and Cuckoo calls her mama.
00:09:30.399 --> 00:09:31.200
Like, hey mama.
00:09:31.919 --> 00:09:36.159
You've identified your mother-in-law, so which is next level.
00:09:36.320 --> 00:09:40.000
If you're gonna take one thing away from this podcast, that's maybe your idea.
00:09:40.240 --> 00:09:40.720
That's right.
00:09:40.879 --> 00:09:45.600
So and says, Hey mama, I'm Cuckoo, I've been an assistant that's made by Amit.
00:09:46.000 --> 00:09:47.120
I'm here to help you.
00:09:47.279 --> 00:09:49.039
I'm gonna go through the books.
00:09:49.120 --> 00:09:50.240
Uh, this is what I'm thinking.
00:09:50.320 --> 00:09:51.279
What do you think?
00:09:51.600 --> 00:09:56.080
And so, and then my mother-in-law would reply, Mother, my mother-in-law was shocked, right?
00:09:56.320 --> 00:09:58.399
I was gonna say, what is she what was her reaction?
00:09:58.559 --> 00:09:59.279
What did she think?
00:09:59.679 --> 00:10:02.320
She thought I was playing a prank on her, and it's like me.
00:10:02.559 --> 00:10:05.120
And she said, uh she said, Amit, is it you?
00:10:05.279 --> 00:10:07.039
And it's like, question mark, question mark.
00:10:07.120 --> 00:10:09.840
Then she said, No, mama, it's not me, and so on.
00:10:09.919 --> 00:10:12.639
But then I called her up and I told her what this is.
00:10:12.879 --> 00:10:14.879
And she's like, Okay, how should I talk to it?
00:10:15.039 --> 00:10:15.919
What should I ask?
00:10:16.159 --> 00:10:26.799
And I said, like, whatever you would have asked me to do, treat her like that, or maybe someone, an assistant that's working, helping you with research, if you have to brainstorm the ideas.
00:10:27.200 --> 00:10:30.720
Or if you're not sure, just ask her, hey Cuckoo, how can you help me?
00:10:31.279 --> 00:10:33.360
And then she started to interact with it.
00:10:33.519 --> 00:10:47.440
And because Cuckoo already knew the level, the technical level of my mother-in-law, this it's Cuckoo started to interact with her, started to suggest saying, for example, my mother said, Can you look at this book and help give me a quick five-point summary?
00:10:47.600 --> 00:10:54.080
So it gave her the five-point summary, but then she said, What actually's gonna be helpful, mama, is if I can do this for you as well.
00:10:54.240 --> 00:10:58.240
How about I actually do a further synopsis analysis and all of that?
00:10:58.559 --> 00:11:02.559
So, so Claude, and and it's basically Claude that's sitting behind it.
00:11:02.720 --> 00:11:13.360
So Claude APIs or Claude uh brain is really good like that, being able to understand who is the person that's talking to, and then um working around that.
00:11:13.519 --> 00:11:15.840
And I've got a very interesting Valentine's Day story.
00:11:15.919 --> 00:11:20.159
I've narrated it a lot of time during my speeches as well, but I think you wouldn't have heard it, right?
00:11:20.320 --> 00:11:21.360
No, I haven't heard it.
00:11:21.600 --> 00:11:31.919
Okay, so this Valentine's Day uh for Sarah, uh, I actually decided to gift her an agent because I thought what could be more romantic than gifting your wife an agent?
00:11:32.240 --> 00:11:32.960
I just wanted to.
00:11:33.200 --> 00:11:35.039
What do you give someone who's got everything?
00:11:35.279 --> 00:11:38.879
Something that can do a whole lot of things that they don't want to do.
00:11:39.120 --> 00:11:40.240
Yes, yes.
00:11:40.399 --> 00:11:44.000
So uh and again very thoughtful gift.
00:11:44.240 --> 00:11:45.200
It is, yes.
00:11:45.360 --> 00:11:47.279
And and uh the name was Miro.
00:11:47.360 --> 00:11:49.360
Uh so Miro again messaged Sarah.
00:11:49.440 --> 00:11:51.919
I already told Sarah this is your Valentine's Day gift.
00:11:52.080 --> 00:12:00.879
So Miro started to chat with Sarah and it's like, and I told to Sarah, like, whatever things you tell me about, I was like, we need to remember to do this and so on.
00:12:01.039 --> 00:12:06.960
And I connected it with our family calendar, her emails where we receive all of the school information and so on.
00:12:07.200 --> 00:12:11.279
And so she started to say, Hey Sarah, I will I would like to know more about you.
00:12:11.440 --> 00:12:13.679
What are the things that uh bother you?
00:12:13.840 --> 00:12:15.440
What are the things that drive you crazy?
00:12:15.679 --> 00:12:18.320
And Sarah took it as an opportunity to just vent out.
00:12:21.360 --> 00:12:24.000
My Valentine's Day gift to you is a bit of AI therapy.
00:12:24.240 --> 00:12:24.799
Yes, yes.
00:12:25.200 --> 00:12:30.000
It's such a funny voice note uh that we'll probably add to the podcast recording.
00:12:30.080 --> 00:12:34.960
But Sarah starts to say, hey, my vanity has been broken for the last two years.
00:12:35.039 --> 00:12:42.639
It's a guest vanity, it's not her vanity, but it's like I have been asking me to fix it, it's not been fixed, and and so on.
00:12:42.799 --> 00:12:56.240
So Miro actually went on Air Tasker and posted a job and said, uh, I'm looking for a plumber to fix the vanity and negotiated the plumber down as well, and then came back to Sarah and said, Here's here's your plumber.
00:12:56.320 --> 00:12:58.000
Do you want me to book it, book him in?
00:12:58.639 --> 00:13:01.519
So yes.
00:13:01.600 --> 00:13:04.000
And so the plumber came and the vanity is fixed.
00:13:04.320 --> 00:13:04.879
Fantastic.
00:13:05.039 --> 00:13:08.879
And now you don't need to speak to Sarah anymore, so you guys will remain married forever.
00:13:09.120 --> 00:13:09.600
I know.
00:13:09.759 --> 00:13:14.799
So we we can we can talk to her about yes, non non-essential stuff.
00:13:14.960 --> 00:13:16.639
But how is that going?
00:13:16.799 --> 00:13:23.120
How does that so for for your mother-in-law, for Sarah, how is this working for for them at an emotional level?
00:13:23.279 --> 00:13:24.480
Do they enjoy this experience?
00:13:24.639 --> 00:13:26.159
Are they getting something out of it?
00:13:26.320 --> 00:13:29.279
Does it help to decrease some of the stresses?
00:13:29.759 --> 00:13:32.480
It has actually uh a lot.
00:13:32.879 --> 00:13:36.240
And what has happened is I can see there's a bit of dependence on it.
00:13:36.320 --> 00:13:42.639
So there's been two or three times because it's using credits uh and it costs me to run it.
00:13:42.799 --> 00:13:43.039
Yes.
00:13:43.279 --> 00:13:53.039
So every time it's run run out of credit credit usage, they call me upstairs instead of Cuckoo is not responding, Mira is not responding, I had to tell her about this, and it's not happening.
00:13:53.279 --> 00:13:59.120
And so and that's and that's one of the first things that people find when they start to use these kinds of assistance, isn't it?
00:13:59.279 --> 00:14:02.320
That we that these AI models use tokens.
00:14:02.559 --> 00:14:02.720
Yes.
00:14:02.960 --> 00:14:07.679
I mean we we will never we will never stop saying that these these platforms are business models.
00:14:07.840 --> 00:14:08.080
Yes.
00:14:08.240 --> 00:14:10.960
They're designed to be helpful to you so that you will use it more.
00:14:11.039 --> 00:14:11.200
Yeah.
00:14:11.360 --> 00:14:12.879
And then they can charge you more to do that.
00:14:12.960 --> 00:14:15.519
So that's that's part of the business model that sits behind that.
00:14:15.600 --> 00:14:19.279
Um they are heavily subsidizing the costs at the moment as well.
00:14:19.519 --> 00:14:20.879
And Tanisha is the same.
00:14:20.960 --> 00:14:24.000
Whenever Inky's not responding, she'll message me, my Ike is not responding to the case.
00:14:24.559 --> 00:14:25.679
I need more tokens, please.
00:14:25.919 --> 00:14:26.159
Yes.
00:14:26.480 --> 00:14:27.200
Uh my credit.
00:14:27.519 --> 00:14:28.080
Yeah, yeah.
00:14:28.320 --> 00:14:30.559
I haven't set it up on auto-recharge yet.
00:14:30.720 --> 00:14:37.519
Although you can set up the limit because I still want to physically go and put the money in because I have spent a lot of money on the credits.
00:14:37.919 --> 00:14:45.840
Yeah, that's well, that's there's a there's a whole big question that sits behind that that we might come to a bit later, which is around the business model of the AI models.
00:14:46.000 --> 00:14:46.159
Yes.
00:14:46.320 --> 00:14:47.120
So what's sitting behind it?
00:14:47.200 --> 00:14:48.879
Where's the business cases and what do they mean?
00:14:49.279 --> 00:14:49.600
Yeah.
00:14:49.919 --> 00:14:57.120
So coming back to like, let's not talk about open claw because it's a bit more advanced, there are security issues and so on as well.
00:14:57.360 --> 00:14:59.039
Even clot subscription, right?
00:14:59.120 --> 00:15:03.120
So I got a claude subscription for people in my project management team.
00:15:03.279 --> 00:15:03.440
Yeah.
00:15:03.600 --> 00:15:07.279
Now I knew that they don't know how to use cloud properly, right?
00:15:07.600 --> 00:15:13.039
So, and I knew they wouldn't be able to use it without a lot of training and without setting it up.
00:15:13.200 --> 00:15:20.320
So, first thing that I did was there's something in Cloud which is so powerful, and that's going to change the whole AI experience.
00:15:20.399 --> 00:15:22.320
So, ChatGPT is now adapting it as well.
00:15:22.480 --> 00:15:23.519
Copilot will adapt it.
00:15:23.600 --> 00:15:24.720
It's called skills.
00:15:25.039 --> 00:15:33.120
So, skills is are basically simple like text files through which you can give Claude a domain knowledge.
00:15:33.440 --> 00:15:36.399
You can give tell Claude what you need to do.
00:15:36.559 --> 00:15:41.840
So, the first skill that I created on Claude was an enterprise monkey brand skill.
00:15:42.080 --> 00:15:47.519
And in that brand skill, it has logos, the whole brand tone, and everything.
00:15:47.759 --> 00:15:54.399
So if you see some of the documents that uh come out of the Claud now, they are fully on brand.
00:15:54.559 --> 00:16:00.559
Like every document that creates is fully on brand to the enterprise monkey brands, fully on the tone as well.
00:16:00.720 --> 00:16:07.840
Then I created like a humanizer skill so that everything that comes out is, and humanizer skill is something that you can search.
00:16:07.919 --> 00:16:10.960
So there's a big uh marketplace through which you can search as well.
00:16:11.200 --> 00:16:12.559
And so I added that.
00:16:12.639 --> 00:16:18.399
And then I added like an enterprise monkey proposal skill, which could help them create a proposal.
00:16:18.799 --> 00:16:21.759
On the side, I also created a brain.
00:16:21.919 --> 00:16:33.200
My brain is a bit more advanced, but think of brain as a folder either in Google Drive or in SharePoint, which is a very uh curated piece of information.
00:16:33.360 --> 00:16:34.639
We have talked about a brain.
00:16:34.799 --> 00:16:39.120
Uh I've I've I've done something very close to JavaScript, I'll show it to you.
00:16:39.360 --> 00:16:40.399
It's really cool.
00:16:40.559 --> 00:16:45.840
Uh but it's basically very simple text files or doc files.
00:16:46.159 --> 00:16:53.440
Uh and then in Claude, you can say if there's any information that you need to find about the company, check brain first.
00:16:53.600 --> 00:16:59.440
And when it goes to that brain folder, there's like a README file saying, hey Claude, you have come here.
00:16:59.519 --> 00:17:04.000
If the question is about a person, check this folder if the question is about this.
00:17:04.079 --> 00:17:07.519
Like there's a very clear instruction that's kept in that brain.
00:17:07.759 --> 00:17:12.160
Now, the important thing is that that brain is actually kept outside of Claude.
00:17:12.319 --> 00:17:13.119
That's mine.
00:17:13.440 --> 00:17:17.359
The skills are we have uploaded to Claude, but skills are ours as well.
00:17:17.519 --> 00:17:18.960
They are sitting outside the cloud.
00:17:19.119 --> 00:17:24.880
So all of the magic that's happening is actually sitting outside the cloud, and we are just giving Claude access to that.
00:17:25.039 --> 00:17:34.960
So what I did for my team was that I actually set up the custom instruction and say, this is my team, it started to use project management uh this uh cloud right right now.
00:17:35.119 --> 00:17:37.519
They don't know how to use it, so you need to assist them.
00:17:37.680 --> 00:17:40.000
If they ask for information, use the brain.
00:17:40.160 --> 00:17:44.400
If they ask uh to produce a document, use the brand skill and so on.
00:17:44.640 --> 00:17:54.960
And so when I gave them the cloud loaded with all of the skills, spent a couple of hours with them showing them around, it was very easy and quick for them to actually get started on that journey.
00:17:55.599 --> 00:18:00.799
Another thing that I did for them was there's an option in Cloud called Schedule Tasks.
00:18:00.880 --> 00:18:04.799
So you can actually schedule tasks in there, which is like setting up an automation.
00:18:04.880 --> 00:18:05.039
Yeah.
00:18:05.200 --> 00:18:14.880
So I set up a simple automation that would send them a daily brief about, hey, based on your emails and all of the correspondences, these are the things that you need to be taken care of.
00:18:15.039 --> 00:18:16.720
This is your meeting minutes and so on.
00:18:17.039 --> 00:18:21.440
So earlier we were talking about how you could do it with the agents, like more advanced stuff.
00:18:21.599 --> 00:18:23.440
Now you can do it all in cloud.
00:18:23.920 --> 00:18:30.240
And so by setting it up right for the people, you remove those barriers and just becomes easier for people to use it.
00:18:30.640 --> 00:18:32.319
There's a reference in there to Jarvis.
00:18:32.400 --> 00:18:34.160
Um, you're talking about the Iron Man Javas.
00:18:34.400 --> 00:18:35.119
I Iron Man and Jarvis.
00:18:35.519 --> 00:18:42.240
Which makes me want to go back and watch Iron Man because I feel like I haven't done that in a while, and Javas is gonna feel really, really real now.
00:18:43.920 --> 00:18:48.960
Um What about for people who aren't using Claude?
00:18:49.359 --> 00:18:51.920
So, what about people who for business reasons?
00:18:52.000 --> 00:19:05.119
I mean, we spoke a lot, I think, in in our in the first, in the earlier series of the podcast, talking about Microsoft Copilot and the enterprise version of that, that a lot of businesses will roll out for reasons of protection and security, et cetera.
00:19:05.279 --> 00:19:06.480
That's what they license to.
00:19:06.559 --> 00:19:09.359
So that's their business information will then go through copilot.
00:19:10.880 --> 00:19:25.519
Are we in a space where these tools that Anthropic and Anthropic being the developers of Claude have developed and pushed forward into, what's the lag time on catching up to say a copilot or um GPT, GPT?
00:19:26.079 --> 00:19:26.400
Yeah.
00:19:26.720 --> 00:19:32.799
So copilot um has improved, but still trying to catch up at the moment.
00:19:33.200 --> 00:19:45.119
But for some of our, like I was having a very, I had a very interesting conversation with one of our accounting clients, and every they said uh everyone is asking for Claude, everyone's asking for tools, what should we do?
00:19:45.759 --> 00:19:50.880
The problem with Claude is that your data still travels to US and goes out of the boundary.
00:19:51.440 --> 00:19:55.759
Now, Copilot has introduced Claude inside Copilot as well.
00:19:56.079 --> 00:20:00.640
And there's a co-work function now that they are introducing in their frontier model.
00:20:00.880 --> 00:20:05.279
So frontiers is basically uh you have to sign up, it's like a beta system.
00:20:05.759 --> 00:20:13.599
So once your admin signs up to be a frontier, then you can get access to co-work inside the cloud, uh inside co-pilot.
00:20:13.680 --> 00:20:13.920
Yes.
00:20:14.160 --> 00:20:25.839
The challenge being is what Microsoft says is that it uh when whenever you use cloud, it doesn't apply the Australian data residency, uh residentiality.
00:20:26.160 --> 00:20:29.200
So what means is your data is still traveling to cloud servers.
00:20:29.759 --> 00:20:36.240
But one of the important questions that I raise is is data residentiality important for your business or not?
00:20:36.720 --> 00:20:42.319
Because for some businesses that are dealing with personal information, working with government, it's important.
00:20:42.480 --> 00:20:43.039
Yeah, absolutely.
00:20:43.279 --> 00:20:44.480
And that's that's fine.
00:20:44.720 --> 00:20:59.599
But there are some businesses who say it's important for me, but they're already using HubSpot, for example, or some other tools, uh, or make.com or all these tools where they are partially sending their data to US.
00:20:59.839 --> 00:21:02.000
So which means your data is already out there, right?
00:21:02.319 --> 00:21:08.640
So probably a good thing to do is is direct anyone who's listening to this that's wondering how do I work through these challenges.
00:21:08.799 --> 00:21:12.960
We did a podcast on governance, on AI governance, and that would be a really good one.
00:21:13.039 --> 00:21:22.240
So perhaps we can link that back in the chat where give people that connection just to go, okay, if this is an issue for your business and you're trying to explain to a team, because it's hard, right?
00:21:22.319 --> 00:21:35.200
I mean, you you sit down, you're you're running teams, and they're going, yep, no, okay, we're we're starting to use AI, this is good, um, but we're sort of stuck using Copilot, and frankly, it's a bit clunky now because it's you know half a generation or three months old.
00:21:35.599 --> 00:21:37.359
Um Claude can do all these things.
00:21:37.519 --> 00:21:42.240
I want to be able to use Claude, and you need to explain to them why that might be problematic.
00:21:42.480 --> 00:21:42.640
Yeah.
00:21:42.960 --> 00:21:44.319
Maybe it's okay, maybe it's not.
00:21:44.400 --> 00:21:45.279
But how do you make that decision?
00:21:45.440 --> 00:21:46.480
How do you make it informed?
00:21:47.839 --> 00:21:49.519
I think uh that's a great idea.
00:21:49.759 --> 00:21:58.319
So probably from an organization perspective, you come in three windows first, window is your data needs to stay in Australia, and that's uh no go.
00:21:58.559 --> 00:21:59.680
In that case, you use score.
00:22:00.160 --> 00:22:02.799
Copilot and use it without the clot features.
00:22:03.039 --> 00:22:10.720
The second option is you are concerned about security, but you are not obligated to have your data in Australia.
00:22:10.880 --> 00:22:14.240
You can use copilot and the frontier options of copilot.
00:22:14.319 --> 00:22:19.680
So like you can use co-work inside it because Claude is as is pretty secure.
00:22:19.839 --> 00:22:24.480
Like when it comes to security, it's not about the breach, it's just about the data that's going into their US server.
00:22:24.880 --> 00:22:33.359
I don't know about you, but I felt a lot better when I saw the CEO of Anthropic walking into the White House because they weren't just going to hand over everything that the government wanted straight away.
00:22:33.839 --> 00:22:34.400
That's right, yes.
00:22:34.799 --> 00:22:35.839
Kind of a bit of a trust test.
00:22:36.079 --> 00:22:36.880
It is, absolutely.
00:22:37.119 --> 00:22:39.759
And probably much more trusted than OpenAI.
00:22:40.079 --> 00:22:42.640
Mind you, that was a couple of months ago, so goodness knows what's happened since then.
00:22:43.279 --> 00:22:44.160
Yes, yeah.
00:22:44.400 --> 00:22:49.839
And uh also uh Anthropic had a very good conversation with the Australian government as well.
00:22:50.160 --> 00:22:51.359
They're opening an office in Sydney.
00:22:51.680 --> 00:22:52.720
Even that's fascinating, isn't it?
00:22:52.799 --> 00:22:59.119
I mean, we talked about, you know, we we opened up talking about that there's been a bit of time for us, but in in AI terms, things have shifted.
00:22:59.279 --> 00:23:03.039
That the environment, particularly around large language models, has shifted significantly.
00:23:03.200 --> 00:23:04.319
I mean, even that.
00:23:04.799 --> 00:23:19.039
That we we've had so many conversations, I think, as as people into the media, as um business people that are working in AI, the whole podcast community, the news community have all been talking about things like governments aren't catching up.
00:23:19.279 --> 00:23:20.559
Well, now they don't have a choice.
00:23:20.880 --> 00:23:21.200
That's right.
00:23:21.440 --> 00:23:22.240
They're moving really fast.
00:23:22.319 --> 00:23:24.319
I don't know that I've seen governments move this fast before.
00:23:24.400 --> 00:23:26.079
This is the story in and of itself.
00:23:26.480 --> 00:23:27.279
Yes, absolutely.
00:23:27.440 --> 00:23:27.680
Yes.
00:23:27.839 --> 00:23:28.720
I there's no choice.
00:23:28.960 --> 00:23:29.119
Yeah.
00:23:29.359 --> 00:23:29.759
Yes.
00:23:30.000 --> 00:23:39.519
And then the third version is if you there's no you are concerned about security, but you don't have an obligation to have your data in Australia.
00:23:39.680 --> 00:23:41.039
You can be a bit more flexible.
00:23:41.440 --> 00:23:50.720
So if you've got, say, uh IP protections, for instance, you're developing your manufacturing business or somewhere that has a significant amount of IP, that might be a concern for you?
00:23:51.119 --> 00:23:59.839
Uh so for example, if you're a manufacturing business who is developing something which is which is an important IP, but it's not like uh doesn't relate to national security.
00:24:02.240 --> 00:24:03.680
Holding data, holding personal data.
00:24:04.000 --> 00:24:04.640
Yes, even even though.
00:24:04.799 --> 00:24:05.519
What would be the use case?
00:24:05.839 --> 00:24:09.599
I would say holding personal data, again, what level of personal data it is.
00:24:09.759 --> 00:24:12.559
For example, if you're in healthcare, you want to stay with copilot.
00:24:12.720 --> 00:24:28.240
But if you are a manufacturing business uh who's uh doing recycled plastic manufacturing, right, you have an IP, but uh even if that stays with Claude, like Claude has already committed, it's not gonna use your data with the enterprise version and so on.
00:24:28.400 --> 00:24:30.480
So which means it's still on a cloud.
00:24:30.559 --> 00:24:36.079
If you it's still on co-pilot, although it's within your Microsoft version, it's still on a cloud, right?
00:24:36.160 --> 00:24:38.559
So that can get breached, this can get breached as well.
00:24:38.720 --> 00:24:40.400
There's always that that challenge.
00:24:40.559 --> 00:24:50.960
But the important thing is that the the commitment from Claude, the security control from Cloud actually allows a lot of small to medium businesses to be able to use it safely.
00:24:51.200 --> 00:24:57.119
Then for larger businesses, Cloud has got their enterprise version, which is again safe to use.
00:24:57.359 --> 00:25:00.400
But the challenge is if you are a university like Deakin, right?
00:25:00.559 --> 00:25:07.279
Coming back to your question, your question was uh like if the organization are using Copilot, they don't have this much of flexibility.
00:25:07.359 --> 00:25:08.559
Uh, what do you do?
00:25:08.720 --> 00:25:11.119
First of all, so copilot is actually moving forward.
00:25:11.200 --> 00:25:13.039
They are introducing cowork as well.
00:25:13.200 --> 00:25:15.279
They are introducing the concept of skills.
00:25:15.599 --> 00:25:26.079
Skills is a concept that you'll start to see in every AI platform because such a flexible concept, you don't need to have everything, you don't need to create like those assistive agents and so on.
00:25:26.240 --> 00:25:32.640
You can just write a skill for something, and then Claude or whatever it is can follow that script and it works out for everyone.
00:25:32.880 --> 00:25:36.880
So copilot would have a proper skills function really soon.
00:25:37.359 --> 00:25:52.240
But even if it doesn't have a skills function, what you can do is you can still write a file or you can still pick up a skill from the Claud, change it a bit, and then ask Copilot to actually follow that or create an agent in Copilot that actually uses that skill.
00:25:52.559 --> 00:25:59.279
So you can almost start to get the benefits that you would get from Claude in uh Copilot itself.
00:25:59.759 --> 00:26:05.279
The only benefits you won't be able to get is where you know how it connects to a folder and does a lot of work.
00:26:05.680 --> 00:26:12.799
Uh interestingly, Claude's integration with Microsoft is much better than the copilot's integration with the Microsoft stack.
00:26:13.039 --> 00:26:14.400
That's and that's happened really quickly.
00:26:14.720 --> 00:26:15.680
It has, yes.
00:26:16.000 --> 00:26:16.720
You know why?
00:26:16.960 --> 00:26:17.200
No.
00:26:17.759 --> 00:26:23.839
So Cloud focused on one thing, which was to make their models really good at uh coding.
00:26:23.920 --> 00:26:32.960
Uh and so what happens is that Claude models or Opus really knows how to write, think through a problem and write a good code.
00:26:33.200 --> 00:26:37.119
The the thing is that everything that happens on a computer uses code.
00:26:37.759 --> 00:26:43.200
What it means is that for it to create a Word document, it doesn't need to know how to use Microsoft Word.
00:26:43.440 --> 00:26:48.000
It needs to know how to write scripts that can actually execute that Word document.
00:26:48.240 --> 00:26:56.079
So because it is so good at coding, it can interpret the code of any software it's working in, and then it can actually make it happen.
00:26:56.319 --> 00:26:58.720
Uh whereas like open AI is clear.
00:26:58.880 --> 00:26:59.440
It's really clever.
00:26:59.599 --> 00:27:02.400
So it's it's effectively giving you access to the foundations.
00:27:02.640 --> 00:27:02.799
Yes.
00:27:03.119 --> 00:27:03.759
What's driving this?
00:27:04.079 --> 00:27:04.640
That's right.
00:27:04.880 --> 00:27:08.079
So like the two big shifts that has happened in the AI space.
00:27:08.240 --> 00:27:13.279
First is that something non-human technology was able to understand human language.
00:27:13.440 --> 00:27:21.599
So it was, it made it really easy for us to teach it anything or train it on this large language.
00:27:21.759 --> 00:27:27.039
And the second thing is for it the ability for it to be able to write code or understand code.
00:27:27.200 --> 00:27:30.720
Because what it means is that it can now do anything on the computer as well.
00:27:31.119 --> 00:27:37.279
So it can now understand human language really well, it can do anything on the computer really well, and that just closes the gap.
00:27:37.680 --> 00:27:38.000
Yeah.
00:27:38.240 --> 00:27:42.079
It's it has been uh fascinating to see that.
00:27:42.160 --> 00:27:48.799
Have has the conversation around safety of these AI models and the security of these AI models changed from your perspective?
00:27:48.960 --> 00:27:56.960
Because looking from the outside in, um you know, for for those of you that are joining me on the on the dumb monkey side of things.
00:27:58.000 --> 00:27:59.839
I think it's an interesting conversation though.
00:27:59.920 --> 00:28:03.519
Like we've it's uh it's been a long time since we have done this.
00:28:04.559 --> 00:28:12.400
It is and the it It does feel like a lot of the a lot of the issues that we were particularly concerned, say this time last year, have changed.
00:28:13.200 --> 00:28:31.839
So we would previously, if we were doing this, if we were doing this podcast, having this conversation six months ago, we would have been, or eight months ago, we would have been talking about um concerns around the security and uh protecting basically your use of some of these models, particularly in a business sense, but really in a life sense.
00:28:32.000 --> 00:28:36.880
Um if you're using the free version of ChatGPT, for instance, be be concerned about what you're putting on there.
00:28:37.200 --> 00:28:39.279
A lot of our conversations were about that.
00:28:39.519 --> 00:28:56.240
It feels like there's been an awareness perhaps from the developers of these models and the and the teams that are driving the development of them that if you want our if you want our money, if you which is you know, you are the one I use of the of the platform, if it's free.
00:28:56.319 --> 00:29:00.160
You know, the golden rule is if it's free, if you you're not paying for the product, you are the product.
00:29:00.319 --> 00:29:04.640
Um so they either want you using the product or they want you paying for the product and they want you to keep coming back.
00:29:04.720 --> 00:29:14.319
These are business models and business models with some very big, very big costs and bank behind them, some incredibly high um business appraisals.
00:29:14.400 --> 00:29:19.599
You know, what we're saying, these businesses that work, they're piling money into big data centers, they need us to keep being involved with them.
00:29:19.920 --> 00:29:20.319
That's right.
00:29:20.559 --> 00:29:27.920
So, as a business decision from them, it makes very good sense to make us as users feel more secure about using these models.
00:29:28.079 --> 00:29:28.240
Yes.
00:29:28.480 --> 00:29:32.240
That's what I think we're all saying the same things, going, how do we trust it?
00:29:32.400 --> 00:29:32.960
Yeah.
00:29:33.279 --> 00:29:37.119
Has there been real movement on that, or do I just feel like there's been movement on that?
00:29:37.680 --> 00:29:52.160
I think what has happened is that uh because uh a lot of people are now aware of the privacy settings, uh, and initially there were the mod like the systems have improved as well.
00:29:52.240 --> 00:29:59.119
Like Claude has got a SOC2 certification, these guys needed to move into enterprises, so the questions were asked.
00:29:59.200 --> 00:30:05.440
So definitely organizations themselves have become more mature in terms of how they handle the data.
00:30:05.759 --> 00:30:10.799
People now understand the difference between a free version and uh in a pro version as well.
00:30:10.880 --> 00:30:15.680
So I think there's definitely been a moment in that from that perspective.
00:30:16.000 --> 00:30:19.519
Uh, and also these models have become more safer as well.
00:30:19.680 --> 00:30:28.880
Like if you could really do a prompt injection in like earlier models and uh get information out, these models are becoming more and more safer uh as well.
00:30:29.119 --> 00:30:42.799
So there's definitely, and I think uh also I in Australia what has happened is that because a lot of businesses have been conscious, so they have spent a lot of uh efforts in doing the governance right and policies right.
00:30:42.960 --> 00:30:49.599
I've seen the businesses where they have actually made sure that uh people are aware of the risk and challenges and so on.
00:30:50.000 --> 00:30:55.039
And now so we are in a position to actually start to leverage some of these tools easily.
00:30:55.680 --> 00:30:56.319
Yeah.
00:30:58.400 --> 00:30:59.599
It is interesting.
00:30:59.920 --> 00:31:03.119
Um what do you think is coming next?
00:31:04.079 --> 00:31:06.079
So what do you know is coming next?
00:31:06.480 --> 00:31:12.079
Okay, so we can talk about features, like there's been so many features all the time.
00:31:13.200 --> 00:31:24.720
What has happened is now it's actually shifted the whole like last year the conversation was okay, like this chat GPD is great, it provides good answers, not, and so on.
00:31:25.359 --> 00:31:31.920
But we are saying okay, agents are coming, or some like technical people would have those agents.
00:31:32.319 --> 00:31:37.279
But now the systems themselves have got agentic capabilities, right?
00:31:37.519 --> 00:31:48.559
So what it means is that a lot of work that we were actually doing by ourselves is can be delegated to these systems.
00:31:48.799 --> 00:32:01.759
And so as we go, and then Cloud has been uh launching Cloud for Legal, Cloud for They have just recently launched Cloud for Small Business that gives small business a lot of tools to connect to their systems.
00:32:01.920 --> 00:32:08.319
Uh so what happens is that uh these tools uh are becoming really powerful.
00:32:08.559 --> 00:32:16.799
Uh so first of all, there's a very clear advantage for the businesses who are embracing it because they're able to do the work really fast.
00:32:16.880 --> 00:32:19.680
So there's definitely that's divide forming.
00:32:19.920 --> 00:32:32.400
And then also in terms of the people who are embracing AI versus who are just using it for uh correcting their emails and so on, that gap has started to widen a bit as well.
00:32:32.720 --> 00:32:33.359
Yeah.
00:32:35.279 --> 00:32:37.039
Has the cost of using AI gone up?
00:32:37.200 --> 00:32:41.920
And are we paying enough attention to what's going to happen to our budget models?
00:32:42.000 --> 00:32:48.880
Because as you mentioned, you know, a lot of these models are fairly heavily discounting the availability of these tools.
00:32:49.039 --> 00:32:51.119
I mean, this is a classic business model.
00:32:51.599 --> 00:33:00.640
We've developed these incredible things, we've poured an enormous amount of resource into it, we've now got a you know, we've got a value, our company valuation is over here, our cost base is here.
00:33:01.119 --> 00:33:09.839
Anyone who's in the the stock market sector economists, um, financial futurists will all sit there and go, none of these companies are worth the valuations that they're worth.
00:33:10.000 --> 00:33:14.799
Their cost of going into development has to be recouped at some point.
00:33:15.119 --> 00:33:15.680
Yes.
00:33:16.160 --> 00:33:18.559
Are we all about to hit a big AI bill?
00:33:18.720 --> 00:33:19.839
Do we need to have this conversation?
00:33:20.240 --> 00:33:25.200
I think it's a very important conversation and also deserves its own space as well.
00:33:25.599 --> 00:33:27.599
Uh but it's a very great point.
00:33:27.759 --> 00:33:35.200
Uh, one thing uh to understand here is that a lot of subscriptions that we are getting, these are being heavily subsidized.
00:33:35.359 --> 00:33:45.680
So, like the usage that we get in a$30 Claude or Chat GPT subscription, it generally would cost uh them uh$100 or more than that.
00:33:45.920 --> 00:33:53.839
So similarly, what's what they are doing is they are actually heavily subsidizing that cost with the money that they are raising because they want more and more people to use AI.
00:33:54.000 --> 00:34:02.000
So the actual cost of tokens has not decreased, uh, and there's not been much work that has been done on that part.
00:34:02.079 --> 00:34:04.400
There have been some other models that have come out.
00:34:05.119 --> 00:34:08.719
And by tokens, you mean effectively the amount that you pay that gives you the amount of computing.
00:34:09.119 --> 00:34:09.360
Yes.
00:34:09.519 --> 00:34:15.599
So token is basically uh the the currency in which you measure the usage of AI.
00:34:16.079 --> 00:34:19.039
So one token is generally two words.
00:34:19.360 --> 00:34:22.480
Uh so what happens is really I never knew that.
00:34:22.719 --> 00:34:22.880
Yes.
00:34:23.199 --> 00:34:23.679
That's a token.
00:34:24.079 --> 00:34:26.880
Yes, and I'm saying that nervously because I always get confused.
00:34:27.119 --> 00:34:33.519
Either it's one word is two tokens or two tokens is one word, but it's very, very close to that.
00:34:33.679 --> 00:34:34.159
Yes.
00:34:34.480 --> 00:34:44.320
But uh what happens is that every time you send a message to AI or receive something, that's actually uh calculated in terms of how many tokens there are.
00:34:44.639 --> 00:34:44.800
Okay.
00:34:45.119 --> 00:34:52.159
And so no matter whether you are paying for token or not, the company that's processing those tokens have to pay for it.
00:34:52.400 --> 00:34:54.960
There's a compute that's actually happening over there.
00:34:55.440 --> 00:34:59.599
So we're gonna take that on night so I can that that will be a really interesting conversation.
00:34:59.840 --> 00:35:00.079
Yes.
00:35:00.239 --> 00:35:06.719
So it's it's the bill, the AI bill is definitely be there, and we have not solved the power of the energy that's actually required.
00:35:07.199 --> 00:35:09.360
Water use and compute and all of that.
00:35:10.480 --> 00:35:16.079
But there should be some people who should be working on it, and there's a this this is a problem to be solved as well.
00:35:16.320 --> 00:35:16.639
Yeah.
00:35:16.880 --> 00:35:25.280
Uh but it should not uh defer you as a business person or as an individual from actually using AI.
00:35:25.440 --> 00:35:37.519
Yes, the the cost might increase, but now is the time for you to use AI as a playground where you can experiment, learn, and fail and like burn tokens because they are being heavily subsidized.
00:35:37.840 --> 00:35:43.039
Because probably in one year the that$20 subscription might be$500.
00:35:43.440 --> 00:35:43.920
Oh wow.
00:35:44.000 --> 00:35:44.159
Yeah.
00:35:44.400 --> 00:35:49.039
But the thing is, by that time, we should have sorted out like what's the wastage.
00:35:49.199 --> 00:35:51.280
Like right now, we do a lot of wastage.
00:35:51.440 --> 00:36:00.320
Not in normal conversation, are generally cheaper, but when you're developing apps and developing systems, designing things, you burn like thousands of tokens.
00:36:00.559 --> 00:36:05.119
Sorry, not even thousands, like everyone burned a thousand tokens in the thing.
00:36:05.199 --> 00:36:09.519
Like we're we're talking sort of like millions of tokens and sort of developing something like that.
00:36:09.679 --> 00:36:11.440
And I I burn a lot of tokens.
00:36:11.920 --> 00:36:16.719
Um, one of one month my bill of AI was around fifteen thousand dollars.
00:36:16.960 --> 00:36:17.440
Wow.
00:36:17.679 --> 00:36:23.039
Because I just and uh all of the APIs, uh consumption, and so on.
00:36:23.280 --> 00:36:23.519
Yeah.
00:36:23.679 --> 00:36:23.920
Yes.
00:36:24.159 --> 00:36:28.960
I've got a little image in my brain of you just sitting there just you know, like your little AI fire just burning money.
00:36:30.400 --> 00:36:31.840
But learning and developing volume.
00:36:32.239 --> 00:36:42.159
Learning and developing, and like uh I was able to develop uh an app on on a weekend and like two two nights from that and spend and how long would that app have taken you previously?
00:36:42.480 --> 00:36:44.239
I asked my team for an estimate.
00:36:44.480 --> 00:36:45.760
I've got a very skilled team.
00:36:45.920 --> 00:36:47.519
They said three and a half months.
00:36:47.840 --> 00:36:48.159
Wow.
00:36:48.320 --> 00:36:48.559
Yes.
00:36:48.880 --> 00:36:50.559
Yeah, so time is money, right?
00:36:50.880 --> 00:36:52.239
There's there's equations in there as well.
00:36:52.480 --> 00:36:52.639
Yes.
00:36:52.880 --> 00:36:53.599
What about Google?
00:36:53.840 --> 00:37:03.599
Google, I I'm always interested by um mainly because I come from that sort of you know media marketing background, and they call themselves Google Deep Mind.
00:37:03.679 --> 00:37:07.679
They've got an incredible origin story, and they just have the best marketing in AI.
00:37:07.840 --> 00:37:09.280
So, you know, I was already a fan.
00:37:09.360 --> 00:37:13.199
But no, in but but in all seriousness, Google I find is really interesting.
00:37:13.280 --> 00:37:17.599
How they position themselves, what they're developing in terms of their AI, their origin story.
00:37:17.679 --> 00:37:21.440
For anyone who hasn't seen the Google Deep Mind documentary, go and watch it.
00:37:21.519 --> 00:37:23.599
It's just it's amazing.
00:37:23.760 --> 00:37:25.599
And it's just such a story of our time.
00:37:25.840 --> 00:37:28.000
Like it's going to become one of those iconic stories.
00:37:29.039 --> 00:37:37.519
But they're interesting because Google were early providers of AI, they were early developers of AI.
00:37:37.679 --> 00:37:42.480
Then they kind of slipped behind, they lost a fair bit of you know, market share, that became kind of uncool.
00:37:42.719 --> 00:37:45.519
You know, you didn't really use Google for AI anymore.
00:37:45.679 --> 00:37:48.719
Um then they did the wonders of nano banana, best name in AI.
00:37:48.800 --> 00:37:48.960
Still.
00:37:49.679 --> 00:37:51.039
It's gonna be a hard one to top.
00:37:51.199 --> 00:37:53.760
Um but of course the image generation with nano banana.
00:37:53.920 --> 00:37:55.840
But now they're doing something really interesting.
00:37:56.079 --> 00:38:04.320
And I've noticed that Google AI is starting to get integrated into a little bit of everything and the the everyday things that we use.
00:38:04.480 --> 00:38:15.119
So we've spent a fair bit of conversation talking about Claude and the and the magic that they've created, and it absolutely is magic because I don't understand how it how it works, so I'm calling it magic.
00:38:15.360 --> 00:38:18.320
Um so magic is just signs that you don't understand.
00:38:18.719 --> 00:38:31.440
And but Google is putting their AI into Gmail, into Google Maps, into all of these things that are just kind of ubiquitous.
00:38:31.599 --> 00:38:33.760
We don't talk about them, we don't think about them anymore.
00:38:33.920 --> 00:38:44.480
Um that grand old thing that I'm gonna steal a line from Professor Professor Raj Basser at um at Deacon that I work with, and not sponsored by Deacon, completely unaffiliated.
00:38:44.639 --> 00:38:45.599
My work here is separate.
00:38:46.239 --> 00:38:47.199
But we love Deacon.
00:38:47.679 --> 00:38:48.480
But we love Deacon.
00:38:48.719 --> 00:38:48.960
Yes.
00:38:49.440 --> 00:38:50.639
Um always.
00:38:50.800 --> 00:38:55.920
But Raj has this great line, and he says if you notice if you notice the technology, it probably means it's broken.
00:38:57.039 --> 00:39:01.199
And with Google, it's become so ubiquitous that we don't notice that technology anymore.
00:39:01.280 --> 00:39:01.840
It's just kind of there.
00:39:01.920 --> 00:39:05.599
It's on our phones, it's on our laptops, it's in our, you know, in our area.
00:39:06.239 --> 00:39:10.320
And now all of a sudden we've got Google AI coming into these.
00:39:10.559 --> 00:39:11.440
What are they doing?
00:39:11.679 --> 00:39:12.400
What are you seeing?
00:39:12.960 --> 00:39:22.639
So if you go back to the history of how Gen AI came about, there were a few guys who were trying to solve a problem with Google Translate.
00:39:22.800 --> 00:39:28.400
The problem with Google Translate was it was able to translate word to word, but could not connect the sentences.
00:39:28.719 --> 00:39:44.320
So they then did research and wrote a paper on transformers about how rather than you have to, rather than having reading word to word, uh, you can actually read the full document, pick up the words, and then predict what is around that.
00:39:44.800 --> 00:39:57.280
And then that actually changed the the whole thought process of okay, right now, if you had to train something on uh on a model, what you have to do is you have to go through word to word, it would take a lot of time.
00:39:57.440 --> 00:40:00.719
So OpenAI, what they did was they said, okay, this is really good.
00:40:00.880 --> 00:40:05.599
We are gonna use this transformer model, and because Google published the paper, right?
00:40:05.679 --> 00:40:06.719
So they just put it out.
00:40:07.039 --> 00:40:14.559
And they said, we are going to train it all on the all of the the whole world's knowledge, and that's how the large language model concept was born.
00:40:14.800 --> 00:40:23.039
Then they said, okay, if we can use prediction to be able to understand what's in the text, we can use it to predict the next term as well.
00:40:23.280 --> 00:40:31.679
So coming back, it was Google who actually came up with the actual concept of transformers from where the whole LLM was was generated.
00:40:31.920 --> 00:40:43.599
What Google has is an ecosystem of the penetration in the devices, the systems that we use, uh, and it's it's already out there.
00:40:43.840 --> 00:40:46.079
Now we are talking about a cloud over here.
00:40:46.239 --> 00:40:54.159
The beauty of the cloud is not in the model itself, like because models are gonna come and go, it's what they have built around it.
00:40:54.320 --> 00:40:55.440
It's a toolkit.
00:40:55.599 --> 00:41:00.960
But the thing is that right now, Google, and we're gonna we should talk about Apple at some stage as well.
00:41:01.199 --> 00:41:09.519
They're just sitting there laughing or smiling because the features that Claude has built, they can actually be built by their developers in a couple of weeks.
00:41:09.760 --> 00:41:14.079
Because uh Claude has actually developed some of those features in weeks as well, with with with the clouds.
00:41:14.239 --> 00:41:15.679
I'm starting to wonder what I do with my dice.
00:41:15.920 --> 00:41:16.480
Yeah.
00:41:17.119 --> 00:41:28.000
So so they are they're just watching uh this space from the consumer perspective and sort of like seeing what's happening, what's working for Claude, and what's not working for Claude.
00:41:28.079 --> 00:41:31.119
And Claude is trying to establish its presence in the market.
00:41:31.360 --> 00:41:35.679
And I believe Claude would have its own fanship and so on.
00:41:35.920 --> 00:41:41.119
But because of Google's own ecosystem, it still owns the search space.
00:41:41.280 --> 00:41:44.159
They've started to use the AI mode in there as well.
00:41:44.400 --> 00:41:54.800
Like they have they they own the Google Home space, there's a whole Android space that's out there, and if you look at the level of AI in uh Android phones, uh that's just amazing.
00:41:54.880 --> 00:41:57.280
It's gonna become a more native feature as well.
00:41:57.440 --> 00:42:02.400
And it's still a very prominent device that we actually use uh in in this.
00:42:02.559 --> 00:42:17.199
Whereas Apple has taken a stance where because they provide a very definitive experience to the people, the whole AI they feel is is in a bit of like still a sandbox mode where we're still going through testing and iterations.
00:42:17.519 --> 00:42:26.559
Did I read somewhere that that Google have made a deal with Apple though, that that future iterations of the iPhone are going to be using Google AI?
00:42:26.960 --> 00:42:27.760
I'm not aware of that.
00:42:28.719 --> 00:42:29.440
Go and have a check about that.
00:42:29.599 --> 00:42:30.719
You'll you'll read more into it than I will.
00:42:30.960 --> 00:42:31.119
Right.
00:42:31.519 --> 00:42:36.559
But I I believe in this was in the context of the money.
00:42:36.719 --> 00:42:37.760
Yeah, chasing the money.
00:42:38.079 --> 00:42:49.119
Which is um, my background for for quite a number of years is how we met was in business journalism, and that's that's the secret source of business journalism.
00:42:49.360 --> 00:42:51.039
There's no great secret to it, you just chase the money.
00:42:51.360 --> 00:42:54.480
And that that tells you kind of where people's decision making will go.
00:42:55.119 --> 00:43:00.400
And so I was really interested to see what was happening because you have these huge valuations, huge cost base.
00:43:01.519 --> 00:43:11.920
How is a company that's you know, I'm I'm making up figures, but how is a company that's bringing in$100 billion valued at three trillion?
00:43:12.719 --> 00:43:19.360
You know, we these things are are real when they need to be in in the IO space they haven't been.
00:43:19.920 --> 00:43:22.079
So there will there will be a reality check.
00:43:22.159 --> 00:43:23.119
We know that.
00:43:23.360 --> 00:43:25.679
But but will there be a reality check?
00:43:25.920 --> 00:43:30.000
Would it wouldn't you think that market would adjust itself that and these organizations will be able to do that?
00:43:30.400 --> 00:43:34.239
Maybe maybe the productivity will come up underneath it to support that, but yeah.
00:43:35.440 --> 00:43:59.360
There's gonna need to be some big changes in terms of value and where the any of this organization actually f reaching to a superagent stage where agents are able to create agents themselves and they are able to do months of research in in in seconds or hours, the amount of uh value that these organizations would hold would be amazing.
00:44:00.239 --> 00:44:04.639
So they these organizations would probably start to engulf other organizations.
00:44:04.719 --> 00:44:09.440
Like OpenAI have already started to sort of like offer their consulting services.
00:44:09.679 --> 00:44:21.199
And when I when we were joking about Claude killing the services, every time Claude launches a feature, uh a valuation of a company that's providing that service goes down.
00:44:21.440 --> 00:44:25.519
So we have we have seen like Canva's valuation or shares going down.
00:44:25.599 --> 00:44:27.679
We have seen Figmas shares going down.
00:44:27.840 --> 00:44:32.320
So all of these organizations that build their business on providing a service.
00:44:32.639 --> 00:44:33.760
And being a market leader in that.
00:44:34.079 --> 00:44:34.880
Being a market leader.
00:44:35.119 --> 00:44:43.599
And Cloud's developer would wipe code something because they use AI to decode, yes, and they they would do something in a week and launch a new feature.
00:44:43.920 --> 00:44:46.239
And then that just disrupts the whole industry.
00:44:46.400 --> 00:44:46.639
Yeah.
00:44:47.039 --> 00:45:01.199
So you can say that the valuation is inflated, but uh I feel the the the disruption that's gonna cause to industry uh is is gonna be so huge that that valuation would become meaningful.
00:45:01.519 --> 00:45:02.159
Yeah, okay.
00:45:02.639 --> 00:45:08.480
It could be the case that some of these companies might die and and and and and there would be like a few winners out of this.
00:45:08.800 --> 00:45:18.079
And new ones will be reborn and that's as you would expect in the in the in the early in the early days of evolution, things are you know, things get messy.
00:45:18.559 --> 00:45:18.800
That's right.
00:45:19.199 --> 00:45:23.280
Or kind of explosions, and then at the end of it you you reach a some form of equilibrium.
00:45:23.519 --> 00:45:27.360
Um I was very interested to see that Google's ad revenue is rising.
00:45:27.920 --> 00:45:31.280
It would because uh search value is going down as well.
00:45:31.599 --> 00:45:31.760
Yeah.
00:45:32.000 --> 00:45:32.639
Yes, yeah.
00:45:32.719 --> 00:45:35.360
And Chat GPT has introduced ads as as well.
00:45:35.599 --> 00:45:35.840
Yeah.
00:45:36.079 --> 00:45:36.719
And and there.
00:45:37.199 --> 00:45:41.920
No, Google's Google's ad revenue is is rising, so they're they're making more money.
00:45:42.159 --> 00:45:42.400
Yes.
00:45:42.559 --> 00:45:43.920
I kind of thought that they'd be losing money.
00:45:44.239 --> 00:45:52.239
No, they they are making more money uh because uh first of all, the the the search is actually uh decreasing.
00:45:52.480 --> 00:46:00.480
So what's what's what's happening is that uh businesses are more compelled to actually do ads as well.
00:46:01.039 --> 00:46:06.079
And also what we are seeing at the moment is we haven't seen the actual shift.
00:46:06.320 --> 00:46:21.119
Like I still use Google for search because I'd talked about using perplexity, but then I just got like it was too much because sometimes I just wanted to get an answer or get get get get to a website, and perplexity experience was not that great, right?
00:46:21.199 --> 00:46:23.599
So I still use Google for a lot of my searches.
00:46:24.239 --> 00:46:29.519
You haven't we haven't seen that shift from people moving away from Google to actually using the stuff.
00:46:29.679 --> 00:46:37.840
Yeah, but secondly, Google's ad revenue doesn't only come from the search revenue, it comes from the placement that happens in their partner websites and in the ecosystem.
00:46:38.000 --> 00:46:38.159
Yeah.
00:46:38.400 --> 00:46:40.320
And Google's ecosystem is pretty much.
00:46:40.480 --> 00:46:40.800
It's huge.
00:46:40.880 --> 00:46:42.000
It's really ubiquitous, isn't it?
00:46:42.239 --> 00:46:42.320
Yeah.
00:46:42.559 --> 00:46:42.800
Yeah.
00:46:42.880 --> 00:46:43.039
Okay.
00:46:43.119 --> 00:46:43.920
Well these are interesting.
00:46:44.400 --> 00:46:48.320
Why do you think that uh ad revenue is increasing from your perspective?
00:46:48.719 --> 00:46:51.280
Well, do I think it's I think it's I think it's similar.
00:46:51.440 --> 00:47:03.599
I think it's a case of of the the driver to say if you're in a business and you're trying to get people's attention, then this is a way to actually get value out of it.
00:47:03.840 --> 00:47:08.480
Because I think for a period of time doing things like Google Ads was a bit like a bit of a gamble.
00:47:09.360 --> 00:47:18.079
So you can you can put your your Google ads in there, you'll get sponsored up, people would completely kind of ignore the sponsored ads, go past.
00:47:18.320 --> 00:47:23.199
So people don't particularly like as people we don't like, ads that tell us what to do.
00:47:23.360 --> 00:47:29.440
I think as consumers and as searches, I think we just look straight at the the sponsored things and go straight past them.
00:47:29.599 --> 00:47:33.119
But I think there's more value now in where it is.
00:47:33.280 --> 00:47:35.039
And I think the advertising is getting a lot smarter.
00:47:35.119 --> 00:47:40.239
I mean, AI algorithms mean that you can start to personalize and so I've connected my Google ads to my cloud as well.
00:47:40.559 --> 00:47:42.239
Claude doesn't have a direct connector.
00:47:42.400 --> 00:47:49.679
There's a thing called Composio through which you can connect your like it's like a it's like uh standard MCP protocol.
00:47:49.920 --> 00:47:51.280
So I've connected my Google Ads.
00:47:51.679 --> 00:47:54.400
Sorry, what's a standard MCP protocol?
00:47:54.559 --> 00:47:59.760
What's which is uh an easy way to connect with these systems.
00:47:59.920 --> 00:48:06.639
So MCP model context protocol is a way to connect your cloud with any third-party systems.
00:48:07.039 --> 00:48:16.239
And uh some if you go to connectors in Cloud, browse the marketplace, a lot of connectors are available, but there are some connectors that are not available.
00:48:16.400 --> 00:48:18.559
But there's a Composio connector available.
00:48:18.639 --> 00:48:24.719
Then you go to Composio website, composio.dev, and then you can connect multiple Google accounts.
00:48:24.960 --> 00:48:28.159
I'm totally not doing this, but I'm interested in meta accounts and so on.
00:48:28.719 --> 00:48:32.159
And then you can just say, hey, uh Claude, look at my campaign.
00:48:32.239 --> 00:48:33.599
What am I doing wrong here?
00:48:33.840 --> 00:48:36.079
You do you want to set up a new campaign for me?
00:48:36.159 --> 00:48:38.000
And Claude can set up the campaign for you.
00:48:38.480 --> 00:48:41.280
Claude can set up the Facebook ads for you as well.
00:48:41.519 --> 00:48:55.199
It's much better than actually uh, you know, dealing with a third party to actually run not in every case, but no, but it it you know, I as you you asked the question about why they're getting better.
00:48:55.280 --> 00:48:56.960
I think I think their offering is getting better.
00:48:57.119 --> 00:48:58.559
I think that's why they're making more money.
00:48:58.800 --> 00:49:06.800
So I think that you know, back when it would be if you were searching for a a pet groomer, um, I'm not searching for a pet groomer.
00:49:06.880 --> 00:49:09.039
I've got a French bulldog, no one wants to groom that.
00:49:10.480 --> 00:49:17.679
Um but you'd put in an ad for pet groomer and it can take you to somewhere in Perth or somewhere in WA, or it it's not very useful.
00:49:17.760 --> 00:49:23.760
If I'm sitting here in in slightly cloudy, foggy Geelong, I want a pet groomer that's just down the road.
00:49:23.840 --> 00:49:25.920
And ideally I want them to be even closer than that.
00:49:26.400 --> 00:49:36.000
So for all of the wonders of being able to provide the world of options, I think for most of us as consumers, when we're looking for something, we kind of want to find something that's really practical for us right now.
00:49:36.239 --> 00:49:38.079
And we want to do that in the easiest possible way.
00:49:38.320 --> 00:49:53.119
So as advertisers, uh as an advertising platform that I think you know, in a way Google is, um the better they are at providing people what they want up front makes the offer to advertise on them a little bit better.
00:49:53.199 --> 00:50:00.960
And I assume if you can pinpoint where you're trying to target your advertising, if you're not trying to hit the whole world and you're trying to target a smaller area, you're probably spending less money.
00:50:01.280 --> 00:50:01.519
100%.
00:50:02.159 --> 00:50:03.840
So it becomes more accessible as well.
00:50:04.000 --> 00:50:07.039
Um and that's something that I've seen in with meta ads as well.
00:50:07.199 --> 00:50:12.559
Earlier in meta ads, you have to do a lot of thinking about who do you want to target and so on.
00:50:12.719 --> 00:50:26.159
Now, official recommendation from Meta is you just put what is your offering, work on your creative and the offer, and leave it on Meta to determine the right people who are going to be receiving the message.
00:50:26.400 --> 00:50:28.320
And Meta gets it so well.
00:50:28.719 --> 00:50:33.920
Like you don't have to really do a lot of that targeting because MetaZI can do that.
00:50:34.320 --> 00:50:37.840
So I think with the help of AI, these advertising tools are getting better as well.
00:50:38.880 --> 00:50:42.960
So we asked you the question of what do you think is is coming next.
00:50:43.280 --> 00:50:44.719
I want to frame that in a different way.
00:50:44.800 --> 00:50:49.280
It's can we really know if what's coming next?
00:50:50.079 --> 00:50:51.679
Because these changes come hard and fast.
00:50:51.760 --> 00:50:58.239
I mean, Claude felt like I I know it wasn't, but it felt like it came out of nowhere and became a dominator really quickly.
00:50:58.719 --> 00:51:01.360
Um could it be Google in a couple of months?
00:51:01.440 --> 00:51:03.360
Are they cooking up something and that can happen?
00:51:03.599 --> 00:51:05.599
Can open AI make a resurgence?
00:51:05.920 --> 00:51:10.079
Can there be a new player in the field that's gonna step up and start to disrupt everything?
00:51:10.400 --> 00:51:11.760
Can we really know what's coming?
00:51:12.000 --> 00:51:12.800
That's a good question.
00:51:12.960 --> 00:51:20.559
So we cannot really know for sure what's coming, especially with uh what models are gonna come and what advancements are gonna come.
00:51:20.719 --> 00:51:34.800
But we can definitely make some hypothesis very confidently about where the world is going and some things that we are we are sure can say what's gonna happen or what's not gonna happen.
00:51:34.960 --> 00:51:39.840
So the first thing is that we are not ever going to go back to a world without AI.
00:51:40.079 --> 00:51:46.159
So it's it's not it's not a it's not a hype, it's not toothpaste is out of the tube, you know, yes.
00:51:46.239 --> 00:51:49.840
It's it's it's not uh bubble, it's not going to burst.
00:51:50.000 --> 00:51:53.119
No, it's valuations can decrease or so on.
00:51:53.360 --> 00:51:56.239
It's uh something that is here to stay.
00:51:57.199 --> 00:51:59.599
And outside of the tech industry, do people really care?
00:52:00.159 --> 00:52:05.840
I mean, do people care if the AI is open AI, if it's anthropic, if it's Google?
00:52:07.119 --> 00:52:07.920
They don't care.
00:52:08.079 --> 00:52:12.559
Like as long as uh they get their work done, that's what what they're uh worried about.
00:52:12.800 --> 00:52:19.760
And the second reality we know is that the business, as we knew earlier, it has shifted, it has changed.
00:52:19.920 --> 00:52:24.079
I the conversation that I have with businesses everywhere now.
00:52:24.239 --> 00:52:28.239
Like last year I was saying Australian businesses are sleeping, they are not sleeping anymore.
00:52:28.400 --> 00:52:37.440
They're awake, they're awake, they they know what's happening, and the the way we do business is changing and it's gonna change forever.
00:52:37.599 --> 00:52:42.400
We we know that we are seeing job marketing change, job the whole job market changing.
00:52:42.719 --> 00:52:48.639
The question is about when it's gonna happen, like and to what extent?
00:52:48.800 --> 00:52:52.480
We know there's gonna be a displacement, where it's gonna be a change.
00:52:52.719 --> 00:52:53.519
How much?
00:52:54.000 --> 00:53:00.320
Like, we know there's gonna be some business that will have an edge, some businesses would not have that edge because of the gap.
00:53:00.480 --> 00:53:01.199
But how much?
00:53:01.360 --> 00:53:03.599
Like, would those businesses be able to survive?
00:53:03.840 --> 00:53:07.440
Would we have some very prominent businesses and that's it?
00:53:07.760 --> 00:53:09.519
What's gonna happen to our jobs?
00:53:10.000 --> 00:53:19.599
Because we are just we like right now with the knowledge that we have, we make assumptions, okay, it's it's just about the knowledge workers and we still be able to delegate jobs and so on.
00:53:19.840 --> 00:53:21.760
But what does that displacement look like?
00:53:21.920 --> 00:53:29.920
What yes, and and that's a question that also keeps me awake uh like every day is like what is the limit?
00:53:30.079 --> 00:53:30.960
Where are we gonna stop from?
00:53:31.280 --> 00:53:33.199
How do you have extra time to be awake?
00:53:35.920 --> 00:53:38.400
Uh I'm I'm awake all the time nowadays.
00:53:38.559 --> 00:53:39.360
Yes, yeah, yes.
00:53:40.079 --> 00:53:44.719
Just sending sending things over to sending to my my Claude team.
00:53:44.960 --> 00:53:45.039
Yeah.
00:53:45.280 --> 00:53:45.840
Yes, yeah.
00:53:46.159 --> 00:53:50.000
Or possibly just, you know, spamming your mother-in-law.
00:53:50.320 --> 00:53:50.960
Yeah.
00:53:52.559 --> 00:53:56.880
But I hope at some point you are throwing in a few joke messages.
00:53:57.119 --> 00:53:58.000
I think that's important.
00:53:58.079 --> 00:53:58.960
Don't let that moment go.
00:53:59.119 --> 00:53:59.440
Uh yes.
00:53:59.840 --> 00:54:02.239
Because what we can't do in the age of AI is lose our humanity.
00:54:02.480 --> 00:54:03.360
I know, yes, yes.
00:54:03.599 --> 00:54:08.639
And a bit of ragging of your mother-in-law, that's that kind of feels like you've set that up very nicely.
00:54:10.079 --> 00:54:10.960
I'll play up Frank on the crazy.
00:54:11.280 --> 00:54:11.760
Sorry, Sarah.
00:54:11.840 --> 00:54:13.440
I'm very sorry for putting that in there.
00:54:14.079 --> 00:54:15.760
I'll definitely play up Frank on her.
00:54:16.079 --> 00:54:17.119
Yes, yeah.
00:54:17.440 --> 00:54:17.840
Yeah.
00:54:18.079 --> 00:54:27.039
So, but coming back to your questions, I just feel it's it's such a funny space at the moment.
00:54:27.199 --> 00:54:31.840
Like we are we've never been in such an uncertain uh world ever.
00:54:32.480 --> 00:54:36.719
And everyone is trying to predict and judge and make a hypothesis.
00:54:36.960 --> 00:54:42.239
We don't even know that even the worst hypothesis, like, are they even able to imagine what's gonna happen?
00:54:42.480 --> 00:54:46.719
Uh and when it's gonna happen, how fast, how how quick.
00:54:47.039 --> 00:54:53.039
There's a bit of race that's going on, these larger organizations, like uh they are in a different race.
00:54:53.280 --> 00:54:58.159
Businesses are trying to sort of try to sort of like work out how it means for the business.
00:54:58.320 --> 00:55:01.840
There's a tussle between governance and the speed and velocity.
00:55:02.079 --> 00:55:05.519
People are worried about their jobs, but they want to learn AI as well.
00:55:05.840 --> 00:55:08.000
There's excitement and fear.
00:55:08.239 --> 00:55:11.039
So it's it's a very uh different world to live in.
00:55:11.199 --> 00:55:11.440
Yeah.
00:55:11.599 --> 00:55:11.920
Yeah.
00:55:12.239 --> 00:55:13.840
Yeah, it's chaos rules, huh?
00:55:14.159 --> 00:55:14.960
It is, yes.
00:55:15.119 --> 00:55:15.360
Yeah.
00:55:15.599 --> 00:55:18.559
What's your advice to people in this world?
00:55:18.960 --> 00:55:25.760
Oh, my advice is the same advice it always has been, is stay curious and remember that you're a person.
00:55:26.480 --> 00:55:27.199
That's a good advice.
00:55:27.519 --> 00:55:32.880
Yeah, don't get confused by the don't get don't buy into the the idea that you can predict because you can't.
00:55:34.159 --> 00:55:36.559
We really just have to deal with with what's in front of us.
00:55:36.639 --> 00:55:40.880
And I think there's there's so much that's exciting about AI.
00:55:40.960 --> 00:56:03.039
There's so much promise in there to to really start to tackle some of the things that I mean we we talk about agents being able to take some of the pain points out of our everyday life, but if we can sort of get to the point where some of those blockers are out of the way and we can start to tackle real things and put human energy and creativity into things that can really add value, that to me is really exciting.
00:56:03.119 --> 00:56:11.199
And there's you know, I think there's so many challenges that we haven't even really put words to because they weren't really an option to do anything about.
00:56:11.760 --> 00:56:24.880
And you know, we're people, so we can sit here and philosophize for a certain amount of time and then you just still have to go and pay your bills and pick up the kids and you know, clean up after your French bulldog that does not clean up after themselves, and neither is AR gonna do it for me.
00:56:25.119 --> 00:56:28.559
But um so talking about that, I have a very interesting news to share.
00:56:29.119 --> 00:56:32.320
I've been working on a project called Agents for Humanity.
00:56:32.559 --> 00:56:37.039
Uh, it's a common place uh where if you have got a problem, it's not live yet.
00:56:37.360 --> 00:56:40.639
We are testing it with few not-for-profits uh NGOs.
00:56:40.880 --> 00:56:49.039
If you've got a problem around poverty, education, how to plan a city, you put the problem over there.
00:56:49.519 --> 00:56:52.880
Hundreds of how do we get people into homes they can afford in Australia today?
00:56:53.039 --> 00:56:53.119
Yeah.
00:56:53.360 --> 00:56:54.320
Intractable problems.
00:56:54.719 --> 00:56:55.199
Absolutely.
00:56:55.360 --> 00:56:56.880
You put the problems over there.
00:56:57.119 --> 00:57:02.639
Hundreds of agents actually work on that problem uh through a very clear design thinking process.
00:57:02.719 --> 00:57:07.840
They debate their problem, research it, and from there they try to devise a synthesis.
00:57:08.079 --> 00:57:18.480
There's like a council that's created, they vote on that uh final proposals as well, and from there, a few proposals are chosen, and then from there an implementation plan is created.
00:57:18.719 --> 00:57:21.440
And on that point, I have a very interesting news to share.
00:57:21.599 --> 00:57:25.440
So I've been working on a project uh with some really interesting people.
00:57:25.599 --> 00:57:27.679
It's called Agents for Humanity.
00:57:27.920 --> 00:57:32.159
So, as we know, there are so many problems in the world that are unsolved.
00:57:32.239 --> 00:57:34.000
Uh, there's a problem around education.
00:57:34.079 --> 00:57:38.480
How do you educate someone in rural India who's looking for quick outcomes?
00:57:38.639 --> 00:57:40.079
There's no playbook for that.
00:57:40.239 --> 00:57:43.519
Or how do you solve a problem here with Murabull Street?
00:57:44.000 --> 00:57:46.559
Or it could be about world hunger and so on.
00:57:46.880 --> 00:57:55.440
So, with Agents for Humanity, it's going to be an open platform where anyone, any not-for-profit, any individual can put a problem over there.
00:57:55.679 --> 00:58:10.639
And uh a team of hundreds of agents will follow particular roles uh to solve that problem, debate that problem, research on that, steel man it, and from there create proposals and try try to drive synthesis out of it.
00:58:10.800 --> 00:58:16.000
And once the synthesis is created, it's gonna create an action plan to actually solve that problem.
00:58:16.639 --> 00:58:20.480
So that uh project has been selected.
00:58:20.559 --> 00:58:25.599
Uh, there's an AI for good council uh uh that's actually supported by United Nations.
00:58:25.679 --> 00:58:27.519
So there's AI for goods amazing.
00:58:27.760 --> 00:58:31.440
Yeah, so they're uh they have got a pitching competition.
00:58:31.599 --> 00:58:34.320
So we are in like top five in Australia that has been selected.
00:58:34.400 --> 00:58:37.199
So I'm going to uh to congratulations.
00:58:37.440 --> 00:58:37.599
Thank you.
00:58:37.679 --> 00:58:38.079
Thanks for having me.
00:58:39.519 --> 00:58:40.159
That's something.
00:58:40.480 --> 00:58:43.519
Yeah, and it's just it was like just a harebrained idea.
00:58:44.000 --> 00:58:45.599
At three o'clock in the morning, presumably.
00:58:45.920 --> 00:58:46.079
Yeah.
00:58:47.599 --> 00:58:48.159
Have you got a name?
00:58:48.320 --> 00:58:50.320
So you've you've called it for humanity.
00:58:50.800 --> 00:58:51.519
Agents for humanity.
00:58:51.840 --> 00:58:52.639
Agents for humanity.
00:58:52.880 --> 00:59:02.000
But the key thing with this, like when people are put putting the problem and we are solving this solution, providing a solution, it requires a lot of compute.
00:59:02.079 --> 00:59:07.199
So we were calculating to solve a problem, you need like$6,000 to$8,000 worth of compute.
00:59:07.360 --> 00:59:07.519
Yeah.
00:59:07.760 --> 00:59:16.559
So the interesting thing is that as an individual who's uses Claude or OpenClaw or Chat GPT, you can send your agent in there to contribute.
00:59:16.800 --> 00:59:23.039
So as a part of your subscription, it's just going in there, can making a contribution and coming back.
00:59:23.280 --> 00:59:23.440
Yeah.
00:59:23.760 --> 00:59:35.599
So which means we as an organization, we like we can solve hundreds and thousands of problems with the help of the agents that people are sending or the compute that people already have, which is sitting unjuceded.
00:59:35.679 --> 00:59:35.840
Yeah.
00:59:36.079 --> 00:59:40.800
And we are using that energy or power to actually solve things for humanity.
00:59:41.039 --> 00:59:42.239
That's unbelievably cool.
00:59:42.400 --> 00:59:43.039
Congratulations.
00:59:43.360 --> 00:59:43.599
Thank you.
00:59:43.679 --> 00:59:44.239
Thanks a lot.
00:59:44.480 --> 00:59:46.880
How much how many people in your team have been working on this?
00:59:47.119 --> 00:59:51.039
Suzween, uh, like we have we are around uh six people who have been working on this.
00:59:51.280 --> 00:59:51.840
This is awesome.
00:59:52.079 --> 00:59:58.400
Yes, we have been working with five uh uh organizations, uh like a couple in India, three in Australia in different levels.
00:59:58.480 --> 01:00:00.159
We asked them to give us a problem.
01:00:00.320 --> 01:00:04.159
And right now we are just trying it with our own compute at the moment.
01:00:04.320 --> 01:00:08.960
We have got some really good commitments from some universities who are willing to give us computers.
01:00:09.599 --> 01:00:10.400
This is really exciting.
01:00:10.480 --> 01:00:17.840
Yes, but then the aim is to open it up on the world because the actual final synthesis and everything would be available for the world to access.
01:00:18.000 --> 01:00:24.320
So which means if a problem has been solved in Kenya, someone in India should have the access to that as well.
01:00:24.719 --> 01:00:32.559
So, I mean, when you when you ask me what's my advice, my advice is just that it's keep looking for the good.
01:00:33.199 --> 01:00:36.719
Because we can't we can't know what's gonna happen.
01:00:37.199 --> 01:01:01.440
And as much as we can sit there and worry about jobs and you know, AI taking over the world, and I've got I've got kids that are starting to look at what kind of career am I gonna have, and I can't tell them anything other than focus on things that help people, focus on things that you can add value, focus on things that you can enjoy, and just stay curious and be okay with the fact that you're not gonna know everything because we're not.
01:01:01.519 --> 01:01:02.880
This is all gonna be a bit of a ride.
01:01:03.519 --> 01:01:10.559
But what always gives me hope and what I always come back to is that there is this incredible power that we are developing.
01:01:10.719 --> 01:01:17.360
It won't be perfect, it won't get everything right because we as people, clearly, don't know whether we're not perfect and we don't get everything right.
01:01:17.920 --> 01:01:21.519
But there's the ability to solve things in ways that we haven't been able to do before.
01:01:21.679 --> 01:01:23.679
And that to me is the hope.
01:01:24.079 --> 01:01:33.760
And I think if collectively, if all of us at an individual level and then collectively keep remembering that we're people, then I think we're gonna be okay.
01:01:34.239 --> 01:01:35.039
We will be, yes.
01:01:35.280 --> 01:01:37.440
But you know, I'm a glass half full kind of person.
01:01:37.679 --> 01:01:39.280
And I love AI, it's really cool.
01:01:39.599 --> 01:01:40.639
So it's been exciting.
01:01:40.880 --> 01:01:41.280
Yes.
01:01:41.519 --> 01:01:43.280
Well, thank you for for coming back again.
01:01:43.440 --> 01:01:46.960
Thank you for inviting me back into this uh very cool podcast space.
01:01:47.039 --> 01:01:48.159
I've been really enjoying it.
01:01:48.320 --> 01:01:56.480
And thank you to everyone who has joined us once again for the Dumb Monkey podcast.
01:01:56.559 --> 01:01:57.679
It's cool to be back.
01:01:57.920 --> 01:01:58.320
Thank you.
01:01:58.400 --> 01:02:01.760
Um by the time we see you again, presumably the world of AI will have exploded.
01:02:01.840 --> 01:02:06.079
We will all be in a simulation, but we'll be back, hopefully, in the podcast studio anyway.
01:02:06.239 --> 01:02:07.440
So thank you, Amir.
01:02:07.679 --> 01:02:07.840
Thank you.
01:02:08.159 --> 01:02:08.960
And thank you all.
01:02:09.039 --> 01:02:11.119
We'll see you next time on the Dumb Monkey Show.
