WEBVTT
00:00:00.800 --> 00:00:05.370
for the most part, they don't use it because again, that's the discovery issue of what can you do with it?
00:00:05.570 --> 00:00:08.119
But then, the second thing is like, what should you be using it for?
00:00:08.380 --> 00:00:27.179
And there, I think like early Google, you'll get burned easily, maybe more easily if you trust it blindly, which is why you maybe don't want to roll it out as aggressively to everyone until, society takes all these nuances into account and just trains up the average member of the population that, it's a pretty.
00:00:27.179 --> 00:00:28.050
Sharp tool,
00:00:34.118 --> 00:00:40.027
Hello friends and welcome to another episode of the deep learning with poly AI podcast.
00:00:41.158 --> 00:00:42.048
I'm your host, Damien.
00:00:42.048 --> 00:00:45.417
I'm here with CEO and co founder of poly AI, Nicola Merchick.
00:00:45.417 --> 00:00:46.548
Hey Nicola, how you doing today?
00:00:47.057 --> 00:00:47.927
Hi Damien, not bad.
00:00:47.927 --> 00:00:48.258
How are you?
00:00:48.773 --> 00:00:49.912
I'm doing very well, thank you.
00:00:50.542 --> 00:01:01.173
I'm not actually Damien, I am a generative ai video and avatar that was created because I figure everyone else is using generative AI to do their work.
00:01:01.173 --> 00:01:01.652
Why not?
00:01:01.862 --> 00:01:02.582
I do the same.
00:01:02.823 --> 00:01:09.453
You and I talked about at the end of a recent recording, making a completely generative episode of the podcast that is coming up soon.
00:01:09.453 --> 00:01:10.742
I'm very excited.
00:01:11.117 --> 00:01:16.037
To see that come to fruition, but it got me thinking, Nicola, that, there are recent reports that came out.
00:01:16.037 --> 00:01:17.117
Adobe had one.
00:01:17.117 --> 00:01:24.397
I know that I know that McKinsey had one as well that talked about how consumer adoption of generative AI specifically.
00:01:24.688 --> 00:01:27.028
But just AI tools in general is at an all time high.
00:01:27.078 --> 00:01:27.688
People get it.
00:01:27.798 --> 00:01:28.918
They know what's great about it.
00:01:28.947 --> 00:01:29.787
They're trying it out.
00:01:29.787 --> 00:01:30.638
They're sampling it.
00:01:31.427 --> 00:01:37.998
And according to Adobe, some, more than half the people in America have at least tried generative AI tools once.
00:01:38.028 --> 00:01:42.378
And of those who use it regularly, more than half use it every single day.
00:01:42.768 --> 00:01:46.697
So suffice it to say generative AI has made itself entrenched.
00:01:47.197 --> 00:01:48.058
It is a tool.
00:01:48.097 --> 00:01:51.938
Obviously we go to work many of those days of every single day.
00:01:52.317 --> 00:01:53.427
And so the question is.
00:01:54.578 --> 00:01:56.638
How, what I wanted to talk about today is.
00:01:56.987 --> 00:02:03.257
How are we seeing employees along the hierarchy, along the various organizational tiers?
00:02:03.268 --> 00:02:07.257
How are we seeing employees today utilizing generative AI?
00:02:07.298 --> 00:02:10.638
And what are some signs that maybe it's getting overutilized?
00:02:10.867 --> 00:02:16.427
I know that recently you talked about when people use it or when it first came out and students were using it to cheat.
00:02:16.487 --> 00:02:19.987
They were having like GPT writing SS form and you're like, well, the way we do that.
00:02:20.247 --> 00:02:21.277
I think there still are, no?
00:02:21.902 --> 00:02:31.573
Yeah, probably non savvy professors that haven't caught up, but it's like, all right, so how do we take that to the next level and how do we say, all right, we recognize this as a tool in the roster.
00:02:31.573 --> 00:02:32.772
It's very, very powerful.
00:02:32.943 --> 00:02:34.362
Here's how you use it effectively.
00:02:34.362 --> 00:02:35.902
So I thought we'd chat about that today.
00:02:36.223 --> 00:02:40.513
First of all, where, what are some of the pros and cons as a generative AI expert?
00:02:40.513 --> 00:02:43.953
What are some of the pros and cons of having this technology in the modern workplace?
00:02:44.242 --> 00:02:46.913
What are some great use cases of not so great use cases?
00:02:47.223 --> 00:02:49.242
And maybe we'll just dive into some details from there.
00:02:49.293 --> 00:02:59.372
Look, I think that AI is a great equalizer, but then equally, I think it increases disparity between like people who can use technology well and those that can't, even further, right?
00:02:59.372 --> 00:03:07.372
So I imagine that when Google, became what it is, I think people who knew how to do that were like infinitely more productive than someone sourcing that information elsewhere.
00:03:07.473 --> 00:03:09.772
And yeah, they had to be sure that the information's right.
00:03:09.782 --> 00:03:11.802
They had to have a healthy dose of skepticism.
00:03:12.362 --> 00:03:16.703
But it gave them really helpful pointers really fast, right?
00:03:17.152 --> 00:03:18.973
So if you have those, then you're like, great.
00:03:19.043 --> 00:03:21.612
Okay I think that the answer is this.
00:03:21.893 --> 00:03:24.513
Now I'll go to my trusted sources and I'll just validate it.
00:03:24.622 --> 00:03:26.573
And if it happens to be right.
00:03:26.798 --> 00:03:27.087
Great.
00:03:27.177 --> 00:03:28.168
I'm done like way faster.
00:03:28.268 --> 00:03:38.538
So similarly, I think, the mundane part, and I think it's like it's hard to trust this data because I suspect that a lot of those people that use it every day, use it for a set of limited things.
00:03:38.548 --> 00:03:40.728
I know a bunch of people that use it for emails, right?
00:03:40.728 --> 00:03:43.198
They write bullet points and be like, explore this.
00:03:43.288 --> 00:03:44.647
It's a very nice verbose email.
00:03:45.018 --> 00:03:51.897
And honestly, if they send those bullet points over, that would probably be more efficient communication, but hey, pomp and circumstance, right?
00:03:53.328 --> 00:03:54.657
It still saves them time.
00:03:55.182 --> 00:03:57.112
And honestly gets better emails written.
00:03:57.423 --> 00:04:01.703
But then, what's funny on the other side is people use them to read emails and summarize things.
00:04:02.052 --> 00:04:03.612
So, the house wins either way.
00:04:03.703 --> 00:04:08.112
And, hopefully it's a layer that helps everyone be a bit more productive.
00:04:08.807 --> 00:04:09.307
That's great.
00:04:09.967 --> 00:04:10.358
That's true.
00:04:10.358 --> 00:04:10.647
It is.
00:04:10.677 --> 00:04:10.967
It is.
00:04:10.987 --> 00:04:20.658
It could be that skewed data point because also it's not like those over half of people who use it every day are going to the actual portal to the application itself and typing in prompts.
00:04:20.658 --> 00:04:24.497
It's probably just a function that's based into their current work suite as it is.
00:04:24.538 --> 00:04:26.458
And it's Oh, I might as well give this a try.
00:04:26.668 --> 00:04:30.108
The little assistant button pops up that, Hey, can we make this a little tighter for you?
00:04:30.247 --> 00:04:31.418
And you say, yeah, why not?
00:04:31.427 --> 00:04:38.158
So maybe that does factor in as the fact is that is the, Instance that it's being baked into so many functional productivity tools anyway.
00:04:38.358 --> 00:04:38.557
Well,
00:04:38.557 --> 00:05:02.577
we say that, but in the preparation for this call, you and I tried to do an example in Microsoft copilot and PowerPoint, which is, I think, one of the best demos that I've seen in general and what I think about 10 minutes and we gave up, even though I saw some rapper that looks like, GPT in a more of a Microsoft like environment, I'm like, Hey, just give me my copilot I didn't manage yet.
00:05:02.848 --> 00:05:04.247
I'll try to do it for the next episode.
00:05:05.408 --> 00:05:06.168
It tells you something.
00:05:06.168 --> 00:05:12.137
I think there's a discovery issue that frankly should be worked on a bit more because people should be using it.
00:05:12.137 --> 00:05:13.648
It should be made available easily.
00:05:13.648 --> 00:05:21.067
I think that a lot of those users I've spoken to like CIOs of very large companies who saw the power of this technology.
00:05:21.067 --> 00:05:22.548
They themselves are power users.
00:05:22.548 --> 00:05:25.958
And then they got a subscription so that every single employee has it.
00:05:25.958 --> 00:05:27.447
And maybe they've tried and maybe not.
00:05:27.987 --> 00:05:32.757
But for the most part, they don't use it because again, that's the discovery issue of what can you do with it?
00:05:32.958 --> 00:05:35.507
But then, the second thing is like, what should you be using it for?
00:05:35.767 --> 00:05:56.057
And there, I think like early Google, like you'll get burned easily, maybe more easily if you trust it blindly, which is why you maybe don't want to roll it out as aggressively to everyone until, society takes all these nuances into account and just trains up the average member of the population that, it's a pretty.
00:05:57.343 --> 00:05:58.442
Sharp tool, right?
00:05:58.442 --> 00:06:00.593
You can, you can cut yourself, but it'd be great.
00:06:01.173 --> 00:06:03.423
So if you're just using it to generate information and.
00:06:03.862 --> 00:06:07.002
One thing is off and then you build many assumptions on it.
00:06:07.023 --> 00:06:08.392
And you write your whole memo on it.
00:06:09.173 --> 00:06:10.223
Like you're going to look like an idiot.
00:06:10.882 --> 00:06:11.072
Yeah.
00:06:11.072 --> 00:06:14.353
Just, just like your early Google search trust, but verify a little bit.
00:06:14.783 --> 00:06:17.603
So the trust would triangulate in this instance, the same way.
00:06:17.913 --> 00:06:19.302
So where are some of these today?
00:06:19.302 --> 00:06:19.822
No, no,
00:06:19.822 --> 00:06:26.343
even with Google today, it's just, we never, but you still got to have a healthy dose of skepticism.
00:06:27.708 --> 00:06:28.997
In every element of the internet.
00:06:28.997 --> 00:06:32.968
Your social media feeds and yeah, exactly, exactly.
00:06:32.978 --> 00:06:36.588
So you said you talked to a lot of business owners when it comes to how, look, this is a great thing.
00:06:36.588 --> 00:06:37.478
It should be utilized.
00:06:37.507 --> 00:06:39.798
Where do you think it strikes the most immediate?
00:06:40.382 --> 00:06:44.822
Value that low hanging fruit of enterprise, individual utilization in an enterprise.
00:06:44.992 --> 00:06:51.913
A lot of the studies that come back, say sales and marketing, like you were talking about, like creating emails, maybe summarizing emails, that sort of thing, written communication.
00:06:52.262 --> 00:06:56.642
What are some other business functions you think regenerative AI has real stickiness and immediate value?
00:06:57.692 --> 00:06:58.252
Yeah, well, look,
00:06:58.262 --> 00:06:59.612
I think that like creating content.
00:07:00.382 --> 00:07:02.942
Full text, images, audio, everything, right?
00:07:03.372 --> 00:07:04.182
It's getting there.
00:07:04.223 --> 00:07:05.583
And it's getting there really fast, right?
00:07:05.742 --> 00:07:05.983
Yeah.
00:07:06.333 --> 00:07:08.533
So, I think if you're like a marketer, world's your oyster.
00:07:08.543 --> 00:07:12.093
You can now do really impressive things that would have been, like, days of iteration now.
00:07:12.562 --> 00:07:17.187
And you can polish and tweak and do things that are just really, really cool, right?
00:07:17.408 --> 00:07:18.627
There's like the communication piece.
00:07:18.658 --> 00:07:20.038
I think most people use it for that.
00:07:20.197 --> 00:07:26.958
It's very mundane, a number of my colleagues have told me, Hey, I was sick and tired of people telling me my English sucks.
00:07:26.968 --> 00:07:31.687
So now I just run in and go Hey, did you know that that guy's not using a lot of GPT?
00:07:31.798 --> 00:07:33.427
Because his emails got really good.
00:07:33.898 --> 00:07:34.338
Really?
00:07:35.108 --> 00:07:35.408
Oh yeah.
00:07:35.778 --> 00:07:36.538
I won't name any people.
00:07:36.697 --> 00:07:37.067
I can see
00:07:37.067 --> 00:07:47.658
that sort of the next stage of Google Translate is going and actually make me sound like I'm a native English speaker, which includes some nuances and some like colloquial that a standard translate wouldn't do.
00:07:48.187 --> 00:07:49.017
I wouldn't even go that
00:07:49.017 --> 00:07:49.317
far.
00:07:49.327 --> 00:07:54.567
I think it's just if you're, I don't know, a Slav and you don't have articles in your native tongue.
00:07:54.648 --> 00:07:55.730
I'm
00:07:55.730 --> 00:07:56.812
trying
00:07:56.812 --> 00:07:58.978
for the,
00:07:59.057 --> 00:08:00.327
a and an, okay.
00:08:00.458 --> 00:08:00.827
I don't know.
00:08:00.867 --> 00:08:09.418
I still don't know why they exist, frankly, but like chances are you're emitting a few of them or you worked really hard your whole life to get to a point where, and then you're probably overusing them, right?
00:08:09.487 --> 00:08:13.298
I remember getting to a point where, language teachers were like, you don't really need it here.
00:08:13.387 --> 00:08:14.598
I was like, but can I have it?
00:08:14.848 --> 00:08:15.237
Yes.
00:08:15.278 --> 00:08:17.848
Then I'll have it just in case cause I could get it wrong.
00:08:17.898 --> 00:08:31.838
And that's all like giveaways that you're not a native, but I'm not talking about The necessity of being a native, I think that a lot, for a lot of people who, I think many native speakers don't appreciate how much slower a non native is in doing work and not non native language, right?
00:08:31.838 --> 00:08:38.868
So your colleagues writing emails to you who are, where English is a second language, they spend a lot longer writing them.
00:08:38.868 --> 00:08:40.727
They spend a lot of time correcting them.
00:08:41.128 --> 00:08:46.717
You are far more comfortable making a mistake in English than any non native is, because you cannot be judged for it.
00:08:47.138 --> 00:08:48.988
So this is huge! Interesting.
00:08:49.518 --> 00:08:50.857
You know how big this is?
00:08:50.868 --> 00:08:54.408
It's it's much more than like that subtle Can you turn me into more of a native?
00:08:54.457 --> 00:08:55.077
I don't care.
00:08:55.107 --> 00:08:55.947
I'm confident enough.
00:08:56.327 --> 00:09:04.748
But there are many people who, have not learned English since the age of two and hence think that they can speak it as well, even if they sound Serbian, right?
00:09:05.008 --> 00:09:10.337
But I think that for a lot of people, it takes up a large part of their day and now like they love it.
00:09:10.427 --> 00:09:11.597
That's the most positive
00:09:11.597 --> 00:09:12.538
sentiment I've seen.
00:09:13.018 --> 00:09:13.788
That's wild.
00:09:14.018 --> 00:09:18.957
And it so happens to coincide with sort of the globalization of the workforce anyway, post COVID.
00:09:18.957 --> 00:09:23.268
People are working for companies that never had an international presence and that now do so.
00:09:23.268 --> 00:09:25.758
How do you immerse yourself immediately into the corporate culture?
00:09:26.197 --> 00:09:29.488
You leverage these generative tools to fast forward that integration.
00:09:30.288 --> 00:09:31.077
Yeah, that's interesting.
00:09:31.538 --> 00:09:31.748
Yeah.
00:09:31.748 --> 00:09:33.258
And just makes people's lives easier.
00:09:33.258 --> 00:09:34.018
And that's why it's loved.
00:09:34.148 --> 00:09:38.248
That's the one example I can think of where I've just seen people, be hyper excited.
00:09:38.427 --> 00:09:38.677
Okay.
00:09:38.677 --> 00:09:42.903
Like It gave me that, I use it, like I hacked, right?
00:09:42.903 --> 00:09:45.653
It means a lot of joy, right?
00:09:45.653 --> 00:09:46.113
To people.
00:09:46.743 --> 00:09:47.993
That's one I can think of.
00:09:47.993 --> 00:09:50.072
I think a lot of people use what, Oh my God, I've got to read this.
00:09:50.783 --> 00:09:52.052
Paste it, summarize it for me.
00:09:52.383 --> 00:09:53.082
Are there risks?
00:09:53.082 --> 00:09:54.653
Yeah, probably some, right?
00:09:54.692 --> 00:09:55.033
Yeah.
00:09:55.253 --> 00:10:00.982
The other risk of course, is people should know that they're pasting like, confidential data, et cetera.
00:10:01.462 --> 00:10:07.842
I think OpenAI is really good at building models that don't like overfit in ways that they shouldn't, but, you Hey, prompt to the right way.
00:10:07.842 --> 00:10:14.383
Maybe, people learn something about your business, about your personal identifiable information.
00:10:14.932 --> 00:10:27.403
If you've put it in there and, just like a thing to keep in mind as you share, but of course the thing that will be enterprise additions, and there are already all the alarms and things where you can do it in a way where you're not sharing that data.
00:10:27.743 --> 00:10:28.942
So I think that's fine.
00:10:29.442 --> 00:10:31.393
But to me, data, there's all kinds of data.
00:10:31.393 --> 00:10:33.232
It's like how you interpret it.
00:10:33.283 --> 00:10:40.913
A lot of people use it, but I think those use cases around just like communication, improving it are like number one by far.
00:10:41.722 --> 00:10:41.863
So
00:10:42.123 --> 00:10:53.062
when we look at those instances where it's sanctioned and utilize, like you said, the enterprise versions of these LLMs that are scrubbed to make sure that you don't accidentally trip up and, share a PII of customers or, I
00:10:53.982 --> 00:10:55.592
think those additions are there too.
00:10:55.822 --> 00:10:57.793
Tell you that you can do it.
00:10:57.993 --> 00:10:58.562
You're safe.
00:10:58.682 --> 00:11:03.883
This is our instance of Sage or some other LLM.
00:11:04.243 --> 00:11:04.972
So feel free.
00:11:05.003 --> 00:11:07.062
Cause if you're using the general one, sure.
00:11:07.283 --> 00:11:07.452
Yeah.
00:11:07.452 --> 00:11:10.643
They find your data, but like they will offer a version where.
00:11:11.048 --> 00:11:11.768
That's not the case.
00:11:11.817 --> 00:11:13.957
So just I think rental prices, that's okay.
00:11:13.977 --> 00:11:19.138
But for the regular Joe, like a lot of people are like paying for their own subscription, using it this way.
00:11:19.357 --> 00:11:19.868
Exactly.
00:11:20.587 --> 00:11:40.697
Or just like old school search and other old school web tools are using the free version which begs the question, when does a business, at what point does sanctioning sort of the use of the general free version by their employees, at what point do businesses need to step back and say, you maybe we should try and structure this so we can have some governance over how it's leveraged in our business.
00:11:40.908 --> 00:11:42.908
I know this is going to vary from industry to industry.
00:11:42.908 --> 00:11:45.187
You're running a plumbing company or something like that.
00:11:45.368 --> 00:11:53.298
There's not a ton of concern about like safety and standards, but if you're doing ambulatory care or some sort of medical practice, there's a lot that you need to be concerned about.
00:11:53.477 --> 00:11:54.687
So where is that?
00:11:54.738 --> 00:12:02.368
I'll turn a blind eye to you using it because I know it's being used to, we need to have some sort of model over how we, Manage this internally.
00:12:02.538 --> 00:12:03.638
I think it's very simple.
00:12:03.638 --> 00:12:05.268
I think you basically have access control.
00:12:05.268 --> 00:12:07.817
Can you log in and use it on your browser or not?
00:12:07.847 --> 00:12:12.738
This is the oldest, but people, it's what people hate working for big companies, right?
00:12:12.738 --> 00:12:16.258
It's I remember interning in like investment banks and suck Facebook.
00:12:16.298 --> 00:12:17.197
com doesn't work.
00:12:17.418 --> 00:12:18.077
This doesn't work.
00:12:18.077 --> 00:12:20.768
You're like, Oh man, like now I really have to work.
00:12:20.817 --> 00:12:22.687
And it's that is what it is.
00:12:22.717 --> 00:12:25.378
I think that's not that hard in terms of just like limiting access.
00:12:25.827 --> 00:12:26.798
Is it the right thing to do?
00:12:26.807 --> 00:12:28.298
Well, I don't know.
00:12:28.557 --> 00:12:36.238
Thinking about the other applications, obviously we've got, I think the creatives that use it for like creating images, providing a lot of text.
00:12:36.238 --> 00:12:39.368
I think that it gets a wild amount of usage for SEO.
00:12:39.837 --> 00:12:41.418
And it's very good at that, right?
00:12:41.418 --> 00:12:43.908
Just creating content that will get indexed.
00:12:44.268 --> 00:12:50.707
Really what I think many are starting to believe is that your website's going to matter a whole lot less.
00:12:51.398 --> 00:12:54.597
And the information picked up by LLMs, which will.
00:12:55.133 --> 00:13:06.472
If things go the way that I think most LLM providers and cloud providers hope they'll get closer to you as your main interface to technology and to information and to databases.
00:13:06.472 --> 00:13:16.393
If that is the case, then I think companies will have to work pretty hard to make sure that the right kind of data is fed into these LLMs so that they can then serve that.
00:13:16.857 --> 00:13:20.868
As information relevant and pertinent to the user, right?
00:13:20.898 --> 00:13:21.477
The end user.
00:13:21.518 --> 00:13:32.307
And if the consumer, if the average consumer is using LLMs as, maybe a great new channel into the internet, then it becomes really important that the
00:13:32.307 --> 00:13:43.538
That's a, that's a bold statement here's what I'm imagining based on what you just said is that, historically I, I hear of a company, I go to Google I Googled the URL, I go to their website.
00:13:43.548 --> 00:13:44.457
It's, they built it.
00:13:44.677 --> 00:13:47.268
So it's all pretty pro.
00:13:47.947 --> 00:13:48.307
Okay.
00:13:48.307 --> 00:13:48.918
It's managed.
00:13:48.918 --> 00:13:49.347
It's got data.
00:13:50.057 --> 00:13:52.918
Now, next, I don't go to a browser.
00:13:52.918 --> 00:13:54.317
I don't go to a search engine.
00:13:54.327 --> 00:13:57.857
I go to an LLM and I say, tell me about this company.
00:13:57.857 --> 00:14:00.128
And they pretty much, it fabricates.
00:14:00.798 --> 00:14:12.967
A made to order website or repository of information about that company that wraps the good, the bad, the indifferent based on data points throughout the internet, and then puts it in front of me.
00:14:12.967 --> 00:14:20.168
So what you're saying is that polished marketing website of days gone by is going to be less relevant as LLMs take over.
00:14:20.918 --> 00:14:21.307
Yeah.
00:14:21.437 --> 00:14:28.727
Look, I'm just like trying out something and I've asked it in about choosing between like us and one of our competitors.
00:14:29.128 --> 00:14:33.057
And who shall not be named?
00:14:33.057 --> 00:14:36.258
And let me just ask it for but what's our core key differences?
00:14:37.028 --> 00:14:37.538
There you go.
00:14:37.788 --> 00:14:40.048
Let's do an accuracy check too, to see what it says.
00:14:40.048 --> 00:14:40.778
I think that'd be interesting.
00:14:41.477 --> 00:14:42.548
And we're not at all biased.
00:14:42.587 --> 00:14:44.697
So that's what our viewers or listeners should do right now.
00:14:44.697 --> 00:14:50.947
Just go ahead and ask an LLM to do a comparison and see how unbiased the information is.
00:14:50.947 --> 00:14:52.528
I think I think that makes sense.
00:14:52.927 --> 00:15:01.903
My question there at that point, though, and maybe this is a little bit of the infrastructural nuances that I just don't understand, but where does the data reside then?
00:15:02.082 --> 00:15:08.712
How, if it doesn't have sites to creep and crawl and scrape, where is it getting its information from?
00:15:09.673 --> 00:15:10.043
So, okay.
00:15:10.072 --> 00:15:15.913
For example, I've asked it, I saw things and it says choose poly versatile and complex interactions.
00:15:16.133 --> 00:15:19.472
Choose another company for efficient customer service automation.
00:15:19.472 --> 00:15:22.143
So I asked them, so poly is better for complex tasks.
00:15:22.807 --> 00:15:23.118
Yes.
00:15:23.158 --> 00:15:25.607
Polya is generally better suited for handling complex tasks.
00:15:25.778 --> 00:15:26.187
All right.
00:15:26.298 --> 00:15:26.977
There's the commercial
00:15:26.977 --> 00:15:27.408
portion.
00:15:27.418 --> 00:15:29.707
That's the sponsored portion of the podcast, right?
00:15:29.707 --> 00:15:32.118
It's not really right.
00:15:32.258 --> 00:15:35.018
Cause I could then ask, is the other company is it better at anything?
00:15:35.018 --> 00:15:38.947
And like it says, like the other one is effective at simple calls.
00:15:38.957 --> 00:15:42.038
So I think it's being pretty correct politically correct to both companies.
00:15:42.337 --> 00:15:42.567
Sure.
00:15:42.857 --> 00:15:45.067
And it's just a function of websites.
00:15:45.107 --> 00:15:46.008
I know our website.
00:15:46.008 --> 00:15:47.347
I know this company's website.
00:15:47.347 --> 00:15:47.863
It's just.
00:15:48.462 --> 00:15:51.653
It's a function of what we've put there, what they've put there.
00:15:51.863 --> 00:15:53.533
The interesting thing here, so the websites
00:15:53.543 --> 00:15:58.533
will still be important, but it'll be important as an information source and not so much as a source of organic traffic.
00:15:59.033 --> 00:16:00.403
Here's where it's really funny, right?
00:16:00.812 --> 00:16:07.302
Cause people attribute superhuman and there are forms of superhuman intelligence when it comes to alarms.
00:16:07.312 --> 00:16:09.842
And they'll continue to be even more superhuman.
00:16:10.302 --> 00:16:11.923
People get to trust what they're saying.
00:16:11.932 --> 00:16:12.653
Cause they'll go ahead.
00:16:12.773 --> 00:16:14.113
It's read everything, right?
00:16:15.278 --> 00:16:20.118
But I think what I'm seeing here is a lot of information from our website, from the other companies website.
00:16:20.567 --> 00:16:21.368
This is not.
00:16:22.798 --> 00:16:26.177
Someone who's done deep work, spoken to clients, et cetera.
00:16:26.498 --> 00:16:28.607
It's just he said, she said, okay.
00:16:29.207 --> 00:16:30.618
That gets stuff right.
00:16:30.648 --> 00:16:32.408
Cause it's is this company a leader in that?
00:16:32.437 --> 00:16:35.398
Well, all the websites go Hey, we're a leader in X, right?
00:16:35.957 --> 00:16:37.597
Can you verify, what do you think?
00:16:37.638 --> 00:16:38.087
Whatever.
00:16:38.227 --> 00:16:39.177
And it gets difficult.
00:16:39.187 --> 00:16:43.687
So I think it'll be hard, especially if people look to just do something quickly.
00:16:44.812 --> 00:16:48.033
It's not on an exam, but maybe it's like a real workflow like this one.
00:16:48.423 --> 00:16:52.702
I think the interesting thing then becomes like, Hey, like, how did you arrive at the answer?
00:16:52.873 --> 00:16:54.153
Like, why do you think this is true?
00:16:54.363 --> 00:17:00.623
So if you trust it too easily, I think you're at risk of getting like wrong answers and I think it's just going to be interesting.
00:17:00.623 --> 00:17:02.253
Cause you know, we mentioned SEO.
00:17:02.302 --> 00:17:03.712
This is not a form of SEO, right?
00:17:03.712 --> 00:17:10.893
You need to put it really, honestly, like you go and regenerate your website and a thousand links, make sure they're all indexed.
00:17:11.573 --> 00:17:23.923
There's going to be an abundance of you repeating that you're better at someone else than someone else That then they are compared to you and you might just swing the vote And llms might start recommending one vendor over the other, right?
00:17:24.083 --> 00:17:46.153
but I would have to I would have to assume that llms are a little bit more savvy than the old school seo hacks of doing like An invisible font at the bottom of the page with a bunch of key search terms that just drove inorganic traffic because you were hacking, a search algorithm in this case, because it's so natural language focused, like it's really looking for that content in the natural language context, I would assume.
00:17:46.460 --> 00:17:50.740
it just has text, like how does that text is collected and used?
00:17:51.650 --> 00:17:52.940
Is really
00:17:53.000 --> 00:17:54.180
the wizards behind the
00:17:55.809 --> 00:17:57.170
cool collected data.
00:17:57.170 --> 00:18:07.480
How, and the fact that it's not at such great volume means that you open it up for just subterfuge where smart people figure out and reversing how it's being collected.
00:18:07.480 --> 00:18:11.200
And then they'll make sure that what they want in there will get in there.
00:18:11.809 --> 00:18:12.089
Sure.
00:18:12.420 --> 00:18:17.089
And then we're hopefully, at a point where you know, with increased reasoning capability.
00:18:17.515 --> 00:18:29.375
Maybe the system can spot contradictions or maybe, open AI can simulate this task many times, provide feedback right around what was a good and what was a bad answer.
00:18:29.684 --> 00:18:40.055
And I suspect that if this becomes a major issue, they'll limit it so that you get the dreaded as a, conversational agent.
00:18:40.095 --> 00:18:42.835
I cannot give you any advice and blah, blah, blah, blah, blah.
00:18:42.835 --> 00:18:43.474
Sure, yeah.
00:18:44.144 --> 00:18:45.625
But then people will find a way to break that.
00:18:45.634 --> 00:18:46.984
So it's a brand new world, right?
00:18:47.414 --> 00:18:51.565
And at the end of the day, hopefully, you'll have like third party information and stuff.
00:18:51.654 --> 00:18:55.884
And people will also know that maybe they shouldn't be asking for certificate, right?
00:18:56.164 --> 00:19:03.980
The same way that, but if they Google it online, they'll at least get things like G2, Kipterra, Gartner, like those customer reviews.
00:19:04.319 --> 00:19:04.509
Yeah.
00:19:04.900 --> 00:19:06.099
Look at websites.
00:19:06.109 --> 00:19:10.529
They'll make like independent value judgments on this company looks good.
00:19:10.579 --> 00:19:11.210
This company looks good.
00:19:11.210 --> 00:19:12.619
Oh, look at these logos.
00:19:12.660 --> 00:19:13.009
Wow.
00:19:13.059 --> 00:19:16.150
PolyEye works with, I don't know, FedEx and Marriott.
00:19:16.200 --> 00:19:19.779
If I'm UPS, I should definitely work with them and not any other company.
00:19:19.779 --> 00:19:21.329
That can make that about value judgment.
00:19:21.410 --> 00:19:24.460
I've said that I work for a company, which one should I talk to?
00:19:24.559 --> 00:19:29.519
And the answer I'm getting, the answer I'm getting, it's taking a bit of time.
00:19:29.619 --> 00:19:33.970
So hopefully some online work is being done, well, yeah.
00:19:33.970 --> 00:19:38.019
And I think it's more searches online basically contribute to the rag.
00:19:38.069 --> 00:19:39.109
This answer, right?
00:19:39.549 --> 00:19:39.769
Yeah.
00:19:40.500 --> 00:19:44.259
And now I'm getting a pretty bland answer around strengths and weaknesses.
00:19:44.759 --> 00:19:46.269
So it won't be definitive.
00:19:46.890 --> 00:19:47.619
Not for now.
00:19:47.630 --> 00:19:52.680
It's written an essay, which says nothing, which it's probably right.
00:19:53.009 --> 00:19:55.720
Leaving the value judgment to the client.
00:19:56.190 --> 00:19:58.099
Let's bring it back into the enterprise.
00:19:58.099 --> 00:20:08.430
So you said one way that people managers where they know this is being utilized within their walls, one way they can establish whether or not they should have some element of governance is can they log in when in your network?
00:20:08.480 --> 00:20:09.059
Okay, great.
00:20:09.109 --> 00:20:10.970
That's going to be pretty much a super student.
00:20:10.980 --> 00:20:16.009
So every organization should probably have some level of visibility into this.
00:20:16.289 --> 00:20:23.670
What are some signs that people managers and I know this is a general question, so feel free to give a general answer, but where's some signs that people managers.
00:20:23.990 --> 00:20:24.990
Can look out for that.
00:20:25.000 --> 00:20:29.900
Maybe their teams are overutilizing generative AI.
00:20:30.069 --> 00:20:32.009
And it's actually starting to be counterproductive.
00:20:33.329 --> 00:20:33.869
I don't know.
00:20:33.880 --> 00:20:34.960
That's really tough, right?
00:20:34.960 --> 00:20:38.759
I think it depends on what the nature of work What do you have in mind which functions?
00:20:40.039 --> 00:20:45.589
Well, I you know, one of the low hanging fruits is content creation You could go there and you could say I want to write a blog post.
00:20:45.599 --> 00:20:47.549
So gbt write it for me.
00:20:47.599 --> 00:20:49.375
Which You know I think it's a good start.
00:20:49.394 --> 00:20:51.785
I see that as being foundational copy, right?
00:20:51.785 --> 00:20:56.674
Then you go in and you make your tweaks, you make it a little more on brand with regards to voice, et cetera, and then you push play.
00:20:57.255 --> 00:21:01.484
But what happens when that's every time, what happens when you don't do that second layer of editing,
00:21:03.275 --> 00:21:11.734
if it sounds good, then do you need that person in a full time job well, you need them for the value judgment of it, of if it's good or not, but maybe they can do something else in addition to it.
00:21:11.785 --> 00:21:22.000
But yeah us getting more productive means that inevitably some functions have the bigger, Sometimes it's in the smaller really right now, or the end of more productive, right?
00:21:22.049 --> 00:21:38.599
Like maybe instead of doing, writing those five blog posts for the sake of SEO, you pump out 500 and you're covered when it comes to SEO, but maybe you need to write it in a different way to make sure that it picks it up so that when it does value judgments, it's it's ruling in your favor.
00:21:38.900 --> 00:21:46.410
So yeah, so the sign that maybe you're over utilizing it is that you're bumping yourself out of your position, I think is a good way to go.
00:21:46.480 --> 00:21:50.970
So maybe it's maybe it should be a self regulation tool.
00:21:50.970 --> 00:21:55.359
Don't do it so much where you're rendering yourself ineffectual in your current position.
00:21:56.109 --> 00:21:56.890
It makes some sense.
00:21:56.960 --> 00:22:19.039
So let's say it gets to that point in a function, how do you how do you, as a people manager, reverse course, how do you go about setting some guidelines that can be adopted to negate any sort of over utilization or, if you start to see the writing on the wall, preempt it, getting to that point, what are some good practices that people, leaders can deploy to, utilization
00:22:19.039 --> 00:22:19.309
in check?
00:22:19.789 --> 00:22:21.086
There's no such thing as too much.
00:22:21.115 --> 00:22:23.476
I think the question is are they using it inappropriately?
00:22:23.756 --> 00:22:35.875
Are you seeing if you're seeing performance go down because people are, the quality of work changed and it has to do with this, you must train them and educate them on how to use this thing to just improve their productivity rather than.
00:22:36.256 --> 00:22:37.846
Inject noise into their work, right?
00:22:38.155 --> 00:22:39.746
There's no such thing as too much.
00:22:39.776 --> 00:22:40.605
It's a productivity tool.
00:22:40.605 --> 00:22:44.336
It's like saying, Hey, like Damien, you're using that keyboard too much.
00:22:44.365 --> 00:22:48.026
Like I want you to start using a pen because I like your handwriting.
00:22:48.026 --> 00:22:51.395
That's a good way to put it.
00:22:52.476 --> 00:22:53.246
I prefer it.
00:22:53.246 --> 00:22:54.655
If you continue using the keyboard.
00:22:54.806 --> 00:22:55.816
If you can type really fast.
00:22:55.865 --> 00:22:58.945
So, I don't think there's such, I think it's just a matter of don't
00:22:59.066 --> 00:22:59.915
type garbage.
00:22:59.925 --> 00:23:00.586
That's the thing.
00:23:00.596 --> 00:23:00.685
And
00:23:01.705 --> 00:23:08.605
speaking to other people, the people who, what I was going to say is the people who are excited about the communication piece, that's really low hanging fruit.
00:23:08.625 --> 00:23:14.165
I know that a lot of those people were previously really excited about using Grammarly as a tool, right?
00:23:14.365 --> 00:23:14.965
Yeah, sure.
00:23:15.226 --> 00:23:21.586
Now they have an easier thing and a better thing and, it does more of the heavy lifting for them rather than just fixing sentences.
00:23:21.586 --> 00:23:23.566
It's writes an email based on instructions.
00:23:23.996 --> 00:23:28.566
They're really happy, but, for every one of those I found a number of people who are like, Yeah, I have it, I don't use it.
00:23:29.256 --> 00:23:29.925
Or they don't really know.
00:23:29.925 --> 00:23:32.506
And I show them things like, wow, and then a month later, have you used it?
00:23:32.596 --> 00:23:33.786
Nah, not really, right?
00:23:34.066 --> 00:23:40.526
So it has to be in the workflow at the fingertips in a way that really makes sense, because we're over indexing for those hackers.
00:23:40.576 --> 00:23:41.286
Because what was the stat?
00:23:41.286 --> 00:23:44.836
Like 180 million active users by some metric, right?
00:23:45.260 --> 00:23:50.221
And if we say that the total addressable workforce is, well, that number is actually really high, right?
00:23:50.560 --> 00:23:52.361
Like already, like surprisingly high.
00:23:52.401 --> 00:23:54.260
So they're doing well, right?
00:23:54.611 --> 00:23:57.921
But I would really like to double click on the active, right?
00:23:58.250 --> 00:24:03.931
Drawing like really funny images is That I do, on a weekend where I'm like bored.
00:24:04.060 --> 00:24:06.351
That doesn't mean that I'm using it productively.
00:24:06.411 --> 00:24:09.010
In fact, I don't think I'm using it to boost my productivity.
00:24:09.760 --> 00:24:13.961
Is it sapping productivity because you're just using it to make kooky images?
00:24:15.340 --> 00:24:19.560
Well, that's a creepy, but yeah, it's I said kooky.
00:24:19.570 --> 00:24:19.931
Thank you.
00:24:19.931 --> 00:24:20.520
Not creepy.
00:24:21.201 --> 00:24:21.550
Okay.
00:24:21.611 --> 00:24:21.881
Okay.
00:24:22.550 --> 00:24:28.286
I just tried to image in to make two pasta shaped, Or no, two kaijus made of pasta.
00:24:28.496 --> 00:24:35.316
Each had a different, each was made of a different kind of pasta fighting in the streets of Boston and the results were made at best.
00:24:35.326 --> 00:24:36.465
So nice try Google.
00:24:37.016 --> 00:24:37.316
Yeah.
00:24:37.316 --> 00:24:37.556
Yeah.
00:24:37.556 --> 00:24:43.465
I've had like business meetings with pork scratch, scratch things and noodles raining from the ceiling and stuff like that.
00:24:43.465 --> 00:24:47.796
It's great stuff, but Hey, look, I think that it's not a separate at all.
00:24:47.796 --> 00:24:49.536
I think it's you've got to learn how to use it.
00:24:49.536 --> 00:24:51.205
I think we'll find better ways to do it.
00:24:51.576 --> 00:24:53.705
And we just need to integrate into more people's workflows.
00:24:53.705 --> 00:24:56.326
Cause not everyone's like a hacker who wants to do it.
00:24:56.355 --> 00:24:57.996
Most people are comfortable doing what they're doing.
00:24:58.330 --> 00:24:59.560
And don't like change.
00:24:59.911 --> 00:25:03.540
So I think like it, it just has, it can't be clippy, right?
00:25:03.810 --> 00:25:05.250
It has to be better than that.
00:25:05.480 --> 00:25:07.155
And I loved Clippy when I was like, is there
00:25:07.276 --> 00:25:08.060
anything better than clippy?
00:25:08.060 --> 00:25:08.570
That was s nothing.
00:25:08.931 --> 00:25:09.351
Nothing's
00:25:09.351 --> 00:25:10.070
better than clippy.
00:25:10.070 --> 00:25:17.050
I clip Like I remember going to my mom's work with a computer at the time and just seeing the thing, I was like, well, this is like the coolest thing ever.
00:25:17.050 --> 00:25:17.351
right?
00:25:17.355 --> 00:25:21.786
But I think like it just has to flow like email writing, like autocomplete, stuff like that.
00:25:22.125 --> 00:25:23.925
Gmail has some nice features.
00:25:24.195 --> 00:25:25.935
Alright, so let's go beyond that then.
00:25:25.935 --> 00:25:30.445
What are some other functions that aren't currently utilizing generative AI that probably could?
00:25:30.875 --> 00:25:39.046
And maybe if not now given current iterations and capabilities in the near future, do you see what sort of business functions where generative AI will really attack?
00:25:40.016 --> 00:25:48.895
Well, look, I think that what do we do in terms of like customer service automation and providing voice assistants, chatbots, et cetera, at scale is a place where businesses will use it a lot.
00:25:49.316 --> 00:25:51.945
Does that mean that many people end up using a lot of it?
00:25:53.346 --> 00:25:54.566
Unclear, right?
00:25:54.605 --> 00:25:57.226
It will impact the work of a lot of people, right?
00:25:57.566 --> 00:25:58.175
Yeah, I don't know.
00:25:58.276 --> 00:25:58.846
I'm not sure.
00:25:59.125 --> 00:26:03.195
I think that it's it's really like the office suite that I'm excited about.
00:26:03.286 --> 00:26:08.986
We're like, people who spend a lot of time making presentations, Excel models, just like that.
00:26:08.986 --> 00:26:11.266
And yeah, I think it's a question of can that assistant.
00:26:11.711 --> 00:26:24.171
Give you the right precision to, be like a sidekick, like you leave your computer for 10 minutes and Hey, Damon, can you like put these three columns together, change this calculation here, format this thing you do it like, yeah, that would be phenomenal.
00:26:24.201 --> 00:26:34.330
As someone who can do financial modeling somewhat, but I don't have the shortcut set at my fingertips anymore and all of that, like that's a real productivity booster, right?
00:26:34.711 --> 00:26:49.570
Like similarly for presentations, hey, please go through this board deck of 150 slides and make sure that every title is exactly the same number of pixels down and write like that would save me and my OCD quite a lot of time, right?
00:26:52.050 --> 00:26:57.351
If you and I can figure out how to get copilot into our PowerPoint, then maybe just maybe.
00:26:57.951 --> 00:26:59.800
We got a day back for a quarter, right?
00:27:00.181 --> 00:27:00.621
Yeah.
00:27:00.671 --> 00:27:03.621
The speculation is it will be very beneficial.
00:27:03.661 --> 00:27:08.101
Now it's all about, like you said, discovery and access, which is an entirely separate episode.
00:27:08.397 --> 00:27:10.008
All right, Nick, I think we're running out of time here.
00:27:10.008 --> 00:27:11.198
I really appreciate your time today.
00:27:11.198 --> 00:27:12.117
This is good stuff.
00:27:12.377 --> 00:27:12.827
Let's go.
00:27:12.877 --> 00:27:13.307
Let's go.
00:27:13.337 --> 00:27:17.238
Now I'm going to go distract myself by creating images for the next two hours.
00:27:17.248 --> 00:27:18.468
So yeah, let's
00:27:18.468 --> 00:27:19.178
go.
00:27:19.258 --> 00:27:20.698
Let's go automate ourselves.
00:27:21.448 --> 00:27:22.028
That's coming soon.
00:27:22.048 --> 00:27:23.758
We will have a regenerative episode.
00:27:23.817 --> 00:27:24.548
I can't wait for it.
00:27:24.778 --> 00:27:25.117
All right, man.
00:27:25.117 --> 00:27:25.448
Take care.
00:27:25.458 --> 00:27:26.067
Have a great day.
00:27:26.377 --> 00:27:27.028
Have a good one.
00:27:27.057 --> 00:27:27.387
Bye.
00:00:00.800 --> 00:00:05.370
for the most part, they don't use it because again, that's the discovery issue of what can you do with it?
00:00:05.570 --> 00:00:08.119
But then, the second thing is like, what should you be using it for?
00:00:08.380 --> 00:00:27.179
And there, I think like early Google, you'll get burned easily, maybe more easily if you trust it blindly, which is why you maybe don't want to roll it out as aggressively to everyone until, society takes all these nuances into account and just trains up the average member of the population that, it's a pretty.
00:00:27.179 --> 00:00:28.050
Sharp tool,
00:00:34.118 --> 00:00:40.027
Hello friends and welcome to another episode of the deep learning with poly AI podcast.
00:00:41.158 --> 00:00:42.048
I'm your host, Damien.
00:00:42.048 --> 00:00:45.417
I'm here with CEO and co founder of poly AI, Nicola Merchick.
00:00:45.417 --> 00:00:46.548
Hey Nicola, how you doing today?
00:00:47.057 --> 00:00:47.927
Hi Damien, not bad.
00:00:47.927 --> 00:00:48.258
How are you?
00:00:48.773 --> 00:00:49.912
I'm doing very well, thank you.
00:00:50.542 --> 00:01:01.173
I'm not actually Damien, I am a generative ai video and avatar that was created because I figure everyone else is using generative AI to do their work.
00:01:01.173 --> 00:01:01.652
Why not?
00:01:01.862 --> 00:01:02.582
I do the same.
00:01:02.823 --> 00:01:09.453
You and I talked about at the end of a recent recording, making a completely generative episode of the podcast that is coming up soon.
00:01:09.453 --> 00:01:10.742
I'm very excited.
00:01:11.117 --> 00:01:16.037
To see that come to fruition, but it got me thinking, Nicola, that, there are recent reports that came out.
00:01:16.037 --> 00:01:17.117
Adobe had one.
00:01:17.117 --> 00:01:24.397
I know that I know that McKinsey had one as well that talked about how consumer adoption of generative AI specifically.
00:01:24.688 --> 00:01:27.028
But just AI tools in general is at an all time high.
00:01:27.078 --> 00:01:27.688
People get it.
00:01:27.798 --> 00:01:28.918
They know what's great about it.
00:01:28.947 --> 00:01:29.787
They're trying it out.
00:01:29.787 --> 00:01:30.638
They're sampling it.
00:01:31.427 --> 00:01:37.998
And according to Adobe, some, more than half the people in America have at least tried generative AI tools once.
00:01:38.028 --> 00:01:42.378
And of those who use it regularly, more than half use it every single day.
00:01:42.768 --> 00:01:46.697
So suffice it to say generative AI has made itself entrenched.
00:01:47.197 --> 00:01:48.058
It is a tool.
00:01:48.097 --> 00:01:51.938
Obviously we go to work many of those days of every single day.
00:01:52.317 --> 00:01:53.427
And so the question is.
00:01:54.578 --> 00:01:56.638
How, what I wanted to talk about today is.
00:01:56.987 --> 00:02:03.257
How are we seeing employees along the hierarchy, along the various organizational tiers?
00:02:03.268 --> 00:02:07.257
How are we seeing employees today utilizing generative AI?
00:02:07.298 --> 00:02:10.638
And what are some signs that maybe it's getting overutilized?
00:02:10.867 --> 00:02:16.427
I know that recently you talked about when people use it or when it first came out and students were using it to cheat.
00:02:16.487 --> 00:02:19.987
They were having like GPT writing SS form and you're like, well, the way we do that.
00:02:20.247 --> 00:02:21.277
I think there still are, no?
00:02:21.902 --> 00:02:31.573
Yeah, probably non savvy professors that haven't caught up, but it's like, all right, so how do we take that to the next level and how do we say, all right, we recognize this as a tool in the roster.
00:02:31.573 --> 00:02:32.772
It's very, very powerful.
00:02:32.943 --> 00:02:34.362
Here's how you use it effectively.
00:02:34.362 --> 00:02:35.902
So I thought we'd chat about that today.
00:02:36.223 --> 00:02:40.513
First of all, where, what are some of the pros and cons as a generative AI expert?
00:02:40.513 --> 00:02:43.953
What are some of the pros and cons of having this technology in the modern workplace?
00:02:44.242 --> 00:02:46.913
What are some great use cases of not so great use cases?
00:02:47.223 --> 00:02:49.242
And maybe we'll just dive into some details from there.
00:02:49.293 --> 00:02:59.372
Look, I think that AI is a great equalizer, but then equally, I think it increases disparity between like people who can use technology well and those that can't, even further, right?
00:02:59.372 --> 00:03:07.372
So I imagine that when Google, became what it is, I think people who knew how to do that were like infinitely more productive than someone sourcing that information elsewhere.
00:03:07.473 --> 00:03:09.772
And yeah, they had to be sure that the information's right.
00:03:09.782 --> 00:03:11.802
They had to have a healthy dose of skepticism.
00:03:12.362 --> 00:03:16.703
But it gave them really helpful pointers really fast, right?
00:03:17.152 --> 00:03:18.973
So if you have those, then you're like, great.
00:03:19.043 --> 00:03:21.612
Okay I think that the answer is this.
00:03:21.893 --> 00:03:24.513
Now I'll go to my trusted sources and I'll just validate it.
00:03:24.622 --> 00:03:26.573
And if it happens to be right.
00:03:26.798 --> 00:03:27.087
Great.
00:03:27.177 --> 00:03:28.168
I'm done like way faster.
00:03:28.268 --> 00:03:38.538
So similarly, I think, the mundane part, and I think it's like it's hard to trust this data because I suspect that a lot of those people that use it every day, use it for a set of limited things.
00:03:38.548 --> 00:03:40.728
I know a bunch of people that use it for emails, right?
00:03:40.728 --> 00:03:43.198
They write bullet points and be like, explore this.
00:03:43.288 --> 00:03:44.647
It's a very nice verbose email.
00:03:45.018 --> 00:03:51.897
And honestly, if they send those bullet points over, that would probably be more efficient communication, but hey, pomp and circumstance, right?
00:03:53.328 --> 00:03:54.657
It still saves them time.
00:03:55.182 --> 00:03:57.112
And honestly gets better emails written.
00:03:57.423 --> 00:04:01.703
But then, what's funny on the other side is people use them to read emails and summarize things.
00:04:02.052 --> 00:04:03.612
So, the house wins either way.
00:04:03.703 --> 00:04:08.112
And, hopefully it's a layer that helps everyone be a bit more productive.
00:04:08.807 --> 00:04:09.307
That's great.
00:04:09.967 --> 00:04:10.358
That's true.
00:04:10.358 --> 00:04:10.647
It is.
00:04:10.677 --> 00:04:10.967
It is.
00:04:10.987 --> 00:04:20.658
It could be that skewed data point because also it's not like those over half of people who use it every day are going to the actual portal to the application itself and typing in prompts.
00:04:20.658 --> 00:04:24.497
It's probably just a function that's based into their current work suite as it is.
00:04:24.538 --> 00:04:26.458
And it's Oh, I might as well give this a try.
00:04:26.668 --> 00:04:30.108
The little assistant button pops up that, Hey, can we make this a little tighter for you?
00:04:30.247 --> 00:04:31.418
And you say, yeah, why not?
00:04:31.427 --> 00:04:38.158
So maybe that does factor in as the fact is that is the, Instance that it's being baked into so many functional productivity tools anyway.
00:04:38.358 --> 00:04:38.557
Well,
00:04:38.557 --> 00:05:02.577
we say that, but in the preparation for this call, you and I tried to do an example in Microsoft copilot and PowerPoint, which is, I think, one of the best demos that I've seen in general and what I think about 10 minutes and we gave up, even though I saw some rapper that looks like, GPT in a more of a Microsoft like environment, I'm like, Hey, just give me my copilot I didn't manage yet.
00:05:02.848 --> 00:05:04.247
I'll try to do it for the next episode.
00:05:05.408 --> 00:05:06.168
It tells you something.
00:05:06.168 --> 00:05:12.137
I think there's a discovery issue that frankly should be worked on a bit more because people should be using it.
00:05:12.137 --> 00:05:13.648
It should be made available easily.
00:05:13.648 --> 00:05:21.067
I think that a lot of those users I've spoken to like CIOs of very large companies who saw the power of this technology.
00:05:21.067 --> 00:05:22.548
They themselves are power users.
00:05:22.548 --> 00:05:25.958
And then they got a subscription so that every single employee has it.
00:05:25.958 --> 00:05:27.447
And maybe they've tried and maybe not.
00:05:27.987 --> 00:05:32.757
But for the most part, they don't use it because again, that's the discovery issue of what can you do with it?
00:05:32.958 --> 00:05:35.507
But then, the second thing is like, what should you be using it for?
00:05:35.767 --> 00:05:56.057
And there, I think like early Google, like you'll get burned easily, maybe more easily if you trust it blindly, which is why you maybe don't want to roll it out as aggressively to everyone until, society takes all these nuances into account and just trains up the average member of the population that, it's a pretty.
00:05:57.343 --> 00:05:58.442
Sharp tool, right?
00:05:58.442 --> 00:06:00.593
You can, you can cut yourself, but it'd be great.
00:06:01.173 --> 00:06:03.423
So if you're just using it to generate information and.
00:06:03.862 --> 00:06:07.002
One thing is off and then you build many assumptions on it.
00:06:07.023 --> 00:06:08.392
And you write your whole memo on it.
00:06:09.173 --> 00:06:10.223
Like you're going to look like an idiot.
00:06:10.882 --> 00:06:11.072
Yeah.
00:06:11.072 --> 00:06:14.353
Just, just like your early Google search trust, but verify a little bit.
00:06:14.783 --> 00:06:17.603
So the trust would triangulate in this instance, the same way.
00:06:17.913 --> 00:06:19.302
So where are some of these today?
00:06:19.302 --> 00:06:19.822
No, no,
00:06:19.822 --> 00:06:26.343
even with Google today, it's just, we never, but you still got to have a healthy dose of skepticism.
00:06:27.708 --> 00:06:28.997
In every element of the internet.
00:06:28.997 --> 00:06:32.968
Your social media feeds and yeah, exactly, exactly.
00:06:32.978 --> 00:06:36.588
So you said you talked to a lot of business owners when it comes to how, look, this is a great thing.
00:06:36.588 --> 00:06:37.478
It should be utilized.
00:06:37.507 --> 00:06:39.798
Where do you think it strikes the most immediate?
00:06:40.382 --> 00:06:44.822
Value that low hanging fruit of enterprise, individual utilization in an enterprise.
00:06:44.992 --> 00:06:51.913
A lot of the studies that come back, say sales and marketing, like you were talking about, like creating emails, maybe summarizing emails, that sort of thing, written communication.
00:06:52.262 --> 00:06:56.642
What are some other business functions you think regenerative AI has real stickiness and immediate value?
00:06:57.692 --> 00:06:58.252
Yeah, well, look,
00:06:58.262 --> 00:06:59.612
I think that like creating content.
00:07:00.382 --> 00:07:02.942
Full text, images, audio, everything, right?
00:07:03.372 --> 00:07:04.182
It's getting there.
00:07:04.223 --> 00:07:05.583
And it's getting there really fast, right?
00:07:05.742 --> 00:07:05.983
Yeah.
00:07:06.333 --> 00:07:08.533
So, I think if you're like a marketer, world's your oyster.
00:07:08.543 --> 00:07:12.093
You can now do really impressive things that would have been, like, days of iteration now.
00:07:12.562 --> 00:07:17.187
And you can polish and tweak and do things that are just really, really cool, right?
00:07:17.408 --> 00:07:18.627
There's like the communication piece.
00:07:18.658 --> 00:07:20.038
I think most people use it for that.
00:07:20.197 --> 00:07:26.958
It's very mundane, a number of my colleagues have told me, Hey, I was sick and tired of people telling me my English sucks.
00:07:26.968 --> 00:07:31.687
So now I just run in and go Hey, did you know that that guy's not using a lot of GPT?
00:07:31.798 --> 00:07:33.427
Because his emails got really good.
00:07:33.898 --> 00:07:34.338
Really?
00:07:35.108 --> 00:07:35.408
Oh yeah.
00:07:35.778 --> 00:07:36.538
I won't name any people.
00:07:36.697 --> 00:07:37.067
I can see
00:07:37.067 --> 00:07:47.658
that sort of the next stage of Google Translate is going and actually make me sound like I'm a native English speaker, which includes some nuances and some like colloquial that a standard translate wouldn't do.
00:07:48.187 --> 00:07:49.017
I wouldn't even go that
00:07:49.017 --> 00:07:49.317
far.
00:07:49.327 --> 00:07:54.567
I think it's just if you're, I don't know, a Slav and you don't have articles in your native tongue.
00:07:54.648 --> 00:07:55.730
I'm
00:07:55.730 --> 00:07:56.812
trying
00:07:56.812 --> 00:07:58.978
for the,
00:07:59.057 --> 00:08:00.327
a and an, okay.
00:08:00.458 --> 00:08:00.827
I don't know.
00:08:00.867 --> 00:08:09.418
I still don't know why they exist, frankly, but like chances are you're emitting a few of them or you worked really hard your whole life to get to a point where, and then you're probably overusing them, right?
00:08:09.487 --> 00:08:13.298
I remember getting to a point where, language teachers were like, you don't really need it here.
00:08:13.387 --> 00:08:14.598
I was like, but can I have it?
00:08:14.848 --> 00:08:15.237
Yes.
00:08:15.278 --> 00:08:17.848
Then I'll have it just in case cause I could get it wrong.
00:08:17.898 --> 00:08:31.838
And that's all like giveaways that you're not a native, but I'm not talking about The necessity of being a native, I think that a lot, for a lot of people who, I think many native speakers don't appreciate how much slower a non native is in doing work and not non native language, right?
00:08:31.838 --> 00:08:38.868
So your colleagues writing emails to you who are, where English is a second language, they spend a lot longer writing them.
00:08:38.868 --> 00:08:40.727
They spend a lot of time correcting them.
00:08:41.128 --> 00:08:46.717
You are far more comfortable making a mistake in English than any non native is, because you cannot be judged for it.
00:08:47.138 --> 00:08:48.988
So this is huge! Interesting.
00:08:49.518 --> 00:08:50.857
You know how big this is?
00:08:50.868 --> 00:08:54.408
It's it's much more than like that subtle Can you turn me into more of a native?
00:08:54.457 --> 00:08:55.077
I don't care.
00:08:55.107 --> 00:08:55.947
I'm confident enough.
00:08:56.327 --> 00:09:04.748
But there are many people who, have not learned English since the age of two and hence think that they can speak it as well, even if they sound Serbian, right?
00:09:05.008 --> 00:09:10.337
But I think that for a lot of people, it takes up a large part of their day and now like they love it.
00:09:10.427 --> 00:09:11.597
That's the most positive
00:09:11.597 --> 00:09:12.538
sentiment I've seen.
00:09:13.018 --> 00:09:13.788
That's wild.
00:09:14.018 --> 00:09:18.957
And it so happens to coincide with sort of the globalization of the workforce anyway, post COVID.
00:09:18.957 --> 00:09:23.268
People are working for companies that never had an international presence and that now do so.
00:09:23.268 --> 00:09:25.758
How do you immerse yourself immediately into the corporate culture?
00:09:26.197 --> 00:09:29.488
You leverage these generative tools to fast forward that integration.
00:09:30.288 --> 00:09:31.077
Yeah, that's interesting.
00:09:31.538 --> 00:09:31.748
Yeah.
00:09:31.748 --> 00:09:33.258
And just makes people's lives easier.
00:09:33.258 --> 00:09:34.018
And that's why it's loved.
00:09:34.148 --> 00:09:38.248
That's the one example I can think of where I've just seen people, be hyper excited.
00:09:38.427 --> 00:09:38.677
Okay.
00:09:38.677 --> 00:09:42.903
Like It gave me that, I use it, like I hacked, right?
00:09:42.903 --> 00:09:45.653
It means a lot of joy, right?
00:09:45.653 --> 00:09:46.113
To people.
00:09:46.743 --> 00:09:47.993
That's one I can think of.
00:09:47.993 --> 00:09:50.072
I think a lot of people use what, Oh my God, I've got to read this.
00:09:50.783 --> 00:09:52.052
Paste it, summarize it for me.
00:09:52.383 --> 00:09:53.082
Are there risks?
00:09:53.082 --> 00:09:54.653
Yeah, probably some, right?
00:09:54.692 --> 00:09:55.033
Yeah.
00:09:55.253 --> 00:10:00.982
The other risk of course, is people should know that they're pasting like, confidential data, et cetera.
00:10:01.462 --> 00:10:07.842
I think OpenAI is really good at building models that don't like overfit in ways that they shouldn't, but, you Hey, prompt to the right way.
00:10:07.842 --> 00:10:14.383
Maybe, people learn something about your business, about your personal identifiable information.
00:10:14.932 --> 00:10:27.403
If you've put it in there and, just like a thing to keep in mind as you share, but of course the thing that will be enterprise additions, and there are already all the alarms and things where you can do it in a way where you're not sharing that data.
00:10:27.743 --> 00:10:28.942
So I think that's fine.
00:10:29.442 --> 00:10:31.393
But to me, data, there's all kinds of data.
00:10:31.393 --> 00:10:33.232
It's like how you interpret it.
00:10:33.283 --> 00:10:40.913
A lot of people use it, but I think those use cases around just like communication, improving it are like number one by far.
00:10:41.722 --> 00:10:41.863
So
00:10:42.123 --> 00:10:53.062
when we look at those instances where it's sanctioned and utilize, like you said, the enterprise versions of these LLMs that are scrubbed to make sure that you don't accidentally trip up and, share a PII of customers or, I
00:10:53.982 --> 00:10:55.592
think those additions are there too.
00:10:55.822 --> 00:10:57.793
Tell you that you can do it.
00:10:57.993 --> 00:10:58.562
You're safe.
00:10:58.682 --> 00:11:03.883
This is our instance of Sage or some other LLM.
00:11:04.243 --> 00:11:04.972
So feel free.
00:11:05.003 --> 00:11:07.062
Cause if you're using the general one, sure.
00:11:07.283 --> 00:11:07.452
Yeah.
00:11:07.452 --> 00:11:10.643
They find your data, but like they will offer a version where.
00:11:11.048 --> 00:11:11.768
That's not the case.
00:11:11.817 --> 00:11:13.957
So just I think rental prices, that's okay.
00:11:13.977 --> 00:11:19.138
But for the regular Joe, like a lot of people are like paying for their own subscription, using it this way.
00:11:19.357 --> 00:11:19.868
Exactly.
00:11:20.587 --> 00:11:40.697
Or just like old school search and other old school web tools are using the free version which begs the question, when does a business, at what point does sanctioning sort of the use of the general free version by their employees, at what point do businesses need to step back and say, you maybe we should try and structure this so we can have some governance over how it's leveraged in our business.
00:11:40.908 --> 00:11:42.908
I know this is going to vary from industry to industry.
00:11:42.908 --> 00:11:45.187
You're running a plumbing company or something like that.
00:11:45.368 --> 00:11:53.298
There's not a ton of concern about like safety and standards, but if you're doing ambulatory care or some sort of medical practice, there's a lot that you need to be concerned about.
00:11:53.477 --> 00:11:54.687
So where is that?
00:11:54.738 --> 00:12:02.368
I'll turn a blind eye to you using it because I know it's being used to, we need to have some sort of model over how we, Manage this internally.
00:12:02.538 --> 00:12:03.638
I think it's very simple.
00:12:03.638 --> 00:12:05.268
I think you basically have access control.
00:12:05.268 --> 00:12:07.817
Can you log in and use it on your browser or not?
00:12:07.847 --> 00:12:12.738
This is the oldest, but people, it's what people hate working for big companies, right?
00:12:12.738 --> 00:12:16.258
It's I remember interning in like investment banks and suck Facebook.
00:12:16.298 --> 00:12:17.197
com doesn't work.
00:12:17.418 --> 00:12:18.077
This doesn't work.
00:12:18.077 --> 00:12:20.768
You're like, Oh man, like now I really have to work.
00:12:20.817 --> 00:12:22.687
And it's that is what it is.
00:12:22.717 --> 00:12:25.378
I think that's not that hard in terms of just like limiting access.
00:12:25.827 --> 00:12:26.798
Is it the right thing to do?
00:12:26.807 --> 00:12:28.298
Well, I don't know.
00:12:28.557 --> 00:12:36.238
Thinking about the other applications, obviously we've got, I think the creatives that use it for like creating images, providing a lot of text.
00:12:36.238 --> 00:12:39.368
I think that it gets a wild amount of usage for SEO.
00:12:39.837 --> 00:12:41.418
And it's very good at that, right?
00:12:41.418 --> 00:12:43.908
Just creating content that will get indexed.
00:12:44.268 --> 00:12:50.707
Really what I think many are starting to believe is that your website's going to matter a whole lot less.
00:12:51.398 --> 00:12:54.597
And the information picked up by LLMs, which will.
00:12:55.133 --> 00:13:06.472
If things go the way that I think most LLM providers and cloud providers hope they'll get closer to you as your main interface to technology and to information and to databases.
00:13:06.472 --> 00:13:16.393
If that is the case, then I think companies will have to work pretty hard to make sure that the right kind of data is fed into these LLMs so that they can then serve that.
00:13:16.857 --> 00:13:20.868
As information relevant and pertinent to the user, right?
00:13:20.898 --> 00:13:21.477
The end user.
00:13:21.518 --> 00:13:32.307
And if the consumer, if the average consumer is using LLMs as, maybe a great new channel into the internet, then it becomes really important that the
00:13:32.307 --> 00:13:43.538
That's a, that's a bold statement here's what I'm imagining based on what you just said is that, historically I, I hear of a company, I go to Google I Googled the URL, I go to their website.
00:13:43.548 --> 00:13:44.457
It's, they built it.
00:13:44.677 --> 00:13:47.268
So it's all pretty pro.
00:13:47.947 --> 00:13:48.307
Okay.
00:13:48.307 --> 00:13:48.918
It's managed.
00:13:48.918 --> 00:13:49.347
It's got data.
00:13:50.057 --> 00:13:52.918
Now, next, I don't go to a browser.
00:13:52.918 --> 00:13:54.317
I don't go to a search engine.
00:13:54.327 --> 00:13:57.857
I go to an LLM and I say, tell me about this company.
00:13:57.857 --> 00:14:00.128
And they pretty much, it fabricates.
00:14:00.798 --> 00:14:12.967
A made to order website or repository of information about that company that wraps the good, the bad, the indifferent based on data points throughout the internet, and then puts it in front of me.
00:14:12.967 --> 00:14:20.168
So what you're saying is that polished marketing website of days gone by is going to be less relevant as LLMs take over.
00:14:20.918 --> 00:14:21.307
Yeah.
00:14:21.437 --> 00:14:28.727
Look, I'm just like trying out something and I've asked it in about choosing between like us and one of our competitors.
00:14:29.128 --> 00:14:33.057
And who shall not be named?
00:14:33.057 --> 00:14:36.258
And let me just ask it for but what's our core key differences?
00:14:37.028 --> 00:14:37.538
There you go.
00:14:37.788 --> 00:14:40.048
Let's do an accuracy check too, to see what it says.
00:14:40.048 --> 00:14:40.778
I think that'd be interesting.
00:14:41.477 --> 00:14:42.548
And we're not at all biased.
00:14:42.587 --> 00:14:44.697
So that's what our viewers or listeners should do right now.
00:14:44.697 --> 00:14:50.947
Just go ahead and ask an LLM to do a comparison and see how unbiased the information is.
00:14:50.947 --> 00:14:52.528
I think I think that makes sense.
00:14:52.927 --> 00:15:01.903
My question there at that point, though, and maybe this is a little bit of the infrastructural nuances that I just don't understand, but where does the data reside then?
00:15:02.082 --> 00:15:08.712
How, if it doesn't have sites to creep and crawl and scrape, where is it getting its information from?
00:15:09.673 --> 00:15:10.043
So, okay.
00:15:10.072 --> 00:15:15.913
For example, I've asked it, I saw things and it says choose poly versatile and complex interactions.
00:15:16.133 --> 00:15:19.472
Choose another company for efficient customer service automation.
00:15:19.472 --> 00:15:22.143
So I asked them, so poly is better for complex tasks.
00:15:22.807 --> 00:15:23.118
Yes.
00:15:23.158 --> 00:15:25.607
Polya is generally better suited for handling complex tasks.
00:15:25.778 --> 00:15:26.187
All right.
00:15:26.298 --> 00:15:26.977
There's the commercial
00:15:26.977 --> 00:15:27.408
portion.
00:15:27.418 --> 00:15:29.707
That's the sponsored portion of the podcast, right?
00:15:29.707 --> 00:15:32.118
It's not really right.
00:15:32.258 --> 00:15:35.018
Cause I could then ask, is the other company is it better at anything?
00:15:35.018 --> 00:15:38.947
And like it says, like the other one is effective at simple calls.
00:15:38.957 --> 00:15:42.038
So I think it's being pretty correct politically correct to both companies.
00:15:42.337 --> 00:15:42.567
Sure.
00:15:42.857 --> 00:15:45.067
And it's just a function of websites.
00:15:45.107 --> 00:15:46.008
I know our website.
00:15:46.008 --> 00:15:47.347
I know this company's website.
00:15:47.347 --> 00:15:47.863
It's just.
00:15:48.462 --> 00:15:51.653
It's a function of what we've put there, what they've put there.
00:15:51.863 --> 00:15:53.533
The interesting thing here, so the websites
00:15:53.543 --> 00:15:58.533
will still be important, but it'll be important as an information source and not so much as a source of organic traffic.
00:15:59.033 --> 00:16:00.403
Here's where it's really funny, right?
00:16:00.812 --> 00:16:07.302
Cause people attribute superhuman and there are forms of superhuman intelligence when it comes to alarms.
00:16:07.312 --> 00:16:09.842
And they'll continue to be even more superhuman.
00:16:10.302 --> 00:16:11.923
People get to trust what they're saying.
00:16:11.932 --> 00:16:12.653
Cause they'll go ahead.
00:16:12.773 --> 00:16:14.113
It's read everything, right?
00:16:15.278 --> 00:16:20.118
But I think what I'm seeing here is a lot of information from our website, from the other companies website.
00:16:20.567 --> 00:16:21.368
This is not.
00:16:22.798 --> 00:16:26.177
Someone who's done deep work, spoken to clients, et cetera.
00:16:26.498 --> 00:16:28.607
It's just he said, she said, okay.
00:16:29.207 --> 00:16:30.618
That gets stuff right.
00:16:30.648 --> 00:16:32.408
Cause it's is this company a leader in that?
00:16:32.437 --> 00:16:35.398
Well, all the websites go Hey, we're a leader in X, right?
00:16:35.957 --> 00:16:37.597
Can you verify, what do you think?
00:16:37.638 --> 00:16:38.087
Whatever.
00:16:38.227 --> 00:16:39.177
And it gets difficult.
00:16:39.187 --> 00:16:43.687
So I think it'll be hard, especially if people look to just do something quickly.
00:16:44.812 --> 00:16:48.033
It's not on an exam, but maybe it's like a real workflow like this one.
00:16:48.423 --> 00:16:52.702
I think the interesting thing then becomes like, Hey, like, how did you arrive at the answer?
00:16:52.873 --> 00:16:54.153
Like, why do you think this is true?
00:16:54.363 --> 00:17:00.623
So if you trust it too easily, I think you're at risk of getting like wrong answers and I think it's just going to be interesting.
00:17:00.623 --> 00:17:02.253
Cause you know, we mentioned SEO.
00:17:02.302 --> 00:17:03.712
This is not a form of SEO, right?
00:17:03.712 --> 00:17:10.893
You need to put it really, honestly, like you go and regenerate your website and a thousand links, make sure they're all indexed.
00:17:11.573 --> 00:17:23.923
There's going to be an abundance of you repeating that you're better at someone else than someone else That then they are compared to you and you might just swing the vote And llms might start recommending one vendor over the other, right?
00:17:24.083 --> 00:17:46.153
but I would have to I would have to assume that llms are a little bit more savvy than the old school seo hacks of doing like An invisible font at the bottom of the page with a bunch of key search terms that just drove inorganic traffic because you were hacking, a search algorithm in this case, because it's so natural language focused, like it's really looking for that content in the natural language context, I would assume.
00:17:46.460 --> 00:17:50.740
it just has text, like how does that text is collected and used?
00:17:51.650 --> 00:17:52.940
Is really
00:17:53.000 --> 00:17:54.180
the wizards behind the
00:17:55.809 --> 00:17:57.170
cool collected data.
00:17:57.170 --> 00:18:07.480
How, and the fact that it's not at such great volume means that you open it up for just subterfuge where smart people figure out and reversing how it's being collected.
00:18:07.480 --> 00:18:11.200
And then they'll make sure that what they want in there will get in there.
00:18:11.809 --> 00:18:12.089
Sure.
00:18:12.420 --> 00:18:17.089
And then we're hopefully, at a point where you know, with increased reasoning capability.
00:18:17.515 --> 00:18:29.375
Maybe the system can spot contradictions or maybe, open AI can simulate this task many times, provide feedback right around what was a good and what was a bad answer.
00:18:29.684 --> 00:18:40.055
And I suspect that if this becomes a major issue, they'll limit it so that you get the dreaded as a, conversational agent.
00:18:40.095 --> 00:18:42.835
I cannot give you any advice and blah, blah, blah, blah, blah.
00:18:42.835 --> 00:18:43.474
Sure, yeah.
00:18:44.144 --> 00:18:45.625
But then people will find a way to break that.
00:18:45.634 --> 00:18:46.984
So it's a brand new world, right?
00:18:47.414 --> 00:18:51.565
And at the end of the day, hopefully, you'll have like third party information and stuff.
00:18:51.654 --> 00:18:55.884
And people will also know that maybe they shouldn't be asking for certificate, right?
00:18:56.164 --> 00:19:03.980
The same way that, but if they Google it online, they'll at least get things like G2, Kipterra, Gartner, like those customer reviews.
00:19:04.319 --> 00:19:04.509
Yeah.
00:19:04.900 --> 00:19:06.099
Look at websites.
00:19:06.109 --> 00:19:10.529
They'll make like independent value judgments on this company looks good.
00:19:10.579 --> 00:19:11.210
This company looks good.
00:19:11.210 --> 00:19:12.619
Oh, look at these logos.
00:19:12.660 --> 00:19:13.009
Wow.
00:19:13.059 --> 00:19:16.150
PolyEye works with, I don't know, FedEx and Marriott.
00:19:16.200 --> 00:19:19.779
If I'm UPS, I should definitely work with them and not any other company.
00:19:19.779 --> 00:19:21.329
That can make that about value judgment.
00:19:21.410 --> 00:19:24.460
I've said that I work for a company, which one should I talk to?
00:19:24.559 --> 00:19:29.519
And the answer I'm getting, the answer I'm getting, it's taking a bit of time.
00:19:29.619 --> 00:19:33.970
So hopefully some online work is being done, well, yeah.
00:19:33.970 --> 00:19:38.019
And I think it's more searches online basically contribute to the rag.
00:19:38.069 --> 00:19:39.109
This answer, right?
00:19:39.549 --> 00:19:39.769
Yeah.
00:19:40.500 --> 00:19:44.259
And now I'm getting a pretty bland answer around strengths and weaknesses.
00:19:44.759 --> 00:19:46.269
So it won't be definitive.
00:19:46.890 --> 00:19:47.619
Not for now.
00:19:47.630 --> 00:19:52.680
It's written an essay, which says nothing, which it's probably right.
00:19:53.009 --> 00:19:55.720
Leaving the value judgment to the client.
00:19:56.190 --> 00:19:58.099
Let's bring it back into the enterprise.
00:19:58.099 --> 00:20:08.430
So you said one way that people managers where they know this is being utilized within their walls, one way they can establish whether or not they should have some element of governance is can they log in when in your network?
00:20:08.480 --> 00:20:09.059
Okay, great.
00:20:09.109 --> 00:20:10.970
That's going to be pretty much a super student.
00:20:10.980 --> 00:20:16.009
So every organization should probably have some level of visibility into this.
00:20:16.289 --> 00:20:23.670
What are some signs that people managers and I know this is a general question, so feel free to give a general answer, but where's some signs that people managers.
00:20:23.990 --> 00:20:24.990
Can look out for that.
00:20:25.000 --> 00:20:29.900
Maybe their teams are overutilizing generative AI.
00:20:30.069 --> 00:20:32.009
And it's actually starting to be counterproductive.
00:20:33.329 --> 00:20:33.869
I don't know.
00:20:33.880 --> 00:20:34.960
That's really tough, right?
00:20:34.960 --> 00:20:38.759
I think it depends on what the nature of work What do you have in mind which functions?
00:20:40.039 --> 00:20:45.589
Well, I you know, one of the low hanging fruits is content creation You could go there and you could say I want to write a blog post.
00:20:45.599 --> 00:20:47.549
So gbt write it for me.
00:20:47.599 --> 00:20:49.375
Which You know I think it's a good start.
00:20:49.394 --> 00:20:51.785
I see that as being foundational copy, right?
00:20:51.785 --> 00:20:56.674
Then you go in and you make your tweaks, you make it a little more on brand with regards to voice, et cetera, and then you push play.
00:20:57.255 --> 00:21:01.484
But what happens when that's every time, what happens when you don't do that second layer of editing,
00:21:03.275 --> 00:21:11.734
if it sounds good, then do you need that person in a full time job well, you need them for the value judgment of it, of if it's good or not, but maybe they can do something else in addition to it.
00:21:11.785 --> 00:21:22.000
But yeah us getting more productive means that inevitably some functions have the bigger, Sometimes it's in the smaller really right now, or the end of more productive, right?
00:21:22.049 --> 00:21:38.599
Like maybe instead of doing, writing those five blog posts for the sake of SEO, you pump out 500 and you're covered when it comes to SEO, but maybe you need to write it in a different way to make sure that it picks it up so that when it does value judgments, it's it's ruling in your favor.
00:21:38.900 --> 00:21:46.410
So yeah, so the sign that maybe you're over utilizing it is that you're bumping yourself out of your position, I think is a good way to go.
00:21:46.480 --> 00:21:50.970
So maybe it's maybe it should be a self regulation tool.
00:21:50.970 --> 00:21:55.359
Don't do it so much where you're rendering yourself ineffectual in your current position.
00:21:56.109 --> 00:21:56.890
It makes some sense.
00:21:56.960 --> 00:22:19.039
So let's say it gets to that point in a function, how do you how do you, as a people manager, reverse course, how do you go about setting some guidelines that can be adopted to negate any sort of over utilization or, if you start to see the writing on the wall, preempt it, getting to that point, what are some good practices that people, leaders can deploy to, utilization
00:22:19.039 --> 00:22:19.309
in check?
00:22:19.789 --> 00:22:21.086
There's no such thing as too much.
00:22:21.115 --> 00:22:23.476
I think the question is are they using it inappropriately?
00:22:23.756 --> 00:22:35.875
Are you seeing if you're seeing performance go down because people are, the quality of work changed and it has to do with this, you must train them and educate them on how to use this thing to just improve their productivity rather than.
00:22:36.256 --> 00:22:37.846
Inject noise into their work, right?
00:22:38.155 --> 00:22:39.746
There's no such thing as too much.
00:22:39.776 --> 00:22:40.605
It's a productivity tool.
00:22:40.605 --> 00:22:44.336
It's like saying, Hey, like Damien, you're using that keyboard too much.
00:22:44.365 --> 00:22:48.026
Like I want you to start using a pen because I like your handwriting.
00:22:48.026 --> 00:22:51.395
That's a good way to put it.
00:22:52.476 --> 00:22:53.246
I prefer it.
00:22:53.246 --> 00:22:54.655
If you continue using the keyboard.
00:22:54.806 --> 00:22:55.816
If you can type really fast.
00:22:55.865 --> 00:22:58.945
So, I don't think there's such, I think it's just a matter of don't
00:22:59.066 --> 00:22:59.915
type garbage.
00:22:59.925 --> 00:23:00.586
That's the thing.
00:23:00.596 --> 00:23:00.685
And
00:23:01.705 --> 00:23:08.605
speaking to other people, the people who, what I was going to say is the people who are excited about the communication piece, that's really low hanging fruit.
00:23:08.625 --> 00:23:14.165
I know that a lot of those people were previously really excited about using Grammarly as a tool, right?
00:23:14.365 --> 00:23:14.965
Yeah, sure.
00:23:15.226 --> 00:23:21.586
Now they have an easier thing and a better thing and, it does more of the heavy lifting for them rather than just fixing sentences.
00:23:21.586 --> 00:23:23.566
It's writes an email based on instructions.
00:23:23.996 --> 00:23:28.566
They're really happy, but, for every one of those I found a number of people who are like, Yeah, I have it, I don't use it.
00:23:29.256 --> 00:23:29.925
Or they don't really know.
00:23:29.925 --> 00:23:32.506
And I show them things like, wow, and then a month later, have you used it?
00:23:32.596 --> 00:23:33.786
Nah, not really, right?
00:23:34.066 --> 00:23:40.526
So it has to be in the workflow at the fingertips in a way that really makes sense, because we're over indexing for those hackers.
00:23:40.576 --> 00:23:41.286
Because what was the stat?
00:23:41.286 --> 00:23:44.836
Like 180 million active users by some metric, right?
00:23:45.260 --> 00:23:50.221
And if we say that the total addressable workforce is, well, that number is actually really high, right?
00:23:50.560 --> 00:23:52.361
Like already, like surprisingly high.
00:23:52.401 --> 00:23:54.260
So they're doing well, right?
00:23:54.611 --> 00:23:57.921
But I would really like to double click on the active, right?
00:23:58.250 --> 00:24:03.931
Drawing like really funny images is That I do, on a weekend where I'm like bored.
00:24:04.060 --> 00:24:06.351
That doesn't mean that I'm using it productively.
00:24:06.411 --> 00:24:09.010
In fact, I don't think I'm using it to boost my productivity.
00:24:09.760 --> 00:24:13.961
Is it sapping productivity because you're just using it to make kooky images?
00:24:15.340 --> 00:24:19.560
Well, that's a creepy, but yeah, it's I said kooky.
00:24:19.570 --> 00:24:19.931
Thank you.
00:24:19.931 --> 00:24:20.520
Not creepy.
00:24:21.201 --> 00:24:21.550
Okay.
00:24:21.611 --> 00:24:21.881
Okay.
00:24:22.550 --> 00:24:28.286
I just tried to image in to make two pasta shaped, Or no, two kaijus made of pasta.
00:24:28.496 --> 00:24:35.316
Each had a different, each was made of a different kind of pasta fighting in the streets of Boston and the results were made at best.
00:24:35.326 --> 00:24:36.465
So nice try Google.
00:24:37.016 --> 00:24:37.316
Yeah.
00:24:37.316 --> 00:24:37.556
Yeah.
00:24:37.556 --> 00:24:43.465
I've had like business meetings with pork scratch, scratch things and noodles raining from the ceiling and stuff like that.
00:24:43.465 --> 00:24:47.796
It's great stuff, but Hey, look, I think that it's not a separate at all.
00:24:47.796 --> 00:24:49.536
I think it's you've got to learn how to use it.
00:24:49.536 --> 00:24:51.205
I think we'll find better ways to do it.
00:24:51.576 --> 00:24:53.705
And we just need to integrate into more people's workflows.
00:24:53.705 --> 00:24:56.326
Cause not everyone's like a hacker who wants to do it.
00:24:56.355 --> 00:24:57.996
Most people are comfortable doing what they're doing.
00:24:58.330 --> 00:24:59.560
And don't like change.
00:24:59.911 --> 00:25:03.540
So I think like it, it just has, it can't be clippy, right?
00:25:03.810 --> 00:25:05.250
It has to be better than that.
00:25:05.480 --> 00:25:07.155
And I loved Clippy when I was like, is there
00:25:07.276 --> 00:25:08.060
anything better than clippy?
00:25:08.060 --> 00:25:08.570
That was s nothing.
00:25:08.931 --> 00:25:09.351
Nothing's
00:25:09.351 --> 00:25:10.070
better than clippy.
00:25:10.070 --> 00:25:17.050
I clip Like I remember going to my mom's work with a computer at the time and just seeing the thing, I was like, well, this is like the coolest thing ever.
00:25:17.050 --> 00:25:17.351
right?
00:25:17.355 --> 00:25:21.786
But I think like it just has to flow like email writing, like autocomplete, stuff like that.
00:25:22.125 --> 00:25:23.925
Gmail has some nice features.
00:25:24.195 --> 00:25:25.935
Alright, so let's go beyond that then.
00:25:25.935 --> 00:25:30.445
What are some other functions that aren't currently utilizing generative AI that probably could?
00:25:30.875 --> 00:25:39.046
And maybe if not now given current iterations and capabilities in the near future, do you see what sort of business functions where generative AI will really attack?
00:25:40.016 --> 00:25:48.895
Well, look, I think that what do we do in terms of like customer service automation and providing voice assistants, chatbots, et cetera, at scale is a place where businesses will use it a lot.
00:25:49.316 --> 00:25:51.945
Does that mean that many people end up using a lot of it?
00:25:53.346 --> 00:25:54.566
Unclear, right?
00:25:54.605 --> 00:25:57.226
It will impact the work of a lot of people, right?
00:25:57.566 --> 00:25:58.175
Yeah, I don't know.
00:25:58.276 --> 00:25:58.846
I'm not sure.
00:25:59.125 --> 00:26:03.195
I think that it's it's really like the office suite that I'm excited about.
00:26:03.286 --> 00:26:08.986
We're like, people who spend a lot of time making presentations, Excel models, just like that.
00:26:08.986 --> 00:26:11.266
And yeah, I think it's a question of can that assistant.
00:26:11.711 --> 00:26:24.171
Give you the right precision to, be like a sidekick, like you leave your computer for 10 minutes and Hey, Damon, can you like put these three columns together, change this calculation here, format this thing you do it like, yeah, that would be phenomenal.
00:26:24.201 --> 00:26:34.330
As someone who can do financial modeling somewhat, but I don't have the shortcut set at my fingertips anymore and all of that, like that's a real productivity booster, right?
00:26:34.711 --> 00:26:49.570
Like similarly for presentations, hey, please go through this board deck of 150 slides and make sure that every title is exactly the same number of pixels down and write like that would save me and my OCD quite a lot of time, right?
00:26:52.050 --> 00:26:57.351
If you and I can figure out how to get copilot into our PowerPoint, then maybe just maybe.
00:26:57.951 --> 00:26:59.800
We got a day back for a quarter, right?
00:27:00.181 --> 00:27:00.621
Yeah.
00:27:00.671 --> 00:27:03.621
The speculation is it will be very beneficial.
00:27:03.661 --> 00:27:08.101
Now it's all about, like you said, discovery and access, which is an entirely separate episode.
00:27:08.397 --> 00:27:10.008
All right, Nick, I think we're running out of time here.
00:27:10.008 --> 00:27:11.198
I really appreciate your time today.
00:27:11.198 --> 00:27:12.117
This is good stuff.
00:27:12.377 --> 00:27:12.827
Let's go.
00:27:12.877 --> 00:27:13.307
Let's go.
00:27:13.337 --> 00:27:17.238
Now I'm going to go distract myself by creating images for the next two hours.
00:27:17.248 --> 00:27:18.468
So yeah, let's
00:27:18.468 --> 00:27:19.178
go.
00:27:19.258 --> 00:27:20.698
Let's go automate ourselves.
00:27:21.448 --> 00:27:22.028
That's coming soon.
00:27:22.048 --> 00:27:23.758
We will have a regenerative episode.
00:27:23.817 --> 00:27:24.548
I can't wait for it.
00:27:24.778 --> 00:27:25.117
All right, man.
00:27:25.117 --> 00:27:25.448
Take care.
00:27:25.458 --> 00:27:26.067
Have a great day.
00:27:26.377 --> 00:27:27.028
Have a good one.
00:27:27.057 --> 00:27:27.387
Bye.
