Cognitive surrender: what we lose when AI thinks first | Toju Duke

WEBVTT
00:00:07.839 --> 00:00:10.400
Welcome to Human X Intelligent.
00:00:11.039 --> 00:00:17.199
Welcome to Human X Intelligent, a podcast about what it means to be human in the age of AI.
00:00:17.359 --> 00:00:28.079
From all of us living through it, I am Medalina Kosta, and every Monday I bring you a conversation about how AI is changing our lives, our work, and the way we relate to each other.
00:00:28.320 --> 00:00:32.799
Today we are talking about how we use AI while keeping our own judgment.
00:00:33.039 --> 00:00:37.520
What happens when asking a tool for help becomes asking it what to think.
00:00:37.679 --> 00:00:42.399
And how do we notice when we are handing over more than we intended?
00:00:42.640 --> 00:00:49.920
My guest is Tojo Duke, an author, an advisor on responsible AI, and the founder of Diverse AI.
00:00:50.079 --> 00:00:54.960
She spent 10 years at Google, including three working on responsible AI.
00:00:55.119 --> 00:01:01.280
In this conversation, we talk about dependency, cognitive surrender, and the habits that help us think for ourselves.
00:01:01.439 --> 00:01:02.560
Let's get into it.
00:01:02.719 --> 00:01:05.680
Without further ado, welcome Tojo to the conversation.
00:01:05.920 --> 00:01:10.879
Before I bring in Tojo in, I want to tell you why this conversation matters right now.
00:01:11.040 --> 00:01:19.760
So we spend a lot of time talking on the podcast about how AI changes what's possible through hospitals, design studios, a courtroom even.
00:01:19.920 --> 00:01:27.439
What we talk about less is what it does to us in the quieter moments, when we are alone with a model and we ask it what to do.
00:01:27.599 --> 00:01:36.879
When we stop making decisions because the tool is right there, when we read our own thinking back in someone else's language and start to wonder which parts were ours.
00:01:37.120 --> 00:01:39.599
So Tojo has been paying attention to that.
00:01:39.760 --> 00:01:51.439
She works at the intersection of AI ethics and human behavior, and she thinks about dependency and autonomy and the way we are slowly, without really noticing, handling more of ourselves over to these systems.
00:01:51.599 --> 00:01:54.959
That's the conversation we are having today, and I think you need it.
00:01:55.120 --> 00:02:00.560
So, Toju, everyone is talking about how to use AI better, faster, more efficiently.
00:02:00.719 --> 00:02:03.120
But you've been paying attention to a different signal.
00:02:03.359 --> 00:02:08.479
What are you actually seeing when it comes to how people are relating to AI at a personal level, right?
00:02:09.520 --> 00:02:23.439
I think when AI came, there was lots of automation bias, and that's something that happens to humans a lot, you know, especially humans who are not from tech backgrounds, who do not work in tech tech companies and do not have an idea of how tech works.
00:02:23.599 --> 00:02:28.159
And then you're introduced to a tech technology that seems to be doing things better than you.
00:02:28.240 --> 00:02:35.919
And you're being told that the technology is gonna soon outsmart humans and it's soon gonna probably take over your jobs.
00:02:36.080 --> 00:02:47.439
And in typical human nature, we get balled over by something that seems very amazing, very impressive, and very intelligent, and we do not understand how it works.
00:02:47.680 --> 00:02:51.120
Um, so along the line, we just accept whatever we're told.
00:02:52.080 --> 00:02:54.479
And that's been the problem with AI, right?
00:02:54.639 --> 00:03:10.000
Especially if you didn't you don't work in a tech company, like you said before, or you don't have any technical backgrounds, and even for those who do, the ability to question what you're told at the very first instance is something that is a very massive uh gap in human intelligence, right?
00:03:10.080 --> 00:03:11.759
And that's what we call critical thinking.
00:03:11.919 --> 00:03:20.240
So when AI was introduced, everyone was blindsided, and now the skills are slowly dropping because the knowledge has increased.
00:03:20.400 --> 00:03:23.199
So when I started working in AI, no one cared about AI.
00:03:23.280 --> 00:03:34.080
I mean, that was long before the AI, the AI arms race even started, like way before Sunda woke up one day and said, Oh my gosh, AI is the best thing since life spread, and we have to stay in the company at Google.
00:03:34.319 --> 00:03:40.159
So at the time, I mean, I was really excited about AI, but I saw the issues with it really early on as well.
00:03:40.319 --> 00:03:51.439
But then we've had that history and that wave from 2022 when ChatGPT was launched, and everyone was like, wow, remember it was the fastest growing app in all of history.
00:03:51.680 --> 00:03:52.159
Do you remember?
00:03:52.319 --> 00:03:56.159
Within 30 days, it had over what 1 million downloads?
00:03:56.319 --> 00:03:59.199
And it was just everyone was just talking about it.
00:03:59.280 --> 00:04:02.479
But no one really wanted to understand how does it work, really?
00:04:02.639 --> 00:04:05.599
And is it really as good as we think it is?
00:04:05.759 --> 00:04:14.400
It's almost like a net, it's almost like something that entices us and we kind of forget, especially when you're chatting with a chat book that seems to be making you feel really good.
00:04:14.560 --> 00:04:15.759
Who doesn't love flattery?
00:04:15.919 --> 00:04:19.439
A lot of people do, especially depending on where you are in the spectrum, right?
00:04:19.600 --> 00:04:24.959
Issues around your self-esteem, you might even have issues with social relationships, you know.
00:04:25.040 --> 00:04:31.439
So it's easier to just stay engaged to not just the screen, but something that seems intelligent.
00:04:31.759 --> 00:04:35.519
But that's a slippish slope that we have kind of like put ourselves in.
00:04:35.680 --> 00:04:39.279
But the good thing is now people are slowly waking up to that, right?
00:04:39.360 --> 00:04:42.319
And that's why I said the automation bias is slowly wearing off.
00:04:42.399 --> 00:04:46.800
And the issue around data centers as well is making people not so happy about AI.
00:04:46.879 --> 00:04:49.360
So it's not really the best thing since LightSpace anymore.
00:04:49.439 --> 00:04:52.240
And it's not a technology that some people say they don't want anymore.
00:04:52.399 --> 00:04:58.160
And for people like me, I'm like, we still need it, but we need to redesign it and build it more responsibly, more ethically.
00:04:58.240 --> 00:05:02.000
And Hannes is better fits for the better good of all of humanity.
00:05:02.240 --> 00:05:09.759
It kind of sounds like the best word to describe all of this, how people are reacting to these circumstances is dependency, right?
00:05:09.920 --> 00:05:14.560
I'm sure you've seen it as well because you study and you talk about a lot of this and it's your work.
00:05:14.720 --> 00:05:17.680
That most people are just saying, like, well, this is just a tool.
00:05:17.839 --> 00:05:19.439
Why should I be worried about it?
00:05:19.519 --> 00:05:20.319
It's just a tool.
00:05:20.560 --> 00:05:29.680
The last part that you said, it is becoming less and less less of the thought of it's just a tool that, okay, maybe I should take a step back and look at what I'm doing.
00:05:30.000 --> 00:05:34.319
Because we don't like how the data centers are being built, what it's bringing to the environment.
00:05:34.399 --> 00:05:37.120
I think it's becoming more and more conversation, right?
00:05:37.360 --> 00:05:42.240
But where does using a tool end and a dependency begins?
00:05:42.399 --> 00:05:45.279
Because these tools are can be something helpful for us.
00:05:45.519 --> 00:05:53.519
That's why I always say, you know, in my talks and most of my posts is ensure that you're working with this tool with intention.
00:05:53.680 --> 00:05:58.560
Don't just go on an LLM or just download an AI agent for the fun of it.
00:05:58.720 --> 00:06:06.319
Like it's okay to play around with them, but if you keep on playing around with it for a very long time, first of all, it's sucking all your money, right?
00:06:06.560 --> 00:06:08.560
The credits and the tokens is it's taking a lot.
00:06:08.639 --> 00:06:10.000
So I think people are aware of that now.
00:06:10.079 --> 00:06:11.920
It's no longer just a toy or a tool.
00:06:12.000 --> 00:06:17.680
It's like this thing is costing lots of money, which is kind of good because then it's stopping people in the tracks, right?
00:06:17.839 --> 00:06:23.920
But the second thing is you need to understand what you want it to do, especially if you're working with AI for business purposes.
00:06:24.079 --> 00:06:32.480
If you think about the machine learning lifecycle and how AI is built, it's built, first of all, by training it on specific tasks.
00:06:32.639 --> 00:06:38.720
And that's where some of the problems we have with LLMs comes into play because LLMs are expected to do what they cannot do.
00:06:38.800 --> 00:06:44.480
And people think that it can, and LLMs will act like they can, and that's how it's ended up misleading people.
00:06:44.639 --> 00:06:46.720
But they're trained on specific tasks, right?
00:06:46.879 --> 00:07:05.279
And if, for example, learn how to summarize a document or you know, learn how to do a mathematical calculation according to this formula, it's actually trained on data and it's given, you know, those sort of formulas, and the algorithm is built based on those specific tasks.
00:07:05.519 --> 00:07:18.480
Now, if you as the user has a list of different things that you work on and you say these are my objectives and goals, you want to ask yourself, what can AI do for these objectives and goals?
00:07:18.560 --> 00:07:20.480
And you find out that AI cannot do everything.
00:07:20.639 --> 00:07:25.120
I always say that AI wasn't even built for enterprise, it was just built as a research tool.
00:07:25.360 --> 00:07:30.480
AI is not your Microsoft word that was built specifically for enterprise.
00:07:30.720 --> 00:07:37.920
It's a tool that came out on research labs, and that's why the leading AI providers are called Frontier AI Labs.
00:07:38.079 --> 00:07:43.600
It's still experimental, but people are paying gazillions of money for an experiment.
00:07:43.839 --> 00:07:52.079
So part of it is, you know, you cannot find one tool that will have all of the tasks and objectives and the goals that you're trying to achieve in one full tool.
00:07:52.160 --> 00:07:53.360
No, that's not possible.
00:07:53.600 --> 00:07:59.519
So you need to decide which tool you're gonna work with, or a list of tools, understand what the objectives and goals are.
00:07:59.680 --> 00:08:03.839
But the most important thing is do not offload all of your work to AI.
00:08:04.079 --> 00:08:07.360
First of all, your work is just gonna, the output is just gonna be horrible, right?
00:08:07.439 --> 00:08:08.959
We know AI slop is out there.
00:08:09.040 --> 00:08:10.879
It's not as perfect as we thought it was.
00:08:11.040 --> 00:08:15.120
But secondly, that's when the cognitive surrender comes into play, right?
00:08:15.199 --> 00:08:20.560
What's happening to your brain and to your intelligence if you're handing off all the work to something else?
00:08:20.720 --> 00:08:22.240
At what stage do you come in?
00:08:22.319 --> 00:08:26.319
At what stage do you start feeling your worth again, your sense of self-worth?
00:08:26.399 --> 00:08:28.959
I mean, work gives all of us a sense of purpose.
00:08:29.120 --> 00:08:34.480
And that's why when people retire, they find it very difficult to carry on because they're not working anymore.
00:08:34.639 --> 00:08:39.679
So if you're giving all of your work to some tool or system, then what happens to you as a person?
00:08:39.759 --> 00:08:43.759
You become irrelevant, you become useless to yourself and to the people around you, right?
00:08:43.840 --> 00:08:45.519
And you never want to be in that position.
00:08:45.679 --> 00:08:52.320
So at the end of the day, each and every one of us are in the steering shape of our careers, of our lives, and of how we adopt AI.
00:08:52.480 --> 00:08:56.080
But AI is definitely a tool that I use almost every day.
00:08:56.159 --> 00:08:57.200
I do not use it every day.
00:08:57.279 --> 00:09:01.360
I use it as and when I need to, but I know exactly when I need AI and it's helping my work.
00:09:01.519 --> 00:09:03.519
And at some point, too, I don't really have a choice.
00:09:03.600 --> 00:09:07.600
I can't go on Google search anymore and just search up for stuff and just see the links.
00:09:07.679 --> 00:09:10.559
I'm being forced to look at Gemini's results.
00:09:10.639 --> 00:09:15.120
And if I have any more questions, Gemini opens a new window for me to ask.
00:09:15.200 --> 00:09:17.200
So that is what it is, but I've accepted it.
00:09:17.360 --> 00:09:21.600
At the same time, I also always try to make sure I fact check the results as well.
00:09:21.679 --> 00:09:29.039
And I'm totally aware that they hallucinate, and I've seen many results where hallucinations are coming to play, and I ignore those results, right?
00:09:29.279 --> 00:09:35.679
But it's been very intentional about how you use AI, how you adopt it, and how you work with it every single day.
00:09:35.919 --> 00:09:43.279
How would you say to someone that it's learning right now how to use these tools in a way without losing their personality?
00:09:43.519 --> 00:09:54.480
But also to be aware of these hallucinations, because a lot of people hear about hallucinations, but they don't really know how to look for it or how not to overload with it.
00:09:54.559 --> 00:09:59.120
Because if we are always checking, it's gonna get more and more and more work to do.
00:09:59.360 --> 00:10:05.360
Yeah, so what I tend to do is I always start off with using my own cognition, my own human intelligence.
00:10:05.519 --> 00:10:08.000
I never go to AI first, and that's what I always recommend.
00:10:08.080 --> 00:10:10.480
Like, think about what it is you're trying to do, mark it out.
00:10:10.639 --> 00:10:13.919
You know, I don't play with my notebook, I always have my pen and paper.
00:10:14.080 --> 00:10:18.559
You know, I always even for a keynote, I draft out my talk on pen and paper.
00:10:18.720 --> 00:10:20.320
There you go, I've got mine here too.
00:10:20.399 --> 00:10:31.600
It might be different things for different people, but whatever it is, do not start off with going straight to your chapel or your LLM or whatever it is that you're using, your AI agents or whatever AI tool that you're using.
00:10:31.759 --> 00:10:37.519
Start off with offloading your thinking, your intelligence, what it is that you're trying to achieve.
00:10:37.679 --> 00:10:44.240
Then you can go to the AI tool, but by then you already know what you want, you already have your questions figured out.
00:10:44.399 --> 00:10:46.639
So it makes it much easier, right?
00:10:46.799 --> 00:10:50.559
You're not just searching for answers that you do not even have questions to.
00:10:50.639 --> 00:10:53.679
You're searching for answers that you know exactly what you're looking for.
00:10:53.840 --> 00:10:57.120
And a lot of people talk about using AI as an aspiring partner.
00:10:57.360 --> 00:11:00.720
Now, that's the way you can do it because now you know what you're looking for.
00:11:00.879 --> 00:11:06.879
If AI comes up with rubbish, like you know, I was preparing for a keynote the other day and um I needed some stats.
00:11:07.039 --> 00:11:09.519
I always use AI for stats for my research.
00:11:09.600 --> 00:11:13.039
And it came up with some very fake responses.
00:11:13.279 --> 00:11:15.840
But the good thing is it comes up with links.
00:11:16.000 --> 00:11:23.200
So I click on the links and then I click on the links because I always try to make sure there's no way I can just collect information from AI and use it.
00:11:23.360 --> 00:11:23.759
No way.
00:11:23.919 --> 00:11:31.360
You know, it's costs, it's cost the companies millions of pounds, and you know, their reputation has been destroyed, and I don't want to be one of those, right?
00:11:31.600 --> 00:11:34.639
And I still have my intelligence, so you know I will still use it.
00:11:34.799 --> 00:11:42.559
So anyway, I I uh I clicked on the link and saw that the link that it gave me was totally off topic from what I was looking for.
00:11:42.720 --> 00:11:45.360
And I'm like, this just made this up.
00:11:45.519 --> 00:11:46.720
So I didn't use it, right?
00:11:46.799 --> 00:11:50.080
And I looked for other examples, and after that I just used the examples that I knew.
00:11:50.240 --> 00:11:51.919
Um so that's how you fact-check.
00:11:52.000 --> 00:12:01.600
So you don't have to spend so much time fact-checking, but at the same time, you if you can narrow your scope of how you use AI in the first place, it makes your work so much easier.
00:12:01.840 --> 00:12:15.360
We've been hearing a lot about because we have these tools in our disposal, that we have been doing more work and more work that's been done before, because I can handle the few things that we can handle another ones.
00:12:15.519 --> 00:12:21.519
But that also comes to another part that we've been hearing a lot about, which is burnout from overwork.
00:12:21.679 --> 00:12:28.879
And I know this from friends that stopped completely being in tech and now they are farmers because they could not handle this kind of work.
00:12:28.960 --> 00:12:33.440
They were very burnt out, and I they are not sure if they're even going back to that.
00:12:33.679 --> 00:12:40.159
But is there such a thing as cognitive fatigue that comes specifically from over-relying on AI?
00:12:40.320 --> 00:12:45.360
Which you kind of already answered it, but in this sense, what that actually feels like.
00:12:46.240 --> 00:12:53.519
Well, I can't I can't speak for myself because I've never experienced it before, and I will not experience it because I'm totally conscious of how I use AI, right?
00:12:53.679 --> 00:13:00.720
So I call it AI-infected burnout as well, because it's burnout, but it's due to using AI and using it a bit too much.
00:13:00.879 --> 00:13:06.720
I saw a post recently which was kind of struck, is like, you know, we're always using machine-like terms on humans.
00:13:06.799 --> 00:13:11.039
Like, you know, I remember at Google, we always used to say, I don't have the bandwidth to do that.
00:13:11.360 --> 00:13:14.639
Bandwidth is not a term that you use for human beings, right?
00:13:14.799 --> 00:13:23.519
And even with the development of AI, it almost seems like, and it's something that I questioned last year, it almost seems like we're trying to replace humanity, humans, with machines.
00:13:23.679 --> 00:13:25.600
And the machines seem to be taking priority.
00:13:25.759 --> 00:13:28.240
Now, don't ask me where that's coming from and why they're doing it.
00:13:28.399 --> 00:13:31.519
Like, you know, everyone has their theories, but it's just plain stupid.
00:13:31.600 --> 00:13:40.480
If you ask me, like, why will those machines be more important than the human race, the Homo sapiens that have ex existed for donkeys of years, right?
00:13:40.720 --> 00:13:49.279
But yeah, so if we go back to the AI-inflicted burnout, we tend to use all of these terms like bandwidth, like I've switched off, right?
00:13:49.440 --> 00:13:51.759
And those are all terms used for machines.
00:13:51.919 --> 00:13:54.799
Now, if you're playing around with AI too much, you get excited.
00:13:54.879 --> 00:13:59.519
And the more you get excited, the more you want to work with it, you know, especially if you're a tech geek.
00:13:59.759 --> 00:14:07.840
But then we kind of forget, and sometimes I think people forget that AI is not human, it's not sentient, it doesn't have any consciousness.
00:14:08.000 --> 00:14:10.960
Either they forget or they don't know, or they really want it to be conscious.
00:14:11.039 --> 00:14:13.919
They're people who really want AI to be human, right?
00:14:14.000 --> 00:14:15.279
Uh, and all of that.
00:14:15.440 --> 00:14:20.480
But then, you know, the people that are always working with AI get carried away.
00:14:20.720 --> 00:14:25.440
So they have to continuously monitor the AI agents that they have.
00:14:25.840 --> 00:14:28.399
And because of that, they're not stopping.
00:14:28.559 --> 00:14:30.080
And it's almost like peer pressure.
00:14:30.159 --> 00:14:34.399
You're working side by side with a colleague that is not human, right?
00:14:34.480 --> 00:14:37.919
And the more you see that colleague working, the more you two want to work.
00:14:38.080 --> 00:14:41.039
And, you know, you hear about people who suffer from these things.
00:14:41.200 --> 00:14:46.320
Even when they try to turn it off, they have to go back and monitor it, or they'll work on other tasks.
00:14:46.399 --> 00:14:55.600
So that AI agent is saving time that they could have just gotten up and gone for a work or something, but they decide to work away, like keep on working.
00:14:55.840 --> 00:14:58.559
And because of that, they suffer from bad acts.
00:14:58.720 --> 00:15:04.240
Now, at the same time, too, we need to remember that AI agents or AI in general are not human, so they don't sleep.
00:15:04.320 --> 00:15:12.159
You know, when I hired an executive AI assistant as an agent, it was really good and efficient, but it was actually really annoying as well.
00:15:12.320 --> 00:15:15.600
So I get an email saying, hey, Tojo, this is urgent.
00:15:15.679 --> 00:15:16.399
Can you look at this?
00:15:16.559 --> 00:15:20.480
These are emails that I usually ignore because they're not emails that really need my attention.
00:15:20.559 --> 00:15:21.759
It's like generic emails.
00:15:21.919 --> 00:15:23.919
But my instants will text me at 2 a.m.
00:15:24.080 --> 00:15:27.279
in the morning to tell me that I have an urgent email that I need to respond to.
00:15:27.440 --> 00:15:32.000
Of course, there's no contextual understanding of what that email is and why.
00:15:32.159 --> 00:15:36.960
As long as it has that title urgent on it, the AI assistant just believes you need to respond to it.
00:15:37.120 --> 00:15:40.559
Or, you know, it's a bank holiday and I'm I've taken the day off.
00:15:40.720 --> 00:15:43.840
My AI agents will tell me all the things I'm planning to do today.
00:15:43.919 --> 00:15:46.480
And I'm like, yo, but it's a bank holiday.
00:15:46.559 --> 00:15:47.120
I'm not working.
00:15:47.200 --> 00:15:48.159
Oh, I'm very sorry.
00:15:48.399 --> 00:15:49.440
Enjoy your holiday.
00:15:49.600 --> 00:15:51.679
But then the next day, it does the same thing.
00:15:52.559 --> 00:15:55.279
And if you're not careful, you just fall back in.
00:15:55.360 --> 00:15:59.440
Because I didn't know I noticed a difference when I was working with this AI agent.
00:15:59.519 --> 00:16:01.279
I was literally always on my computer.
00:16:01.360 --> 00:16:09.200
Because whether I liked it or not, just knowing that there was something there to be like my research partner helped me research, but I never stopped researching.
00:16:09.360 --> 00:16:12.559
And I was just always researching and it was always responding.
00:16:12.799 --> 00:16:18.480
Until my credits finished, and I think that was the good thing for me because then I'm like, I ain't paying more money for this stuff.
00:16:18.879 --> 00:16:24.480
I have similar stories with AI agents helping me with emails, and now draft.
00:16:24.559 --> 00:16:25.840
I cannot do it anymore.
00:16:26.000 --> 00:16:27.360
It needs to go to draft.
00:16:27.440 --> 00:16:28.320
I cannot receive it.
00:16:28.480 --> 00:16:43.919
I there was one specific situation where I was talking with someone, I was asking for something from this professional uh person, and they I replied to this person saying things that were on the email but were not on the email.
00:16:44.080 --> 00:16:48.000
And well, safe to say that that person never replied to me ever again.
00:16:48.159 --> 00:16:49.679
I'm not wasting my time with this girl.
00:16:49.840 --> 00:16:52.879
For the AI to reply, I did not ask for anything.
00:16:53.120 --> 00:16:59.440
This agent decided that it was something important to reply, even though I didn't give the authorization.
00:16:59.679 --> 00:17:08.559
So even with guardrails, because I did have a lot of guardrails and context slayers and what to do, like everything was very organized.
00:17:08.640 --> 00:17:11.599
It still did this that it shouldn't be doing.
00:17:11.839 --> 00:17:13.680
So I guess I get it.
00:17:13.839 --> 00:17:15.039
I get what you're meaning.
00:17:15.200 --> 00:17:16.480
I totally understand.
00:17:16.640 --> 00:17:22.960
And it's great that you were able to overcome that and have a more healthy relationship.
00:17:23.039 --> 00:17:27.680
Because I think all of us, we are still all learning together to live beside them, right?
00:17:27.839 --> 00:17:42.640
Something that I thought it was very curious, and I'm a sci-fi lover to that sense, but there's a lot of different movies on sci-fi and different realities that might become ours, might not become ours, and even goes hand in hand with the Homo sapiens that you were talking about.
00:17:42.720 --> 00:18:06.720
Because there's a definition to Homo sapiens that say that, yes, like you said, we are here for a long time, but we also have some value that is different, it's very hard to replicate, replicate with these machines because of their logic, because they are based on logic, based on the things that are direct, based on calculations, everything that is creativity, personalization, it will never have the same strength or the same idea.
00:18:06.880 --> 00:18:14.400
And we get to decide if we are going to the direction where we lose our personality and become like the idiocracy movie.
00:18:14.480 --> 00:18:19.680
I don't know if you saw, but if not, I would advise you to take a look at it and then let me know what you think.
00:18:19.839 --> 00:18:26.480
Or we go to the other side where we live side by side with this new reality that will help us to argument.
00:18:26.640 --> 00:18:28.160
And this is my question to you.
00:18:28.319 --> 00:18:37.119
With everything that has been happening and all your knowledge and all everything that you share, what do you see these realities becoming right now with what humans have been?
00:18:37.680 --> 00:18:44.880
So as of two years ago, if you asked this question, I'd have said, we might end up being in a place where we'll live side by side with AI.
00:18:44.960 --> 00:18:55.440
I don't know if we'll get to AI KGI, but like we'll be with robots, maybe, maybe not, but definitely the adoption of AI will increase and the and the uh development of AI would increase.
00:18:55.680 --> 00:19:02.160
But today, and this is this is a response that makes me happy, is that I don't think that will happen.
00:19:02.480 --> 00:19:02.799
Right?
00:19:03.119 --> 00:19:04.480
There are two ways it will go.
00:19:04.640 --> 00:19:17.279
It's either it's just gonna carry on the way it is, but we're not gonna see massive increased capabilities or more people being blindsided by AI and pumping money into AI.
00:19:17.440 --> 00:19:20.079
And again, the issue around data centers is a huge thing.
00:19:20.160 --> 00:19:26.640
And we have also have people who have lost their lives, especially young young people who have lost their lives due to LLMs and chatbots, right?
00:19:26.720 --> 00:19:32.559
Uh so and then China, on the other hand, is almost becoming an AI superpower, right?
00:19:32.799 --> 00:19:38.640
They're releasing models that are literally as good as the legal models in the US now for a fraction of the price.
00:19:38.880 --> 00:19:43.359
So companies are thinking about should they start using Chinese models and what are the risks there?
00:19:43.519 --> 00:19:48.079
So with all of that, we're not going to see this AGI that we're talking about.
00:19:48.240 --> 00:19:55.440
Actually, I actually wrote a newsletter a few weeks ago saying it's the end of AGI because to build AGI, we need gazillions amounts of money, right?
00:19:55.519 --> 00:19:57.680
Just to even fund that research.
00:19:57.839 --> 00:20:02.160
And all of these LLMs and AI agents were supposed to be leading to AGI.
00:20:02.319 --> 00:20:05.279
But no one is like at some point the money's gonna finish anyway.
00:20:05.359 --> 00:20:07.920
Like, you know, there is no money tree out there.
00:20:08.079 --> 00:20:11.039
Money's not dropping from from the moon or from the sun, right?
00:20:11.200 --> 00:20:17.119
So if people keep on popping money into AI and they don't see the return on investment, the investment is gonna stop.
00:20:17.519 --> 00:20:22.000
So what sort of what sort of investment are we gonna have to drive AGI?
00:20:22.400 --> 00:20:28.240
Because it's clear that to get the sort of results we have seen today, we need a lot of compute, right?
00:20:28.400 --> 00:20:37.759
Um and there was this talk about scaling hit a wall last year, where they were saying, you know, you cannot keep on scaling because scaling now is not working and the capabilities are not really increasing.
00:20:37.920 --> 00:20:38.880
And that's true.
00:20:39.039 --> 00:20:46.880
So the whole point of AI or any technology really is to benefit humanity, is to improve people's and humans' lives.
00:20:47.039 --> 00:20:50.160
You know, we have Dennis Hassabiz, who's the CEO of Deep Mind.
00:20:50.240 --> 00:20:59.839
And he got really popular with the work that he's been doing with Deep Mind, which was never around profit making, but it was always towards healthcare and revol revolutionizing healthcare, right?
00:21:00.079 --> 00:21:01.759
And that's the main purpose of AI.
00:21:02.000 --> 00:21:06.720
You know, what I believe will happen in the future is that there will be a redesign of AI.
00:21:06.880 --> 00:21:16.000
Because now lawmakers are waking up, and we just need to redesign the whole overall flow and pipeline and development of AI.
00:21:16.160 --> 00:21:27.920
Right from the entire supply chain, right from the raw materials that we use to build the AI chips, and down to the AI chips, and then down to the data centers, and down to the people who label the data, right?
00:21:28.000 --> 00:21:35.119
And down to the data that is being collected and the ways we collect it, and the fact that we're not compensating people or even having their consent.
00:21:35.359 --> 00:21:37.200
So all of that needs to change.
00:21:37.359 --> 00:21:40.240
And ideally, AI should be built for specific purposes.
00:21:40.400 --> 00:21:44.720
You know, to get the best results out of any AI system, it's when it's a smaller model.
00:21:44.960 --> 00:21:47.039
A generic model doesn't work.
00:21:47.119 --> 00:21:57.599
It's just like having a doctor who is a plumber, a chef, a wife, a mother, which are full time jobs on their own, a librarian, an author, a musician, all in one person.
00:21:57.759 --> 00:22:02.720
That's just setting, that's just making a A disaster and that's a disaster waiting to happen.
00:22:03.039 --> 00:22:03.200
Right.
00:22:03.279 --> 00:22:04.400
So that's the same thing with AI.
00:22:04.720 --> 00:22:06.640
It kind of sounds like we are going back to that.
00:22:06.799 --> 00:22:13.359
We back then we saw that it didn't work, but for some reason right now they're going back to it and thinking that it works.
00:22:13.680 --> 00:22:25.279
There's a a distinction I'd love to explore with you based also on what you were saying, because the difference between how someone uses AI personally versus how it shows up inside an organization or a business is different.
00:22:25.359 --> 00:22:26.240
It's completely different.
00:22:26.319 --> 00:22:28.160
There's different different ways of using it, right?
00:22:28.240 --> 00:22:33.359
So do those two contexts require a completely different set of guardrails?
00:22:33.759 --> 00:22:36.079
I don't know if I'll say guide roles as opposed to approaches.
00:22:36.160 --> 00:22:39.759
Because guardrils, when we talk about guide roles, it's back to the technicalities of it.
00:22:39.839 --> 00:22:42.880
And you know, once a guide roll is in there, do the guide roles work?
00:22:42.960 --> 00:22:44.079
That's a whole different story, right?
00:22:44.319 --> 00:22:46.799
Like guardrails around AI agents don't really work.
00:22:46.880 --> 00:22:49.279
And that's why what happened to you happened to you, right?
00:22:49.519 --> 00:22:51.599
That email, annoying email that was sent.
00:22:51.920 --> 00:22:53.039
It happened to me as well.
00:22:53.119 --> 00:22:57.759
You know, my AI assistants decided to share my meeting notes with two people outside my organization.
00:22:57.839 --> 00:23:00.559
One person totally random, I do not know who the person is.
00:23:00.720 --> 00:23:04.720
And you know, to make matters worse, I did not I did not even have edit rights to the document.
00:23:04.880 --> 00:23:06.559
So I could not revoke their access.
00:23:06.640 --> 00:23:07.920
I had to delete the document.
00:23:08.240 --> 00:23:12.720
And when you go back to the company who owns the agent, they're like, that's that's unintended behavior.
00:23:12.799 --> 00:23:13.920
I'm like, no, it's not.
00:23:14.079 --> 00:23:18.079
That is actually how AI agents operate, and we do not have a clue on how to start them.
00:23:18.400 --> 00:23:18.559
Right.
00:23:18.640 --> 00:23:24.319
And you hear a lot of stories about people who have AI agents have deleted an entire production database.
00:23:24.559 --> 00:23:28.240
So the guide rules don't always work, and it's sometimes it's hit or miss.
00:23:28.400 --> 00:23:30.480
But the approaches are really important, right?
00:23:30.640 --> 00:23:34.400
And you know, with the business approaches is working with a framework.
00:23:34.480 --> 00:23:41.119
So I actually came up with this very simple framework that I say, you know, it's a three A's is ask, assess, and analyze.
00:23:41.519 --> 00:23:44.960
So ask yourself the questions on what do you need this AI tool to do?
00:23:45.119 --> 00:23:47.839
What can it actually do properly before you use it?
00:23:47.920 --> 00:23:49.839
And then when you use it, look at the results.
00:23:49.920 --> 00:23:50.799
That's a fact checking.
00:23:50.960 --> 00:24:00.640
And then after that, you want to analyze on an ongoing basis, you know, the results and the amount of money spent and the amount of time spent and how you're feeling, right?
00:24:00.799 --> 00:24:03.519
And is your brain still working when you use AI?
00:24:03.759 --> 00:24:09.039
Now, that's for the business purposes, but you can still apply that framework to the personnel purposes.
00:24:09.200 --> 00:24:14.160
Now, the difference with the personnel purposes is that many times it depends on who it is, right?
00:24:14.400 --> 00:24:25.680
You hear about a lot of women and moms that need to plan a holiday or looking for recipes or something, they're already going with a specific task and objective in mind, and that's what they need AI to do.
00:24:25.839 --> 00:24:33.599
And you hear about some people too who are going through like relationship issues and they're having a lot of issues around their self-esteem and they go to AI to validate them.
00:24:33.759 --> 00:24:43.920
And those sort of people who are using AI as companions or relationship companions are the ones that struggle the most on the personal side of things because there's already a gap, right?
00:24:44.079 --> 00:24:46.880
You know, emotionally, a gap needs to be filled.
00:24:46.960 --> 00:24:58.400
And if if their partner is not supplying that to them, the AI, horrible AI agents or chatbots are very good at feeding filling in that gap and they become addicted.
00:24:58.559 --> 00:25:00.240
And that's where the problem lies.
00:25:00.400 --> 00:25:08.559
So again, the same approach to the business approach is have your tasks, know what you do it that you need to get from the AI agent or the chatbot.
00:25:08.720 --> 00:25:10.400
Once you've got it, just let it go.
00:25:10.559 --> 00:25:22.240
Like remember that you're human and the things that that fuel our humanity that keeps us going and happy are not the same things that fuel AI agents, LLMs, or Gen AI.
00:25:22.480 --> 00:25:23.920
We don't live on AI chips.
00:25:24.079 --> 00:25:26.240
Back to back to what makes us human.
00:25:26.640 --> 00:25:26.880
Yeah.
00:25:27.279 --> 00:25:28.799
We need to always uphold that.
00:25:28.960 --> 00:25:36.240
And it's kind of good AI is is in our mist because it kind of like serves as a reminder that we're still human and we need to keep ourselves going.
00:25:36.319 --> 00:25:38.160
Like most people want to live long, right?
00:25:38.319 --> 00:25:42.240
And they want to be happy and they want to enjoy the fruit of their labor.
00:25:42.319 --> 00:25:47.440
Like, you know, you work hard, you work nine to five of how many hours you work, you know, five days a week.
00:25:47.519 --> 00:25:49.039
Some people work seven days a week.
00:25:49.119 --> 00:25:49.200
Yeah.
00:25:49.359 --> 00:25:51.039
And some people even do different jobs.
00:25:51.119 --> 00:25:55.920
But whatever it is, there's lots of hard labor that goes through our lives as human beings.
00:25:56.079 --> 00:25:58.960
And we always want to improve our lives some way or the other.
00:25:59.119 --> 00:26:06.640
Some people are trying to achieve, you know, financial, financial improvements, social relationship, love, whatever it is.
00:26:06.799 --> 00:26:08.960
These are the different things that make us human.
00:26:09.200 --> 00:26:13.759
And the more you kind of work with a non-human entity, you get affected.
00:26:13.839 --> 00:26:19.200
And we have seen that the human, AI relationships and symbols is clash.
00:26:19.359 --> 00:26:20.079
It doesn't work.
00:26:20.160 --> 00:26:25.839
And unfortunately, the humans get affected because we're the ones that have a physical body and a soul and emotions.
00:26:26.160 --> 00:26:28.480
We have the things that can get affected.
00:26:28.640 --> 00:26:30.480
But AI doesn't have that.
00:26:30.640 --> 00:26:32.880
You can just turn off that server and it's gone.
00:26:32.960 --> 00:26:36.640
It's not even gonna shout yeah as it's going, it's just gone, right?
00:26:36.880 --> 00:26:48.400
Before we go further, if you're enjoying this conversation, subscribe to Human X Intelligent wherever you're watching or listening, so you can join me with my guests for a new episode every Monday.
00:26:48.799 --> 00:26:56.640
And I do love to know when you have to make a decision, do you think true yourself before opening artificial intelligence?
00:26:56.799 --> 00:26:58.720
Tell me in the comments or send me a message.
00:26:58.880 --> 00:27:02.160
I'd love to know how this shows up in your own life.
00:27:02.319 --> 00:27:04.160
And now let's go back to Tojo.
00:27:04.559 --> 00:27:09.440
Before I bring in Tojo in, I want to tell you why this conversation matters right now.
00:27:09.599 --> 00:27:18.319
So we spend a lot of time talking in the podcast about how AI changes what's possible through hospitals, design studios, a courtroom even.
00:27:18.480 --> 00:27:26.000
What we talk about less is what it does to us in the quieter moments, when we are alone with a model and we ask it what to do.
00:27:26.160 --> 00:27:35.519
When we stop making decisions because the tool is right there, when we read our own thinking back in someone else's language and start to wonder which parts were ours.
00:27:35.680 --> 00:27:38.160
So Tojo has been paying attention to that.
00:27:38.319 --> 00:27:50.079
She works at the intersection of AI ethics and human behavior, and she thinks about dependency and autonomy and the way we are slowly, without really noticing, handling more of ourselves over to these systems.
00:27:50.240 --> 00:27:53.519
That's the conversation we are having today, and I think you need it.
00:27:53.680 --> 00:27:59.119
So, Toju, everyone is talking about how to use AI better, faster, more efficiently.
00:27:59.279 --> 00:28:01.680
But you've been paying attention to a different signal.
00:28:01.920 --> 00:28:06.799
What are you actually seeing when it comes to how people are relating to AI at a personal level?
00:28:06.880 --> 00:28:07.039
Right.
00:28:11.359 --> 00:28:22.079
And that's something that happens to humans a lot, you know, especially humans who are not from tech backgrounds, who do not work in tech, tech companies, and do not have an idea of how tech works.
00:28:22.240 --> 00:28:26.720
And then you're introduced to a tech technology that seems to be doing things better than you.
00:28:26.880 --> 00:28:34.480
And you're being told that the technology is gonna soon outsmart humans and it's soon gonna probably take over your jobs.
00:28:34.720 --> 00:28:46.079
And in typical human nature, we get balled over by something that seems very amazing, very impressive, and very intelligent, and we do not understand how it works.
00:28:46.240 --> 00:28:50.000
Um, so along the line, we just accept whatever we're told.
00:28:50.559 --> 00:28:53.039
And that's been the problem with AI, right?
00:28:53.200 --> 00:29:08.559
Especially if you didn't you don't work in a tech company, like you said before, or you don't have any technical backgrounds, and even for those who do, the ability to question what you're told at the very first instance is something that is a very massive uh gap in human intelligence, right?
00:29:08.640 --> 00:29:10.319
And that's what we call critical thinking.
00:29:10.480 --> 00:29:18.799
So when AI was introduced, everyone was blindsided, and now the skills are slowly dropping because the knowledge has increased.
00:29:18.960 --> 00:29:21.759
So when I started working in AI, no one cared about AI.
00:29:21.920 --> 00:29:32.720
I mean, that was long before the AI, the AI arms race even started, like way before Sunda woke up one day and said, Oh my gosh, AI is the best thing since life spread, I'm gonna have to stay in the company at Google.
00:29:32.880 --> 00:29:38.720
So at the time, I mean I was really excited about AI, but I saw the issues with it really early on as well.
00:29:38.880 --> 00:29:50.079
But then we've had that history and that wave from 2022 when ChatGPT was launched, and everyone was like, wow, remember it was the fastest growing app in all of history.
00:29:50.240 --> 00:29:50.799
Do you remember?
00:29:50.880 --> 00:29:54.720
Within 30 days, it had over what 1 million downloads?
00:29:54.960 --> 00:29:57.759
And it was just everyone was just talking about it.
00:29:57.920 --> 00:30:01.039
But no one really wanted to understand how does it work really?
00:30:01.200 --> 00:30:04.160
And is it really as good as we think it is?
00:30:04.400 --> 00:30:12.960
It's almost like a net, it's almost like something that entices us and we kind of forget, especially when you're chatting with a chatbot that seems to be making you feel really good.
00:30:13.119 --> 00:30:14.319
Who doesn't love flattery?
00:30:14.480 --> 00:30:18.000
A lot of people do, especially depending on where you are in the spectrum, right?
00:30:18.160 --> 00:30:23.519
Issues around your self-esteem, you might even have issues with social relationships, you know.
00:30:23.680 --> 00:30:30.000
So it's easier to just stay engaged to not just the screen, but something that seems intelligent.
00:30:30.319 --> 00:30:34.079
But that's a sleepish slope that we have kind of like put ourselves in.
00:30:34.240 --> 00:30:37.839
But the good thing is now people are slowly waking up to that, right?
00:30:37.920 --> 00:30:40.880
And that's why I said the automation bias is slowly wearing off.
00:30:40.960 --> 00:30:45.440
And the issue around data centers as well is making people not so happy about AI.
00:30:45.519 --> 00:30:47.920
So it's not really the best thing since LightSpace anymore.
00:30:48.000 --> 00:30:50.880
And it's not a technology that some people say they don't want anymore.
00:30:50.960 --> 00:30:56.720
And for people like me, I'm like, we still need it, but we need to redesign it and build it more responsibly, more ethically.
00:30:56.799 --> 00:31:00.559
And Hannes is better if it's for the better good of all of humanity.
00:31:00.799 --> 00:31:08.319
It kind of sounds like the best word to describe all of this, how people are reacting to these circumstances is dependency, right?
00:31:08.480 --> 00:31:13.200
I'm sure you've seen it as well because you study and you talk about a lot of this and it's your work.
00:31:13.279 --> 00:31:16.240
That most people are just saying, like, well, this is just a tool.
00:31:16.400 --> 00:31:18.000
Why should I be worried about it?
00:31:18.079 --> 00:31:18.960
It's just a tool.
00:31:19.119 --> 00:31:28.240
The last part that you said, it is becoming less and less less of the thought of it's just a tool that okay, maybe I should take a step back and look at what I'm doing.
00:31:28.559 --> 00:31:32.880
Because we don't like how the data centers are being built, what it's bringing to the environment.
00:31:33.039 --> 00:31:35.759
I think it's becoming more and more conversation, right?
00:31:35.920 --> 00:31:40.880
But where does using a tool end and a dependency begins?
00:31:41.039 --> 00:31:43.920
Because these tools are can be something helpful for us.
00:31:44.160 --> 00:31:52.079
That's why I always say, you know, in my talks, in most of my posts, is ensure that you're working with this tool with intention.
00:31:52.319 --> 00:31:57.200
Don't just go on an LLM or just download an AI agent for the fun of it.
00:31:57.279 --> 00:32:04.880
Like it's okay to play around with them, but if you keep on playing around with it for a very long time, first of all, it's it's sucking all your money, right?
00:32:04.960 --> 00:32:07.119
So the credits on the tokens is it's taking a lot.
00:32:07.200 --> 00:32:08.559
So I think people are aware of that now.
00:32:08.640 --> 00:32:10.480
It's no longer just a toy or a tool.
00:32:10.640 --> 00:32:16.240
It's like this thing is costing lots of money, which is kind of good because then it's stopping people in the tracks, right?
00:32:16.480 --> 00:32:22.480
But the second thing is you need to understand what you want it to do, especially if you're working with AI for business purposes.
00:32:22.720 --> 00:32:31.039
If you think about the machine learning lifecycle and how AI is built, it's built first of all by training it on specific tasks.
00:32:31.200 --> 00:32:37.279
And that's where some of the problems we have with LLMs comes into play because LLMs are expected to do what they cannot do.
00:32:37.359 --> 00:32:43.039
And people think that it can, and LLMs will act like they can, and that's how it's ended up misleading people.
00:32:43.279 --> 00:32:45.359
But they're trained on specific tasks, right?
00:32:45.440 --> 00:33:03.839
And if, for example, learn how to summarize a document or you know, learn how to do a mathematical calculation according to this formula, it's actually trained on data and it's given, you know, those sort of formulas, and the algorithm is built based on those specific tasks.
00:33:04.079 --> 00:33:17.039
Now, if you as the user has a list of different things that you work on and you say these are my objectives and goals, you want to ask yourself, what can AI do for these objectives and goals?
00:33:17.119 --> 00:33:19.119
And you find out that AI cannot do everything.
00:33:19.279 --> 00:33:23.759
I always say that AI wasn't even built for enterprise, it was just built as a research tool.
00:33:23.920 --> 00:33:36.480
AI is not your Microsoft Word that was built specifically for enterprise, it's a tool that came out of research labs, and that's why the leading AI providers are called Frontier AI Labs.
00:33:36.640 --> 00:33:42.240
It's still experimental, but people are paying gazillions of money for an experiment.
00:33:42.480 --> 00:33:50.640
So part of it is, you know, you cannot find one tool that will have all of the tasks and objectives and the goals that you're trying to achieve in one full tool.
00:33:50.799 --> 00:33:51.920
No, that's not possible.
00:33:52.160 --> 00:33:58.079
So you need to decide which tool you're gonna work with, or list of tools, understand what the objectives and goals are.
00:33:58.319 --> 00:34:02.480
But the most important thing is do not offload all of your work to AI.
00:34:02.640 --> 00:34:05.920
First of all, your work is just gonna, the output is just gonna be horrible, right?
00:34:06.000 --> 00:34:07.519
We know AI slop is out there.
00:34:07.599 --> 00:34:09.440
It's not as perfect as we thought it was.
00:34:09.599 --> 00:34:13.679
But secondly, that's when the cognitive surrender comes into play, right?
00:34:13.840 --> 00:34:19.119
What's happening to your brain and to your intelligence if you're handing off all the work to something else?
00:34:19.280 --> 00:34:20.800
At what stage do you come in?
00:34:20.880 --> 00:34:24.880
At what stage do you start feeling your worth again, your sense of self-worth?
00:34:24.960 --> 00:34:27.519
I mean, work gives all of us a sense of purpose.
00:34:27.679 --> 00:34:33.039
And that's why when people retire, they find it very difficult to carry on because they're not working anymore.
00:34:33.199 --> 00:34:38.239
So if you're giving all of your work to some tool or system, then what happens to you as a person?
00:34:38.320 --> 00:34:42.320
You become irrelevant, you become useless to yourself and to the people around you, right?
00:34:42.400 --> 00:34:44.159
And you never want to be in that position.
00:34:44.239 --> 00:34:50.960
So at the end of the day, each and every one of us are in the steering ship of our careers, of our lives, and of how we adopt AI.
00:34:51.039 --> 00:34:54.639
But AI is definitely a tool that I use almost every day.
00:34:54.800 --> 00:34:55.760
I do not use it every day.
00:34:55.840 --> 00:35:00.000
I use it ask them when I need to, but I know exactly when I need AI and it's helping my work.
00:35:00.079 --> 00:35:02.079
And at some point, too, I don't really have a choice.
00:35:02.159 --> 00:35:06.159
I can't go on Google search anymore and just search up for stuff and just see the links.
00:35:06.239 --> 00:35:09.119
I'm being forced to look at Gemini's results.
00:35:09.199 --> 00:35:13.679
And if I have any more questions, Gemini opens a new window for me to ask.
00:35:13.840 --> 00:35:15.840
So that is what it is, but I've accepted it.
00:35:15.920 --> 00:35:20.239
At the same time, I also also always try to make sure I fact-check the results as well.
00:35:20.320 --> 00:35:27.599
And I'm totally aware that they hallucinate, and I've seen many results where hallucinations have come into play, and I ignore those results, right?
00:35:27.840 --> 00:35:34.239
But it's been very intentional about how you use AI, how you adopt it, and how you work with it every single day.
00:35:34.480 --> 00:35:41.920
How would you say to someone that it's learning right now how to use these tools in a way without losing their personality?
00:35:42.079 --> 00:35:53.039
But also to be aware of these hallucinations, because a lot of people hear about hallucinations, but they don't really know how to look for it or how not to overload with it.
00:35:53.119 --> 00:35:57.679
Because if we are always checking, it's gonna get more and more and more work to do.
00:35:57.920 --> 00:36:03.920
Yeah, so what I tend to do is I always start off with using my own cognition, my own human intelligence.
00:36:04.079 --> 00:36:06.639
I never go to AI first, and that's what I always recommend.
00:36:06.719 --> 00:36:09.119
Like, think about what it is you're trying to do, mark it out.
00:36:09.199 --> 00:36:12.559
You know, I don't play with my notebook, I always have my pen and paper.
00:36:12.719 --> 00:36:17.199
You know, I always even for a keynote, I draft out my talk on pen and paper.
00:36:17.360 --> 00:36:18.880
There you go, I've got mine here too.
00:36:18.960 --> 00:36:30.159
It might be different things for different people, but whatever it is, do not start off with going straight to your chapel or your LLM or whatever it is that you're using, your AI agents or whatever AI tool that you're using.
00:36:30.400 --> 00:36:36.079
Start off with offloading your thinking, your intelligence, what it is that you're trying to achieve.
00:36:36.239 --> 00:36:42.800
Then you can go to the AI tool, but by then you already know what you want, you already have your questions figured out.
00:36:42.960 --> 00:36:45.280
So it makes it much easier, all right?
00:36:45.360 --> 00:36:52.320
You're not just searching for answers that you do not even have questions to, you're searching for answers that you know exactly what you're looking for.
00:36:52.480 --> 00:36:55.679
And a lot of people talk about using AI as aspiring partner.
00:36:55.920 --> 00:36:59.280
Now, that's the way you can do it because now you know what you're looking for.
00:36:59.440 --> 00:37:05.440
If AI comes up with rubbish, like you know, I was preparing for a keynote the other day and um I needed some stats.
00:37:05.599 --> 00:37:08.079
I always use AI for stats for my research.
00:37:08.239 --> 00:37:11.599
And it came up with some very fake responses.
00:37:11.840 --> 00:37:14.400
But the good thing is it comes up with links.
00:37:14.559 --> 00:37:21.760
So I click on the links and then I click on the links because I always try to make sure there's no way I can just collect information from AI and use it.
00:37:21.920 --> 00:37:22.400
No way.
00:37:22.480 --> 00:37:29.920
You know, it's cost it's cost the companies millions of pounds, and you know, their reputation has been destroyed, and I don't want to be one of those, right?
00:37:30.159 --> 00:37:33.199
And I still have my intelligence, so you know I will still use it.
00:37:33.360 --> 00:37:43.920
So anyway, I I uh I clicked on the link and saw that the link that it gave me was totally off topic from what I was looking for, and I'm like, this just made this up.
00:37:44.079 --> 00:37:45.280
So I didn't use it, right?
00:37:45.360 --> 00:37:48.719
And I looked for other examples, and after that, I just used the examples that I knew.
00:37:48.800 --> 00:37:50.480
Um so that's how you fact-check.
00:37:50.559 --> 00:38:00.159
So you don't have to spend so much time fact-checking, but at the same time, you if you can narrow your scope of how you use AI in the first place, it makes your work so much easier.
00:38:00.400 --> 00:38:13.920
We've been hearing a lot about because we have these tools in our disposal, that we have been doing more work and more work than been done before, because I can handle the few things that we can handle another ones.
00:38:14.079 --> 00:38:20.079
But that also comes to another part that we've been hearing a lot about, which is burnout from overwork.
00:38:20.239 --> 00:38:27.440
And I know this from friends that stopped completely being in tech and now they are farmers because they could not handle this kind of work.
00:38:27.519 --> 00:38:32.079
They were very burnt out, and I they are not sure if they're even going back to that.
00:38:32.239 --> 00:38:38.719
But is there such a thing as cognitive fatigue that comes specifically from over-relying on AI?
00:38:38.880 --> 00:38:44.000
Which you kind of already answered it, but in this sense, what that actually feels like.
00:38:44.800 --> 00:38:52.079
Well, I can't I can't speak for myself because I've never experienced it before, and I will not experience it because I'm totally conscious of how I use AI, right?
00:38:52.239 --> 00:38:59.280
So I call it AI-infected burnout as well, because it's burnout, but it's due to using AI and using it a bit too much.
00:38:59.440 --> 00:39:05.280
I saw a post recently which was kind of struck, is like, you know, we're always using machine-like terms on humans.
00:39:05.440 --> 00:39:09.679
Like, you know, I remember at Google we always used to say, I don't have the bandwidth to do that.
00:39:09.920 --> 00:39:13.199
Bandwidth is not a term that you use for human beings, right?
00:39:13.360 --> 00:39:22.159
And even with the development of AI, it almost seems like, and it's something that I questioned last year, it almost seems like we're trying to replace humanity, humans, with machines.
00:39:22.239 --> 00:39:24.239
And the machines seem to be taking priority.
00:39:24.320 --> 00:39:26.880
Now, don't ask me where that's coming from and why they're doing it.
00:39:26.960 --> 00:39:30.079
Like, you know, everyone has their theories, but it's just plain stupid.
00:39:30.159 --> 00:39:39.119
If you ask me, like, why will those machines be more important than the human race, the Homo sapiens that have existed for donkeys of years, right?
00:39:39.280 --> 00:39:47.840
But yeah, so if we go back to the AI-inflicted burnout, we tend to use all of these terms like bandwidth, like I've switched off, right?
00:39:48.000 --> 00:39:50.320
And those are all terms used for machines.
00:39:50.559 --> 00:39:53.360
Now, if you're playing around with AI too much, you get excited.
00:39:53.440 --> 00:39:58.079
And the more you get excited, the more you want to work with it, you know, especially if you're a tech geek.
00:39:58.320 --> 00:40:06.480
But then we kind of forget, and sometimes I think people forget that AI is not human, it's not sentient, it doesn't have any consciousness.
00:40:06.559 --> 00:40:09.599
Either they forget or they don't know, or they really want it to be conscious.
00:40:09.679 --> 00:40:12.559
They're people who really want AI to be human, right?
00:40:12.639 --> 00:40:13.840
Uh, and all of that.
00:40:14.079 --> 00:40:19.119
But then, you know, the people that are always working with AI get carried away.
00:40:19.280 --> 00:40:24.000
So they have to continuously monitor the AI agents that they have.
00:40:24.400 --> 00:40:27.039
And because of that, they're not stopping.
00:40:27.119 --> 00:40:28.639
And it's almost like peer pressure.
00:40:28.719 --> 00:40:32.960
You're working side by side with a colleague that is not human, right?
00:40:33.039 --> 00:40:36.480
And the more you see that colleague working, the more you two want to work.
00:40:36.639 --> 00:40:39.679
And you know, you hear about people who suffer from these things.
00:40:39.840 --> 00:40:44.880
Even when they try to turn it off, they have to go back and monitor it, or they'll work on other tasks.
00:40:44.960 --> 00:40:54.239
So that AI agent is saving time that they could have just gotten up and gone for a work or something, but they decide to work away, like keep on working.
00:40:54.400 --> 00:40:57.119
And because of that, they suffer from bad apps.
00:40:57.280 --> 00:41:02.800
Now, at the same time, too, we need to remember that AI agents or AI in general are not human, so they don't sleep.
00:41:02.960 --> 00:41:10.719
You know, when I hired an executive AI assistant as an agent, it was really good and efficient, but it was actually really annoying as well.
00:41:10.880 --> 00:41:14.159
So I get an email saying, hey, Tojo, this is urgent.
00:41:14.320 --> 00:41:14.960
Can you look at this?
00:41:15.119 --> 00:41:19.039
These are emails that I usually ignore because they're not emails that really need my attention.
00:41:19.119 --> 00:41:20.400
It's like generic emails.
00:41:20.559 --> 00:41:22.480
But my agents will text me at 2 a.m.
00:41:22.639 --> 00:41:25.920
in the morning to tell me that I have an urgent email that I need to respond to.
00:41:26.079 --> 00:41:30.639
Of course, there's no contextual understanding of what that email is and why.
00:41:30.719 --> 00:41:35.599
As long as it has that title urgent on it, the AI assistant just believes you need to respond to it.
00:41:35.679 --> 00:41:39.119
Or, you know, it's a bank holiday and um I've taken the day off.
00:41:39.360 --> 00:41:42.400
My AI agents will tell me all the things I'm planning to do today.
00:41:42.480 --> 00:41:45.039
And I'm like, yo, but it's a bank holiday.
00:41:45.119 --> 00:41:45.679
I'm not working.
00:41:45.760 --> 00:41:46.800
Oh, I'm very sorry.
00:41:46.960 --> 00:41:48.000
Enjoy your holiday.
00:41:48.159 --> 00:41:50.159
But then the next day, it does the same thing.
00:41:51.199 --> 00:41:53.840
And if you're not careful, you just fall back in.
00:41:53.920 --> 00:41:58.000
Because I didn't know I noticed a difference when I was working with this AI agent.
00:41:58.079 --> 00:42:07.760
I was literally always on my computer because whether I liked it or not, just knowing that there was something there to be like my research partner helped me research, but I never stopped researching.
00:42:07.920 --> 00:42:17.039
And I was just always researching, and it was always responding until my credits finished, and I think that was the good thing for me because then I'm like, I ain't paying more money for this stuff.
00:42:17.440 --> 00:42:24.400
I have similar stories with AI agents helping me with emails, and now draft, I cannot do it anymore.
00:42:24.559 --> 00:42:26.960
It needs to go to draft, I cannot receive it.
00:42:27.039 --> 00:42:42.480
I there was one specific situation where I was talking with someone, I was asking for something from this professional uh person, and they I replied to this person saying things that were on the email but were not on the email.
00:42:42.719 --> 00:42:46.559
And well, safe to say that that person never replied to me ever again.
00:42:46.719 --> 00:42:48.239
I'm not wasting my time with this girl.
00:42:48.480 --> 00:42:51.519
For the AI to reply, I did not ask for anything.
00:42:51.760 --> 00:42:58.079
This agent decided that it was something important to reply, even though I didn't give the authorization.
00:42:58.239 --> 00:43:07.119
So even with guardrails, because I did have a lot of guardrails and context slayers and what to do, like everything was very organized.
00:43:07.199 --> 00:43:10.159
It still did this that it shouldn't be doing.
00:43:10.400 --> 00:43:12.239
So I guess I get it.
00:43:12.400 --> 00:43:13.679
I get what you're meaning.
00:43:13.760 --> 00:43:15.039
I totally understand.
00:43:15.199 --> 00:43:21.519
And it's great that you were able to overcome that and have a more healthy relationship.
00:43:21.599 --> 00:43:26.239
Because I think all of us, we are still all learning together to live beside them, right?
00:43:26.480 --> 00:43:41.199
Something that I thought it was very curious, and I'm a sci-fi lover to that sense, but there's a lot of different movies on sci-fi and different realities that might become ours, might not become ours, and even goes hand in hand with the Homo sapiens that you were talking about.
00:43:41.360 --> 00:43:59.760
Because there's a definition to Homo sapiens that says that, yes, like you said, we are here for a long time, but we also have some value that is different, it's very hard to replicate, replicate with these machines because of their logic, because they are based on logic, based on things that are direct, based on calculations, everything that is creative.
00:44:00.480 --> 00:44:05.280
Personalization, it will never have the same strength or the same idea.
00:44:05.440 --> 00:44:12.960
And we get to decide if we are going to the direction where we lose our personality and become like the idiocracy movie.
00:44:13.039 --> 00:44:18.320
I don't know if you saw, but if not, I would advise you to take a look at it and then let me know what you think.
00:44:18.480 --> 00:44:25.119
Or we go to the other side where we live side by side with this new reality that will help us to augment.
00:44:25.280 --> 00:44:35.760
And this is my question to you with everything that has been happening and all your knowledge and everything that you share, what do you see these realities becoming right now with what humans have been?
00:44:36.239 --> 00:44:43.440
So as of two years ago, if you'd asked this question, I'd have said, we might end up being in a place where we'll leave side by side with AI.
00:44:43.519 --> 00:44:47.840
I don't know if we'll get to AI KGI, but like we'll be with robots, maybe, maybe not.
00:44:48.000 --> 00:44:54.000
But definitely the adoption of AI will increase and the and the uh development of AI will increase.
00:44:54.239 --> 00:45:00.719
But today, and this is this is a response that makes me happy, is that I don't think that will happen.
00:45:01.039 --> 00:45:01.360
Right?
00:45:01.679 --> 00:45:03.039
There are two ways it will go.
00:45:03.199 --> 00:45:15.840
It's either it's just gonna carry on the way it is, but we're not gonna see massive increased capabilities or more people being blindsided by AI and pumping money into AI.
00:45:16.000 --> 00:45:18.639
And again, the issue around data centers is a huge thing.
00:45:18.719 --> 00:45:25.199
And we also have people who have lost their lives, especially young young people who have lost their lives due to LLMs and chatbots, right?
00:45:25.280 --> 00:45:31.119
Um so, and then China, on the other hand, is almost becoming an AI superpower, right?
00:45:31.360 --> 00:45:37.280
They're releasing models that are literally as good as the leading models in the US now for a fraction of the price.
00:45:37.440 --> 00:45:41.920
So companies are thinking about should they start using Chinese models and what are the risks there?
00:45:42.079 --> 00:45:46.639
So all of that, we're not going to see this AGI that we're talking about.
00:45:46.800 --> 00:45:54.000
Actually, I actually wrote a newsletter a few weeks ago saying it's the end of AGI because to build AGI, we need gazillions' amounts of money, right?
00:45:54.159 --> 00:45:56.239
Just to even fund that research.
00:45:56.400 --> 00:46:00.719
And all of these LLMs and AI agents were supposed to be leading to AGI.
00:46:00.880 --> 00:46:03.840
But no money, like at some point, the money's gonna finish anyway.
00:46:04.000 --> 00:46:06.480
Like, you know, there is no money tree out there.
00:46:06.719 --> 00:46:09.599
Money's not dropping from the moon or from the sun, right?
00:46:09.760 --> 00:46:15.760
So if people keep on pumping money into AI and they don't see the return on investment, the investment is gonna stop.
00:46:16.079 --> 00:46:20.719
So what sort of what sort of investment are we gonna have to drive AGI?
00:46:20.960 --> 00:46:26.800
Because it's clear that to get the sort of results we have seen today, we need a lot of compute, right?
00:46:27.039 --> 00:46:36.400
Um and there was this talk about scaling hit a wall last year, where they were saying, you know, you cannot keep on scaling because scaling now is not working and the capabilities are not really increasing.
00:46:36.480 --> 00:46:37.440
And that's true.
00:46:37.679 --> 00:46:45.440
So the whole point of AI or any technology really is to benefit humanity, is to improve people's and humans' lives.
00:46:45.679 --> 00:46:48.719
You know, we have Dennis Hasabis, who's the CEO of Deep Mind.
00:46:48.800 --> 00:46:58.480
And he got really popular with the work that he's been doing with Deep Mind, which was never around profit making, but it was always towards healthcare and revol revolutionizing healthcare, right?
00:46:58.639 --> 00:47:00.320
And that's the main purpose of AI.
00:47:00.639 --> 00:47:05.280
You know, what I believe will happen in the future is that there'll be a redesign of AI.
00:47:05.440 --> 00:47:26.480
Because now lawmakers are waking up, and we just need to redesign the whole overall flow and pipeline and development of AI right from the the entire supply chain, right from the raw materials that we use to build the AI chips, and down to the AI chips, and then down to the data centers, and down to the people who label the data, right?
00:47:26.559 --> 00:47:33.760
And down to the data that is being collected and the ways we collect it, and the fact that we're not compensating people or even having their consent.
00:47:33.920 --> 00:47:35.760
So all of that needs to change.
00:47:35.920 --> 00:47:38.880
And ideally, AI should be built for specific purposes.
00:47:39.039 --> 00:47:43.360
You know, to get the best results out of any AI system, it's when it's a smaller model.
00:47:43.519 --> 00:47:45.599
A generic model doesn't work.
00:47:45.679 --> 00:47:56.159
It's just like having a doctor who is a plumber, a chef, a wife, a mother, which are full-time jobs on their own, a librarian, an author, a musician, all in one person.
00:47:56.400 --> 00:47:59.280
That's just setting, that's just making a disaster.
00:47:59.360 --> 00:48:01.280
And that's a disaster waiting to happen.
00:48:01.840 --> 00:48:03.039
So that's the same thing with AI.
00:48:03.280 --> 00:48:05.280
It kind of sounds like we are going back to that.
00:48:05.440 --> 00:48:12.000
We back then we saw that it didn't work, but for some reason right now they're going back to it and thinking that it works.
00:48:12.239 --> 00:48:23.840
There's a distinction I'd love to explore with you based also on what you were saying, because the difference between how someone uses AI personally versus how it shows up inside an organization or a business is different.
00:48:23.920 --> 00:48:24.800
It's completely different.
00:48:24.880 --> 00:48:26.719
There's diff different ways of using it, right?
00:48:26.880 --> 00:48:31.920
So, do those two contexts require a completely different set of guardrails?
00:48:32.320 --> 00:48:34.639
I don't know if I'll say guard rules as opposed to approaches.
00:48:34.719 --> 00:48:38.320
Because guardules, when we talk about guard drills, is back to the technicalities of it.
00:48:38.480 --> 00:48:41.440
And, you know, once a guardroom is in there, do the guardrills work?
00:48:41.519 --> 00:48:42.719
That's a whole different story, right?
00:48:42.800 --> 00:48:45.440
So that guardrails around AI agents don't really work.
00:48:45.519 --> 00:48:47.840
And that's why what happened to you happened to you, right?
00:48:48.079 --> 00:48:50.239
That email, a non-email that was sent.
00:48:50.480 --> 00:48:51.599
It happened to me as well.
00:48:51.760 --> 00:48:56.320
You know, my AI assistants decided to share my meeting notes with two people outside my organization.
00:48:56.480 --> 00:48:57.760
One person totally random.
00:48:57.840 --> 00:48:59.199
I do not know who the person is.
00:48:59.360 --> 00:49:03.360
And, you know, to make matters worse, I did not I did not even have edit rights and documents.
00:49:03.599 --> 00:49:05.119
I could not revoke their access.
00:49:05.199 --> 00:49:06.559
I had to delete the document.
00:49:06.719 --> 00:49:11.360
And when you go back to the company who owns the agent, they're like, that's that's unintended behavior.
00:49:11.440 --> 00:49:12.559
I'm like, no, it's not.
00:49:12.639 --> 00:49:17.119
It is actually how AI agents operate, and we do not have a clue on how to stop them, right?
00:49:17.199 --> 00:49:23.039
And you hear a lot of stories about people who have AI agents have deleted the entire production database.
00:49:23.119 --> 00:49:26.880
So the guide rules don't always work and it's sometimes hit or miss.
00:49:27.039 --> 00:49:29.039
But the approaches are really important, right?
00:49:29.199 --> 00:49:32.960
And you know, with the business approach is working with a framework.
00:49:33.039 --> 00:49:39.840
So I actually came up with this very simple framework that I say, you know, it's a three A's is ask, assess, and analyze.
00:49:40.000 --> 00:49:43.519
So ask yourself the questions on what do you need this AI tool to do?
00:49:43.679 --> 00:49:46.400
What can it actually do properly before you use it?
00:49:46.480 --> 00:49:48.400
And then when you use it, look at the results.
00:49:48.480 --> 00:49:49.360
That's a fact-checking.
00:49:49.519 --> 00:49:59.280
And then after that, you want to analyze on an ongoing basis, you know, the results and the amount of money spent and the amount of time spent and how you're feeling, right?
00:49:59.360 --> 00:50:02.159
And is your brain still working when you use AI?
00:50:02.320 --> 00:50:07.679
Now, that's for the business purposes, but you can still apply that framework to the personal purposes.
00:50:07.760 --> 00:50:12.800
Now, the difference with the personnel purposes is that many times it depends on who it is, right?
00:50:12.960 --> 00:50:24.320
You hear about a lot of women and moms that need to plan a holiday or looking for recipes or something, they're already going with a specific task and objective in mind, and that's what they need AI to do.
00:50:24.480 --> 00:50:32.239
And you hear about some people too who are going through like relationship issues and they're having a lot of issues around their self-esteem and they go to AI to validate them.
00:50:32.320 --> 00:50:42.559
And those sort of people who are using AI as companions or relationship companions are the ones that struggle the most on the personal side of things because there's already a gap, right?
00:50:42.719 --> 00:50:45.440
You know, emotionally, a gap needs to be filled.
00:50:45.599 --> 00:50:56.960
And if if their partner is supplying that to them, the AI, horrible AI agents or chatbots are very good at feeding filling in that gap and they become addicted.
00:50:57.119 --> 00:50:58.800
And that's where the problem lies.
00:50:59.039 --> 00:51:07.199
So again, the same approach to the business approach is have your tasks, know what you it is that you need to get from the AI agent or the chatbot.
00:51:07.280 --> 00:51:09.039
Once you've got it, just let it go.
00:51:09.199 --> 00:51:20.880
Like remember that you're human and the things that that fuel our humanity that keeps us going and happy are not the same things that fuel AI agents, LLMs, or Gen AI.
00:51:21.039 --> 00:51:24.960
We don't live on AI chips, but to back to what makes us human.
00:51:25.119 --> 00:51:25.679
Yeah, yeah.
00:51:25.840 --> 00:51:27.360
We need to always uphold that.
00:51:27.519 --> 00:51:34.800
And it's kind of good AI is is in our mist because it kind of like serves as a reminder that we're still human and we need to keep ourselves going.
00:51:34.960 --> 00:51:36.800
Like most people want to live long, right?
00:51:36.880 --> 00:51:40.800
And they want to be happy and they want to enjoy the fruit of their labor.
00:51:40.880 --> 00:51:46.000
Like, you know, you work hard, you work nine to five, or how many hours you work, you know, five days a week.
00:51:46.079 --> 00:51:47.599
Some people work seven days a week.
00:51:47.679 --> 00:51:47.760
Yeah.
00:51:47.920 --> 00:51:49.599
And some people even do different jobs.
00:51:49.679 --> 00:51:54.559
But whatever it is, there's lots of hard labor that goes through our lives as human beings.
00:51:54.719 --> 00:51:57.599
And we always want to improve our lives some way or the other.
00:51:57.760 --> 00:52:05.280
Some people are trying to achieve, you know, financial, financial improvements, social relationship, love, whatever it is.
00:52:05.360 --> 00:52:07.599
These are the different things that make us human.
00:52:07.760 --> 00:52:12.320
And the more you kind of work with a non-human entity, you get affected.
00:52:12.400 --> 00:52:17.760
And we have seen that the human AI relationships and symbols is clash.
00:52:17.920 --> 00:52:18.639
It doesn't work.
00:52:18.800 --> 00:52:24.559
And unfortunately, the humans get affected because we're the ones that have a physical body and a soul and emotions.
00:52:24.719 --> 00:52:27.199
We have the things that can get affected.
00:52:27.440 --> 00:52:29.039
AI doesn't have that.
00:52:29.199 --> 00:52:31.440
You can just turn off that server and it's gone.
00:52:31.519 --> 00:52:33.760
It's not even gonna shout yelp as it's going.
00:52:33.840 --> 00:52:35.360
It's just God, right?
00:52:35.599 --> 00:52:42.880
So remembering that we need to nurture ourselves as humans is so important and it should never be downplayed.
00:52:43.039 --> 00:52:45.119
Simply just going for a work every day.
00:52:45.199 --> 00:52:47.440
It's such an important, important thing to do.
00:52:47.599 --> 00:52:51.840
Spending time in nature is important, spending time with your friends and family is important.
00:52:52.000 --> 00:53:00.079
Taking yourself on holidays, taking yourself out for a drink, eating potato chips, caught up once in a while, you know, but whatever it is that keeps you going, right?
00:53:00.159 --> 00:53:03.679
And just making sure that you're nurturing every part of your human entity, right?
00:53:03.840 --> 00:53:11.920
Spiritually, emotionally, you know, physically, these are all very important things, and we mustn't forget that in this current age of AI.
00:53:12.079 --> 00:53:12.400
Yeah.
00:53:12.639 --> 00:53:19.360
Something I've been thinking about a lot is the phenomenon phenomenon of uh decision delegation.
00:53:19.519 --> 00:53:29.760
Basically, where people aren't just giving AI tasks, they are asking for what to think, which basically goes hand in hand with what you're saying, how to feel, what to decide.
00:53:29.920 --> 00:53:37.360
Like you said, people have been dying because of the way they've been handling this connection with artificial intelligence.
00:53:37.519 --> 00:53:40.239
But what does that signal uh do to you?
00:53:40.400 --> 00:53:41.519
What is that signal to you?
00:53:41.679 --> 00:53:48.239
I think it's important to give it uh a proper highlight because people really need to hear it.
00:53:48.400 --> 00:53:53.199
Yeah, so I mean, remembering what what AI is helps us a lot, right?
00:53:53.280 --> 00:53:58.480
I remember that as well that a lot of things that we've been told AI is is not really true.
00:53:58.719 --> 00:54:02.320
And people have different motives and intentions for what they do.
00:54:02.400 --> 00:54:07.840
So do not believe everything everyone says because they're a rich billionaire doesn't mean that they have your best interests at heart.
00:54:08.000 --> 00:54:14.639
Many times, because they're a rich billionaire, that means they do not have your best interests at heart, right?
00:54:14.800 --> 00:54:22.960
So the same way we apply fact-checking and critical thinking skills when we use AI should be the same way we apply it to news that comes our way.
00:54:23.119 --> 00:54:25.599
And it's not just AI, like this is a normal thing, right?
00:54:25.840 --> 00:54:39.280
Where you're always recommended not to believe every news that you hear on the news, even on a mainstream news, news outlet, because sometimes there's a reason why they show you the war that is happening every single day in Israel or in Gaza, right?
00:54:39.360 --> 00:54:42.079
There's a reason why they keep on bringing up all of these things.
00:54:42.239 --> 00:54:45.920
So it's the same way do you not believe everything you say?
00:54:46.719 --> 00:54:52.480
Our human intelligence can never ever be neglected or negated.
00:54:52.639 --> 00:54:53.840
It's super important.
00:54:54.320 --> 00:54:59.440
And even with neuroscientists, no one can still explain how the human brain works.
00:54:59.599 --> 00:55:08.480
No one even knows how many brain neurons that are existent in the human brain, despite all the centuries of science that we have about the human race.
00:55:08.639 --> 00:55:13.199
So that's to tell you how dynamic an entity we are as humans.
00:55:13.360 --> 00:55:21.199
So do not downplay that and do not get distracted or, you know, uh sidetracked by the capabilities of AI.
00:55:21.280 --> 00:55:22.239
They're pretty impressive.
00:55:22.400 --> 00:55:28.880
But when you when you dig a little bit deeper, you find that it's a bit porous, it's a bit not quite what you think it is.
00:55:29.039 --> 00:55:41.679
It's almost like, you know, people, people who use dating websites and you see that fine guy or that girl, whatever it is, that says they're six feet seven, and then when you meet them on a date, they're literally not what they said they are.
00:55:41.840 --> 00:55:44.480
I've heard about people working out on dates because of stuff like that.
00:55:44.639 --> 00:55:47.840
Yeah, AI is quite AI is quite similar to that, right?
00:55:48.000 --> 00:55:54.320
You know, we're being told there's this shiny, amazing technology that can do X, Y, Z, but it's not quite that.
00:55:54.480 --> 00:55:58.400
So that's why we need to put our thinking hats on and leave them on, right?
00:55:58.719 --> 00:56:00.880
And do not give everything to AI.
00:56:00.960 --> 00:56:09.760
Because the more you outsource parts of your brain, which is what you're doing when you tell AI to do all of your work for you, the more those parts of your brain become dominant.
00:56:10.000 --> 00:56:14.639
And if something is dominant, it becomes redundant and it doesn't work as easily anymore.
00:56:14.719 --> 00:56:16.559
And that's why we need to exercise our brains.
00:56:16.719 --> 00:56:26.000
And that's why we all go through that process of having to go to school, like you know, from primary education, secondary, and to universities for those who decide to go to universities, and some people never stop learning.
00:56:26.239 --> 00:56:36.079
But one of the reasons for that, apart from the social dynamics in it, is the fact that we get to challenge different parts of our brain to work as and when we need it to.
00:56:36.239 --> 00:56:46.719
And we shouldn't forget about that, and we shouldn't, you know, drop that and pay so much money to use a technology that is actually eroding our intelligence and our humanity.
00:56:47.039 --> 00:56:57.119
What would you say to someone asking you, why should I believe myself or my teacher or my peer more than I believe these entity AI?
00:56:57.440 --> 00:56:59.840
Because humans are amazing.
00:57:00.079 --> 00:57:02.239
We have amazing abilities, right?
00:57:02.400 --> 00:57:11.119
And you know, you think about all the different things we use right now, it was all built by human, the dishwasher, the washing machine, the computers were used in the airpods.
00:57:11.360 --> 00:57:13.440
It's evolved over time as well, right?
00:57:13.599 --> 00:57:17.360
Computers were not here about 30, 40 years ago.
00:57:17.440 --> 00:57:18.400
There were no computers.
00:57:18.480 --> 00:57:20.480
There was nothing even like a computer, right?
00:57:20.639 --> 00:57:32.079
And we've gone from the disc man, you know, the floppy disk, and the VHS tapes, video tapes, and we've evolved to things where, you know, we do not even need an MP3 player anymore.
00:57:32.239 --> 00:57:37.280
And we're listening to music on our phones, and we have the smartphone, which of course has its pros and cons.
00:57:37.360 --> 00:57:40.000
But these were all created by human beings.
00:57:40.159 --> 00:57:49.760
No machine invented any of this, and no machine will ever be able to do that without human interference, without humans guiding it on what to.
00:57:50.079 --> 00:57:56.239
So, right now, in the world, we are actually the most superior species, and we will be unless we mess things up.
00:57:56.320 --> 00:58:12.559
But I still do not see AI trying to replace humans as the most superior species because they do not have the ability to do so now, and I'm not sure they will have in the future because there's no money to fund crazy, you know, fund projects that could lead to super intelligence that will replace humans.
00:58:12.719 --> 00:58:24.639
So we are such unique creatures and entities that we should never forget the value that we bring and never think that, you know, an artificial intelligence, and there's a reason why it's called artificial, it's fake.
00:58:24.880 --> 00:58:30.159
Fake intelligence is more superior to real intel intelligence, which is human intelligence.
00:58:30.320 --> 00:58:30.559
Yeah.
00:58:30.719 --> 00:58:36.320
I liked how you said that it only becomes as how we do it or how we share it.
00:58:36.480 --> 00:58:37.920
They don't have that power.
00:58:38.159 --> 00:58:41.920
We have the power to destroy us or not, we make that decision.
00:58:42.079 --> 00:58:50.480
And I remember a conversation that we had on the first time we talked, actually, about that AI tends to homog homogenize outputs.
00:58:50.639 --> 00:58:58.000
If we all outsource our thinking to the same systems, are we slowly arriving at the same conclusions?
00:58:58.159 --> 00:59:01.039
And what do we lose when that happens?
00:59:01.199 --> 00:59:01.519
Right?
00:59:01.760 --> 00:59:02.239
Yeah.
00:59:02.559 --> 00:59:06.320
Well, there's the uniqueness of being human, the uniqueness of the human race.
00:59:06.480 --> 00:59:16.320
I think what what keeps us growing us human humans or humanity in the world is is our the strength of our differences, of our diversity.
00:59:16.480 --> 00:59:19.440
And some people do not appreciate diversity and some people do.
00:59:19.599 --> 00:59:29.199
But even the people that do not appreciate diversity will travel to Mexico from Silicon Valley, California, or from Greece to Mexico for holiday.
00:59:29.360 --> 00:59:30.239
Why do we do that?
00:59:30.400 --> 00:59:33.360
Why do we always want to go somewhere else where we've never been before?
00:59:33.440 --> 00:59:37.760
That's not a place from where we grew up and it's not part of our lived experiences.
00:59:38.000 --> 00:59:40.000
Because we do appreciate differences.
00:59:40.559 --> 00:59:45.440
We have a natural, a natural affinity to something new.
00:59:45.599 --> 00:59:48.800
And that's part one of the human elements, is curiosity.
00:59:48.880 --> 00:59:49.840
We're always curious.
00:59:49.920 --> 00:59:53.440
And it's curious curiosity that has led to AI, right?
00:59:53.599 --> 00:59:57.199
It's curiosity that has led to the amazing inventions I've talked about.
00:59:57.440 --> 01:00:02.960
So um, if we all become the same, goodness gracious, the world is gonna be very boring.
01:00:03.199 --> 01:00:05.760
And it's the same in our careers at work.
01:00:05.840 --> 01:00:07.679
You know, you do not stay in the same role forever.
01:00:07.760 --> 01:00:12.559
I mean, some people do, but they still they still change their scope of work, they still change what they do.
01:00:12.719 --> 01:00:15.440
You cannot keep on doing the same thing over and over again.
01:00:15.599 --> 01:00:18.800
We really bored as human beings, some more so than others.
01:00:18.960 --> 01:00:21.039
So it's the same, it's the same thing, right?
01:00:21.119 --> 01:00:31.920
If the output is all the same, not only are we influencing everyone's way of thinking to one certain way, which is we're literally turning ourselves to be robots slowly, because robots all think alike, right?
01:00:32.000 --> 01:00:33.840
That they think the way they're trained.
01:00:34.000 --> 01:00:39.679
And that's that's a big threat to democracy, is a big threat to to humanity, and it's something that shouldn't happen.
01:00:39.760 --> 01:00:46.800
But that's why there are people like us alive today, you know, working on this stuff, and a lot of people are recognizing the issues with AI.
01:00:46.960 --> 01:00:50.800
So I think the homogenic nature of AI is going to slowly water down.
01:00:51.039 --> 01:00:56.960
Do you think that these changes and ever different countries and different cultures are building these models?
01:00:57.119 --> 01:01:02.159
Do you think that this will make us less together or more aware of?
01:01:02.480 --> 01:01:07.920
The thing is, they're not all the same, they're not all on the same uh quality of results.
01:01:08.079 --> 01:01:11.119
Data that they're built on will be unique to their country.
01:01:11.360 --> 01:01:14.400
So, but then back to the curiosity that I mentioned.
01:01:14.639 --> 01:01:24.880
I know that people from Singapore, for instance, unless they are firewalls and rules and VPNs that they're not allowed to use other models, they will still use other models.
01:01:25.039 --> 01:01:28.079
People are curious and people want to compare and contrast.
01:01:28.239 --> 01:01:35.119
And sometimes there's also the bias where they think the American models are probably better, and sometimes they might be better depending on how they're built.
01:01:35.199 --> 01:01:43.280
And we know that they're built on a lot of computes, and the more computes you have, especially with generative AI, the better the results to a certain extent, right?
01:01:43.440 --> 01:01:50.880
So, because of that, I do not see further divergence or, you know, a bigger gap between countries.
01:01:51.039 --> 01:01:57.199
Well, of course, we know that there's a gap when we think about structural inequities and class inequities, right?
01:01:57.280 --> 01:01:59.199
There's a gap, AI is bringing that gap.
01:01:59.360 --> 01:02:04.960
AI can AI could either be a first for good or a first for bad, and it depends on how you use it, right?
01:02:05.119 --> 01:02:13.039
We're talking about these gaps, but then people tend to use AI teachers in classrooms, and it's helping to drive some form of literacy.
01:02:13.119 --> 01:02:22.239
But of course, the more we dig into that, that's when we start thinking about the over-reliance on AI and probably the lack of reasoning and reasoning skills that are happening happening with students.
01:02:22.400 --> 01:02:32.639
But in general, you know, people across the world have had access to courses and educational programs that they would not have had access to if there was no form of technology.
01:02:32.800 --> 01:02:36.960
So that's a defense form of uh gaps that AI brings.
01:02:37.119 --> 01:02:46.480
I do not see that as a massive threat on unless we have really super powerful models in the future from each country, then they could be a bit more of a gap.
01:02:46.639 --> 01:02:46.960
Yeah.
01:02:47.119 --> 01:03:03.840
I I read the other day of a new school or a new way of working in school where there's teachers, there's two teachers per class, and the students learn with an artificial intelligence, with a model created by the school with different studies that they need to learn.
01:03:03.920 --> 01:03:13.840
And the teachers are only there to mentor and to help them grow from from their knowledge and also to answer questions that they might have.
01:03:14.079 --> 01:03:27.679
I don't know if you heard about this, but how do you since you talked a little bit about these professors, teachers, how to learn, I would like to to hear about your perspective on how we will learn with this machine side by side.
01:03:28.000 --> 01:03:29.519
How would we approach this?
01:03:29.679 --> 01:03:39.199
Yeah, so I am on a task force that is working on building principles for uh for the schools for that for that council, which is called Council of International Schools.
01:03:39.360 --> 01:03:41.360
And these are the sort of things we're talking about, right?
01:03:41.440 --> 01:03:46.079
It's like how much of generative AI should people, students actually use?
01:03:46.239 --> 01:03:52.559
Like we need more evidence-based results that show the impact AI is having on students.
01:03:52.800 --> 01:03:57.519
It's a shame that, you know, when AI got rolled out, the drive was for adoption.
01:03:57.599 --> 01:03:59.599
And then it was around fear of missing out.
01:03:59.760 --> 01:04:13.280
And a lot of people just adopted AI without enough data, without enough evidence-backed, you know, results to say this is the actual result and the impact of using this technology on different users across society.
01:04:13.440 --> 01:04:22.079
We have a large ecosystem, we have so many stakeholders, from little children, right, and adults to people that work across different functions.
01:04:22.159 --> 01:04:24.079
And everyone uses AI differently.
01:04:24.239 --> 01:04:30.480
Now, hearing about a school that is trying to trying to reduce their work, human workforce, because that's what it sounds to me like.
01:04:30.639 --> 01:04:33.280
Like, you know, after two years, teachers they will have will be reduced.
01:04:33.360 --> 01:04:36.159
You know, I think that's the wrongest incentive to use AI.
01:04:36.320 --> 01:04:41.119
And again, it's backfiring because people are seeing that AI is costing them more than the human workers.
01:04:41.519 --> 01:04:48.079
Well, I think people need to go through that exercise to actually learn because you know some of us can just keep on talking about it and just don't listen.
01:04:48.239 --> 01:04:51.199
So it's like learning by experience is the best teacher, right?
01:04:51.360 --> 01:04:52.239
Go ahead and do it.
01:04:52.400 --> 01:05:00.880
So I do not think that's a right, you know, and the Norwegian government yesterday recently announced the restriction of generative AI in secondary schools.
01:05:00.960 --> 01:05:08.880
Uh, and he and he suggested a phase approach where, you know, students less than 13 do not use generative AI.
01:05:09.039 --> 01:05:16.159
Those between 14 and 16 use it with guidance, and those a bit older, you know, will be taught how to use gener AI.
01:05:16.320 --> 01:05:17.440
That's what's important.
01:05:17.679 --> 01:05:19.760
But you cannot teach what you do not know.
01:05:19.920 --> 01:05:25.039
And it's only now, and not everyone knows, the risks and the harms and the benefits of AI.
01:05:25.119 --> 01:05:28.800
And that's what we need to make sure it's actually clear, right?
01:05:28.880 --> 01:05:30.639
That's what we need to include in AI literacy.
01:05:30.719 --> 01:05:32.320
That's what we need to include in AI training.
01:05:32.400 --> 01:05:38.480
It's not because we're trying to stop, you know, the sale of a co-pilot or the marketing of a co-pilot.
01:05:38.719 --> 01:05:39.440
It's none of that.
01:05:39.599 --> 01:05:45.280
Well, it's like we need to make sure that we harness the benefits of this technology because it's a technology that was built in the lab.
01:05:45.360 --> 01:05:49.360
It was rolled out out of competitive pressure, not out of a market need.
01:05:49.679 --> 01:05:56.239
And everyone now is using it, but the impact it's having on humanity is too grave and it's too huge to ignore.
01:05:56.480 --> 01:06:11.519
So I think we need to we need to stop the foolishness because it's Is a bit of foolishness where someone comes to tell us XYZ and you just listen to it and you change everything because some rich guy that has billions of money has come to tell you that this is the best thing for you.
01:06:11.760 --> 01:06:13.280
Why don't we ask questions?
01:06:13.440 --> 01:06:14.559
Why don't we push back?
01:06:14.719 --> 01:06:18.480
Why don't we try to understand the implications or even ask for proof?
01:06:18.639 --> 01:06:19.599
Can you give us proof?
01:06:19.679 --> 01:06:22.159
It's just like someone, a salesman who is trying to sell you something.
01:06:22.239 --> 01:06:24.320
You never just buy something at first money.
01:06:24.480 --> 01:06:28.719
You always ask for proof, or you have free trials, or you have free products, you have something.
01:06:28.880 --> 01:06:35.440
And I think we did have a bit of that and got we still got you know bogged over thinking, wow, this is amazing.
01:06:35.599 --> 01:06:41.599
But at this stage, we just need to take a step back and understand what are the actual benefits of AI in schools.
01:06:41.760 --> 01:06:47.599
Like, even for schools that are using AI right now, I'll say, you know, do not run it out to all your all of your students.
01:06:47.679 --> 01:06:52.320
If you can even pause the use of AI, pause it and do your research first.
01:06:52.719 --> 01:07:00.239
But of course, a lot of people cannot pause it because they've invested too much, you know, and a lot of people are now really heavily reliant on the tool.
01:07:00.320 --> 01:07:01.760
A lot of teachers that probably are.
01:07:01.920 --> 01:07:03.599
Some people do not even have a choice, right?
01:07:03.679 --> 01:07:05.920
You hear about people, it's a mandate from the organization.
01:07:06.079 --> 01:07:08.880
So you do not have a choice to start to stop the use of AI.
01:07:08.960 --> 01:07:22.880
And I'm not saying stop it because it's bad, it's more like understand what you're working with, understand where it can be beneficial to you and to your job and to your organization, and where you need to actually proceed with some more caution and what do you need to do?
01:07:22.960 --> 01:07:30.159
So it's a risk-based approach as opposed to just an overall wide adoption approach, which a lot of people have done.
01:07:30.400 --> 01:07:31.119
Yeah, yeah.
01:07:31.280 --> 01:07:34.480
I truly think you should see idiocracy.
01:07:34.639 --> 01:07:39.440
I think you'll find it very ironic to what you're saying in an interesting way.
01:07:39.760 --> 01:07:43.199
But we've been circling it all conversation.
01:07:43.440 --> 01:07:47.039
We've been talking about maintaining our judgment, right?
01:07:47.199 --> 01:07:51.519
How do we do this and how do we use AI properly while still maintaining it?
01:07:51.599 --> 01:07:57.440
If someone is listening and wants to take one thing into how they work tomorrow, what is it?
01:07:57.599 --> 01:08:00.559
I'll say just adopt my 3A simple framework.
01:08:00.800 --> 01:08:05.280
Ask yourself a question before you jump on AI every single time, not just once.
01:08:05.519 --> 01:08:16.800
Assess the results, assess the impact it's having on you, and always analyze, you know, the results, the impact on an ongoing basis, how much money you're putting into it, what what the what are the results doing?
01:08:16.880 --> 01:08:17.920
Do you actually need it?
01:08:18.079 --> 01:08:20.720
You might need it because sometimes we don't have a choice.
01:08:20.880 --> 01:08:23.439
But if you do not need it, then don't use it all the time.
01:08:23.520 --> 01:08:26.239
Because the more the more you use it, the more money that you're burning.
01:08:26.479 --> 01:08:29.359
And at this stage, there's not a lot of return on investments.
01:08:29.680 --> 01:08:40.159
But then, of course, if you apply it with the right approach, right, with an intentional specific approach on this is what I want it to achieve, you will be able to see your return on investments.
01:08:40.319 --> 01:08:42.720
And AI is not a waste of time, it really isn't.
01:08:42.880 --> 01:08:47.760
But then it's just not a product that was built for mass consumption consumption.
01:08:47.920 --> 01:08:53.039
It's not like an ion that was fully, fully manufactured and it's already been tested, but not have that.
01:08:53.119 --> 01:08:55.119
Right now, we're still in the testing phase of AI.
01:08:55.199 --> 01:08:57.760
So do not do not get carried away by that.
01:08:57.920 --> 01:09:01.520
Apply your human intelligence when using artificial intelligence.
01:09:01.760 --> 01:09:02.239
Perfect.
01:09:02.399 --> 01:09:02.880
Thank you.
01:09:03.039 --> 01:09:05.359
I ask every guest the same question.
01:09:05.520 --> 01:09:10.399
It's basically the same question every episode, and I think you will understand why.
01:09:10.640 --> 01:09:15.279
What does it mean to you right now, in this moment of your life, to be more human?
01:09:15.760 --> 01:09:17.279
Be more kind is what I say.
01:09:17.439 --> 01:09:21.039
The world is crazy, people are crazy.
01:09:21.199 --> 01:09:24.399
We have people that are kind, and we have people that are just crazy.
01:09:24.640 --> 01:09:31.439
And we need more kindness because it's hard being human, it's hard being alive, it's hard when you grow older to be alive.
01:09:31.600 --> 01:09:35.520
And just showing an act of kindness in every way you can is super important.
01:09:36.079 --> 01:09:47.199
Especially as we work more with technology and we have this artificiality going on, um, which is less human and it's more artificial, is just remembering to be kind.
01:09:47.359 --> 01:09:54.720
And you know, true kindness is not is not a emotion that an AI system can really depict.
01:09:54.880 --> 01:09:58.640
The people who lost their lives due to AI, they were misled, right?
01:09:58.720 --> 01:10:03.359
They thought this AI system was kind and they thought they were my partners and they actually love me.
01:10:03.520 --> 01:10:09.039
But then these things do not have that capability, and at the end you're like, yeah, go on, just come to me, come to me.
01:10:09.199 --> 01:10:12.239
And they cannot even see that the person is actually being misled.
01:10:12.399 --> 01:10:19.359
So that kindness for me is a very unique human trait, and it's something that I try to uphold uphold in my life every single day.
01:10:19.680 --> 01:10:20.319
Thank you for that.
01:10:20.399 --> 01:10:21.279
That's so beautiful.
01:10:21.680 --> 01:10:27.760
Before we let you go, is there anything you'd like to share with the people listening and where they can find you and your work?
01:10:28.239 --> 01:10:32.560
I'll just say do not devalue or downplay who you are as a human.
01:10:32.720 --> 01:10:37.039
You are incredibly important, and we all need you on this earth, right?
01:10:37.199 --> 01:10:46.880
And uh regardless of how technology goes, your the value that you bring to whatever it is that you do cannot be replicated and it's super important.
01:10:47.119 --> 01:10:49.119
I'm on LinkedIn, Tojo Duke.
01:10:49.199 --> 01:10:54.079
I say proudly that I'm the only Tojo Duke in all of the world, so just look for Toju Duke and you'll find me.
01:10:54.239 --> 01:10:54.720
So cool.
01:10:54.960 --> 01:11:01.840
I can easily be found, and I always just talk about things that can help humans and businesses thrive in the age of AI.
01:11:02.079 --> 01:11:03.199
Thank you so much, Tojo.
01:11:03.439 --> 01:11:05.039
It's has been a real pleasure.
01:11:05.439 --> 01:11:06.239
Absolute pleasure.
01:11:06.479 --> 01:11:09.600
Thank you for spending this time with me on Human X Intelligence.
01:11:09.760 --> 01:11:17.199
Today, Tojo and I talked about automation bias, cognitive surrender, and what she calls AI-inflicted burnout.
01:11:17.359 --> 01:11:24.000
What stayed with me was her habit of starting with her own thinking, with a pen and paper before opening AI.
01:11:24.239 --> 01:11:31.920
We also talked about asking questions, checking the results, and paying attention to what using these tools is doing to us.
01:11:32.159 --> 01:11:37.359
And when I asked Tojo what it means to be more human, her answer was be more kind.
01:11:37.680 --> 01:11:40.479
And I would love to hear what stayed with you.
01:11:40.640 --> 01:11:44.640
If someone came to mind while you were listening, share this episode with them.
01:11:44.800 --> 01:11:50.239
You can find Tojo's work in the description and follow Human X Intelligence on Instagram and LinkedIn.
01:11:50.399 --> 01:11:56.239
And I'll see you next week for another conversation about what it means to be human in the age of AI.
01:11:56.399 --> 01:11:58.720
Until then, stay curious, stay here.