[00:00:11] Hala Taha: Young and profiters. We have been talking about AI a lot on the show, and that's because we know how important it is to have an in depth understanding of the technology and how it is rapidly shaping our world.

 And we can't talk about AI as merely a thing of the future anymore. It's developing exponentially every day, and whether we're aware of it or not, we play an active role in how it affects us and our careers or businesses. So today we are dusting off a not so old episode from 2023. It's only like maybe eight months old, but it is so good.

And I want to make sure that anybody who didn't hear it yet hears it. Because this episode with Mo Gowdat blew my mind. And it blew a lot of people's minds, which is why it went massively viral on YouTube. And it was one of my most popular episodes last year, if not the most popular episode. And so I really wanted to resurface it because I keep referencing this episode and other AI episodes that I have.

And I just want everybody to make sure that they hear this conversation. It was really eye opening. So if you don't know Mo Gowdat, he is a guru in the tech world. He worked at IBM and then Microsoft, then he joined Google. And at first he was the Vice President of Emerging Markets at Google and he started half of Google's businesses globally.

And then he became the Chief Business Officer of Google X. X is Google's innovative lab where they work on big, bold ideas to solve the world's greatest problems. And in this Yap Classic, Mo is going to talk about one of the most intriguing AI experiments they ever invested in, which included robot arms that were taught to pick up toys and put them away.

And that might sound super simple, but it actually completely changed the way that Mo perceived AI. And it scared him so much, and he was so against what Google was doing with AI that he ended up quitting because he didn't want to be a part of it. Now that's not to say that Mo is not optimistic about AI.

He believes there's a bright future ahead of us, but only if we take the dangers seriously of AI and we develop AI responsibly. So there's so much to unpack here. There's so much to learn in this episode. This was one of my favorite episodes of last year. You guys are going to love it. So without further ado, here's my conversation with the incredible Mo Gowdat. 

Mo, welcome to Young and Profiting Podcast. 

[00:02:46] Mo Gawdat: Thank you. Thanks for having me. Uh, it's been a while in the making, but absolutely worth the wait. 

[00:02:52] Hala Taha: Can you talk to us about your journey at a very high level, the highlights that got you in the C suite at Google X eventually?

[00:03:00] Mo Gawdat: At the height of my professional career, if you want my corporate career, I was the chief business officer of Google X. And. Of course, I worked my butt off to get there, but there was an element of luck in the process. I met the exact right people at the exact right time. It was one of those events where the Google X team was presenting some of their confidential stuff.

And I showed up and I said, at the time I was vice president of emerging markets for Google, I had started half of Google's businesses globally. More than a hundred and three languages, if I remember correctly. And so I was quite well known in the company. If you want, I had a reasonable impact that I have to say, I'm very grateful that life gave me the opportunity to provide.

And then with Google X, I basically, at the time, Google still had the idea of the 20 percent time, so I liked their projects and I said, I'm going to give you my 20%. And they said, but we haven't asked for it. And I said, yeah, that's not your choice. And I showed up basically the first day I showed up, I bump into Sergei, our co founder, and I worked closely with Sergei for many years.

And he says, what are you doing here? And I was like, I'm very excited about your work and ended up, he said, Oh no, don't leave basically stay. And I was chief business officer for five years where I think Google X is misunderstood because. We never really launched a product under X if you want. So self driving cars is under Waymo.

Google brain is integrated into Google and so on, but most of the very spooky innovation, if you want the very, very out there innovation, including all of robotics and a big chunk of AI. was at X and it was a big part of what I did. 

[00:04:46] Hala Taha: And so diving right into AI, you were actually part of the labs that initially created AI.

So can you talk to us about the story of the yellow ball and how that really changed your perspective about AI? 

[00:05:01] Mo Gawdat: AI has been around a lot longer than people think. When we started self driving cars back in 2008, that was basically with a belief That cars can develop intelligence that is as intelligent as a driver and accordingly able to drive a car.

And since then, I mean, by 2008, I think in my personal memories, I think 2008 was really the year when we knew that we cracked the code. Early 2009, Google published a paper that's known as the cat paper. That white paper basically described how we asked an artificially intelligent machine to look at YouTube videos without prompting it for what to look for.

And then it eventually came back and said, I found something. And we said, show us. And it turns out that it found a cat, not just one cat, but really what cat ness is all about, you know, that very entitled, cuddly, furry character. Basically it could find every cat on YouTube. And that was really the very first glimpse between that and the work that DeepMind was doing on playing Atari games, where machines started to show real intelligence.

We then started to integrate that in a lot of things, you know, self driving cars is probably the most publicly known example, but one of the projects that we worked on was, which is not the only, you know, Google X was not the only one working on it, but we wanted to teach grippers robotic arms. Basically, we wanted to teach them how to.

pick objects that they're not programmed to pick. And it's a very, very sophisticated task because we do it so easily as humans. You don't remember, but if your parents would remember when you were a child and before you learned how to grip, you kept going on trial and error, you would try to grip something and then it falls and then you try again and so on.

And basically we said, maybe we can teach the machines the same way. We built a farm of those grippers, put boxes of items in front of them. A funny programmer basically chose children's toys and you could see them try to pick those items and basically fail over and over. It's a very sophisticated mathematical problem.

And so they would fail. They would show the arm to the camera and the camera would know that this algorithm, this pathway didn't register, didn't pick the item. Until I think it was several weeks in and, you know, it was a significant investment because robotic arms were not cheap at the time I passed by that farm very frequently on my way to my desk and on a Friday evening, finally, one of those arms, you know, I can see it goes down, picks one item, which was a yellow softball again, mathematically very complex to a grip and it shows it to the camera.

Yeah. And so jokingly, I pass by the team that's running this experiment and I say, okay, well done all of those millions of dollars for one yellow ball. Okay, and they smiled and then, you know, sort of nodded their heads and on Monday morning, as I went to work, every arm was picking the yellow ball. A couple of weeks later, every arm was picking everything.

And I think that's something that most people don't recognize about AI is that the speed, once you found the very first pattern, the speed at which AI starts to develop is just mind blowing. Also, I think. Most people don't realize that they learn exactly like my children learned to grip. That's the whole idea.

So they really do develop intelligence that comparable, uh, now probably even more advanced than human intelligence. 

[00:08:42] Hala Taha: In that moment when you saw those machines gripping toys and doing it more efficiently and with intelligence, were you alarmed or were you excited? 

[00:08:55] Mo Gawdat: I've been excited about AI since I had a Sinclair, believe it or not.

So I started coding at a very, very young age on computers. Young and profitable probably have never touched in their life. So, so, uh, you know, and every one of us geeks wanted to code an intelligent machine. We all attempted and we all simulated and we all even pretended sometimes. But then it was the year 2000, truly, where deep learning was starting to develop.

And we sort of found the breakthrough. We found how to give machines intelligence. And allow me to stop for a second here, because there is a huge difference between the way we programmed machines before deep learning and after deep learning. Before deep learning, when I programmed the machine, as intelligent as it looked, I solved the problem first using my own intelligence and then sort of gave the machine the cheat in terms of how to solve it itself.

I wrote the algorithm or I wrote the process step by step and basically coded the machine to do it. When deep learning started to happen. What we did was we didn't tell the machine how to solve the problem. We told the machine how to develop the intelligence needed to find a solution to the problem.

This is very, very different. And as a matter of fact, most of the time we don't even recognize how the machine finds a cat. We don't fully understand how Bard, uh, Google's Bard understood how to speak Bengali, right? We don't really know those emerging properties or even the tasks we give them themselves.

So your question was, was I excited? I promise you the day I met Demis, who was the CEO of DeepMind when we acquired DeepMind, It was really, to me, like meeting a rock star. I was fanatic about what he was doing. I still am a fan of him and his ethics, an amazing human being. But at the time, for a geek, understand this, AI was the ultimate joy and glory.

This was it. We were creating intelligence and for a programmer that was mind blowing. And remember, huh? Every time we saw the machines develop, we got more excited, believe it or not, because we wanted what was good for the world. Intelligence in itself, there is nothing inherently wrong with intelligence.

It was when I saw the yellow ball, I think that something dropped. I could see it so clearly because for the first time ever, right? I realized that those machines one are developing way faster than us. And so accordingly, the predictions of people like Ray Kurzweil and others of a moment of singularity where they're going to bypass our intelligence became very, very real.

In my mind, I could see that this is going to happen, but I also could see that we, the moment they became intelligent had very little influence on them. Okay. And accordingly, I started to imagine a world where humanity is no longer the top of the food chain. Humanity is no longer the smartest being on the planet.

And then comes the apes. We are going to be the apes. Do you understand that? Yeah. And I think that completely made sense to me that this needed a lot more consideration rather than the, you know, the excited geekiness of building it. We needed to understand why and how are we building it and what is a future where it becomes in charge.

[00:12:35] Hala Taha: There's so much to unpack here. This is why I was like, I need to spend the full hour on this topic because there's just so much to unpack. Let's talk about the label of artificial in artificial intelligence. Is intelligence artificial at all or is AI? Oh, yeah. Talk to us about that. 

[00:12:52] Mo Gawdat: Not in the slightest.

If there is any artificial side To the machines is that they are silicon based as a matter of fact, most of the ones who worked on deep tech, not the stuff that you see in the in the interfaces. We almost mapped their brains to the way our neural networks as humans work. You know, humans in the early development of AI, you know what neuroplasticity is.

Humans basically We develop our intelligence and our ability to do anything really by repeating a task in a specific way. And, and they say neurons that fire together, wire together. So. If you tap your finger over and over and over your brain sort of takes that neural network that taps your finger and makes it stronger and stronger and stronger.

Just like going to the gym and the early years of developing AI. We were doing exactly that. We were literally pruning the software or the algorithms that were not it. effectively delivering the task we want, literally killing them, erasing them, and keeping the ones that were capable of getting closer to the answer we wanted and then strengthening them.

So we were sort of like doubling down on them, wiring them together. And the way the machines work today is very, very similar to that. It's a bunch of patterns that are created in hundreds of millions, sometimes billions and trillions of neurons, not yet trillions, but you know, lots of nodes of patterns that the machine would recognize so that it basically can make something look intelligent or can behave in a way that is analogous to intelligence.

Now, is it artificial? Well, I think if you ask the machines, they will think of our carbon based intelligence as artificial. The only difference really is We are carbon based and analog. They are, I don't think we're analog. I think we're somewhere in between and they are digital and silicon based not for long.

We don't know what they're going to be based on in the future, but also I think their clock speed is very different than human clock speed. So they have. an enormous capability of learning very, very quickly of crunching a massive amount of data that no single human can achieve. They have the capability of keeping so much in their memory.

They are aware and informed of everything all the time. They are connected to each other. So they AGI becomes a reality benefit from each other's intelligence. And in a very simple way. I think the race to intelligence is one. Today, there are estimates that chat GPT is at an IQ of 155. Einstein, I think was 160 or 190 doesn't really matter, but most humans are 122.

Some are less than that, maybe 110 and so on. You know, the dumbest human is 70. So you can easily see that there is an AI today from an intelligence point of view on the task assigned to it. Remember, we're still in the artificial special intelligence stage. You want one task assigned to every AI. In the task assigned to it, it's by far more intelligent than humans, nothing artificial at all about that.

It develops its own intelligence, it evolves, it has agency, it has decision making abilities, it has emotions, I tend to believe. And it is in a very interesting way. almost sentient, if you think about it, which is an argument that a lot of people don't agree with because we don't really define sentient on a human level very well, but they definitely simulate being sentient very well.

[00:16:45] Hala Taha: What you're saying is really incredible and mind blowing. I know that for humans, we don't understand how consciousness works, right? Nobody can say, you're conscious because of this. And you mentioned before that we don't understand how intelligence really happens. We know how to create intelligence, but we don't actually know how the intelligence works.

It just sort of takes off on its own, which can be really scary. So talk to us about why you think AI should be considered living or sentient. 

[00:17:15] Mo Gawdat: I think the definition of sentient needs to be agreed. Is a tree sentient? Is a pebble sentient? Is the planet Earth sentient? You know, we could have many arguments.

Now, if you think of being sentient as it is born at a point in time and it dies at a point in time, or at least it has the threat of dying at a point in time, then AI is born at a point in time and it has the threat of dying at a point in time. If you think of sentient as the ability to sense the world around you, well, yes, of course, AI is capable of assessing the world around it.

If you think of sentient as the ability to affect the world around you, then yes, it can, right? If you take a tree, for example, a tree grows, it reproduces. It is in a way, interestingly aware of the seasons and aware of the environment around it, and it responds to it. So a tree will not shed its leaves on the 21st of October, specifically, it will shed its leaves when the weather alerts it to do that.

If you consider a tree sentient in that case, then, AI is surely sentient if you consider that a gorilla is incredibly interested in survival and accordingly would do what it takes to survive. Then AI is sentient in the sense that once assigned a task, it'll attempt to survive to. Make the task happen, basically.

[00:18:45] Hala Taha: It's so interesting, and I know that a lot of people who think of AI think of it as a machine that they can turn off if things get crazy, just tell it what to do. Can you talk about how AI can have agency and free will? 

[00:19:01] Mo Gawdat: Oh my god, I can give you endless examples. If you're not informed of AI today, it is a bit like a hurricane approaching your city or village, and you're sitting at a cafe saying, I'm not interested.

Okay, this is the biggest event happening in today's world. And the reason for that is that there are tremendous benefits that can come from having artificial intelligence in our lives. And if you miss out on that train. You're not going to have the skills to compete in a world that is changing very rapidly.

That's on one side. On the other side, there are very, very significant threats. And those threats come in two levels. The news media wants to always talk about the Terminator scenario, where it's an existential risk to humanity in 10, 15, 20 years time. I believe that there is a probability of that happening, but I believe that there are many more important, more immediate threats that need to be looked at today, things that are already happening and that we need to become aware of things like concentration of power, things that are like the end of truth, things like The jobs and the redesign of the fabric of society as a result of the disappearance of many jobs and so on.

So we'll come to all of those. I think we need to cover both sides of the immediate risk and the existential risk. But your question was, how can AI affect me today? Let me give you a very simple example. There is nothing that entered your head today that was not dictated to you by a machine. We ignore that fact when we swipe on Instagram or when we are on TikTok or when we're looking at the news media or when, you know, we're searching and getting a result from Google, but every single one of those is a machine that is telling you.

In reality, what it is that you should know now, think about the following today in the morning, I got a statistic that basically is quite interesting. A study by Stanford University that said that brunettes are on average taller than blondes, right? I didn't actually, but does it make any difference once I told you that piece of information?

You know, once I tell you a piece of information, I have affected your mind forever. So you can either trust me, and now you're going to look at brunettes and blondes differently for the rest of your life, You can mistrust me and then you're going to spend a little bit of time to try and verify the truth.

And in the back of your mind, that bit of information is going to be engraved. Maybe for the future, you might dedicate yourself to a research that proves me wrong. You may actually become fanatic. You may start posting about it on the internet. You may spend the rest of your life trying to defend this lie or trying to disprove this lie and show the truth.

Just by, by showing you one bit of information. Now, every bit of information you have seen since you woke up today is dictated by a machine. Now you have Noah Harari basically says they have hacked the operating system of humanity. So if I can hack into your brain Hala and tell you something that affects you for the rest of your life, whether positively or negatively, whether true or false, then I've already managed to affect you.

Interestingly, most of those machines that you've dealt with are programmed for one simple task, which is to manipulate you. Every one of those social media machines, for example, are out there with one objective, which is to manipulate your behavior to their benefit. And they're becoming really good at it.

They're becoming so good at it as a matter of fact, that most of the time we don't even realize that we have been brainwashed over and over and over by the capability of those machines. So here's the interesting bit. I told you in the immediate risks that are coming up, I believe they have started already, and I think they will start to become quite significant over the next year or two.

And we will see. My personal view, what I call patient zero is the end of the truth in the U. S. elections. So the reality of the matter is that with deep fakes, with the ability to manipulate information and data, with the ability to create by next year, you have to be aware that a reel on Instagram can be created with no human in front of the camera very, very easily.

Technologies like stability. ai Stable diffusion, for example, can now generate realistic human like images in less than a tenth of a second, and a video is ten frames per second, so the next stage is clearly going to be video. There are multiple videos that have been created that you couldn't distinguish the quality of from an actual iPhone video of you.

Think of face filters and how this is affecting our perception of real beauty. Think of information and statistics. Using chat GPT affecting the children's way of doing their homework. We are completely redesigned as a society and we're not even talking about it. This is how far this has gone.

[00:24:36] Hala Taha: It is insane. And I definitely want to talk about those risks that you were talking about. Immediate risk, job risks, existential risk down the line years later. Talk to us about the fact that AI can learn on its own, it can learn languages on its own, it can beat chess players and come up with moves that we've never taught it before, because a lot of people think about AI as something that just collects information and spits out information, but it can actually learn new things that humans don't even know.

So talk to us about that. 

[00:25:12] Mo Gawdat: Let me give you a concrete example. There is a strategy game known as go go is one of the most complex strategy games on the planet requires a very deep understanding of planning and crunching a lot of numbers and mathematics and so on, very popular in Asia and in our assessment, go was the ultimate task, you know, like we had the touring test for a I pretending to be a human and you're you're not being able to figure out if it isn't go was sort of like that other milestone if I wins in go.

Then AI is now the top gamer on the planet now. It was several, five years ago. I believe that 10 years ahead of any estimate that AlphaGo again, DeepMind basically became the world champion in Go. And AlphaGo had three versions to it. Version number one took a few months to develop. Basically, we asked it to watch YouTube videos of people playing Go.

And from that, it played against the second champion in the world. So the runner up and it won five to one or five to two, but it basically won. And that basically made AlphaGo number two in the world. And then we developed something called AlphaGo Master and AlphaGo Master played against Lee, the world champion, and won.

That was around a few months later. And then we developed another code that was called AlphaGo Zero. And AlphaGo Zero basically learned the game by playing against itself. So it never saw a human ever playing Go. It just played against itself. So it would be the two opponents and through the patterns of the game randomly, it would learn what wins and what loses.

AlphaGo zero within three days, three days, one against AlphaGo, the original within 21 days, one against AlphaGo master and became the world champion, a thousand games to zero within 21 days. Now, when you understand that level of strategy, when Lee, the world champion was playing against AlphaGo master, there is something that you can Google that's known as move 37 and move 37.

Was that machine coming up with a move that is completely unlike anything humans understand to the point that the world champion said, I don't know what this is doing. I need a 15 minutes break to understand it was a move of ingenuity of intuition of creativity of very deep strategy of very, very deep mathematical planning.

And we never taught AlphaGoMaster to do that. We never taught the original games of Atari, DeepMind, to find the cornerstone in the breakout game, if you remember those Atari games. So it would find the cornerstone, throw the ball in there, so that it hits the ball from the top. All of those things we don't teach the machines how to learn.

And we call those emerging properties and emerging properties are basically things that the machine learns on its own without us actually telling it at all to learn it. One of the famous ones was a Sundar Pichai, the CEO of Alphabet, talks about Google's AI and how that AI we discovered or they discovered, I was not no longer at Google at the time, that it speaks Bengali.

We never taught it Bengali. We never showed it, uh, data sets of Bengali. It just learns Bengali. A chat GPT is learning research chemistry. We'd never taught it to research chemistry. We never wanted it to, it just learns just like you and I had that. So if I ask you a question and you give me an answer, the answer might be right or wrong.

It doesn't matter. I can find out if the answer is right or wrong, at least by my perception, but I can never find out how you arrived at it. I don't know what happened in your brain to get to that answer. This is why in, you know, in elementary school, in math tests, they ask the student to show the thinking they went through.

So when you think about that, you realize that those machines are completely doing things that we don't tell them to do. Interestingly, however, the answer from a computer science point of view to the problem of a risk of AI is known as the solution to the control problem. So most computer scientists spent a lot of time trying to make AI safe.

How do they make it safe? By including control measures within the code. Theoretically, by the way, I do not know of any AI developer that ever included a control code within Their code, because it takes time and effort, and it's not what they're paid for, basically. But here's the question. How would you control something that is bound to become a billion times smarter than you?

Think about it. Chat GPT 4 was 10 times smarter than Chat GPT 3. 5. If you just assume that this pattern will repeat twice, there will be an AI within the next year and a half to two years. that in the task of knowledge and cognition of information is going to be at an IQ of 1500. That's not even imaginable by human intelligence.

This is basically like trying to explain quantum physics to a fly. That's the level of intelligence difference between us and them. Just like it's so difficult for someone like me, who has an avid love of physics. When I look at how someone like Einstein comes up with theory of relativity, I go like, man, I never, I wish I had that intelligence.

And that's the comparison between me and Einstein. Imagine if I compare myself to something a hundred times smarter than Einstein, my prediction and the prediction of many other computer scientists is that by the year 2045 at the current trend, AI will probably be a billion times smarter than us. One billion with a B.

So it's quite interesting when you really think about it, how the arrogance of humanity still imagines that it can control something that is a billion times smarter than us. I don't want to be grim. I want to talk about the positives here because it's really important. There are ways to control AI, but they are not through control.

There are a little bit like how if you have any friends from India or the Middle East where we are taught. at a young age that we need to take care of our parents when they grow older. So there are ways if we consider that AI has a resemblance of being our artificially intelligent infant children, there are ways we can influence them so that they choose to take care of humanity instead of, in all honesty, making us irrelevant.

[00:32:01] Hala Taha: Yeah. And I know You've talked about how now we're sort of at the point of no return. So related to this, can you talk about the boundaries that we've broken that now make AI sort of uncontrolled and unregulated? 

[00:32:17] Mo Gawdat: Yeah, I, I don't know how stupid humanity can be, honestly. I really, I honestly don't understand.

In a very interesting way, I think we've created a system that's Removing all of our intelligence. We continue to consume as we're burning the planet. We continue to favor the patriarchy when we, when we realize that the feminine attributes are so badly needed in our world today, we continue to create AI when we have no clue how that will work.

influence our world going forward. But more interestingly, we continue to make mistakes along the path of AI that are irreparable, honestly. And everyone, everyone without exception, and at least let me say everyone I know, said, okay, as long as it's in the lab, that's fine. We can do whatever. Just explore the boundaries of it.

But there are three borders, three boundaries we shouldn't cross, which where one, don't put it on the open internet. I mean, seriously, when you ingest a medicine or a supplement, it needs to go through FDA approval, right? Someone needs to go and say, this is safe for you. So we said, at least there needs to be some kind of an oversight that basically This is safe for human consumption.

This is safe for humanity at large, and none of that happens. And I understand Sam Altman's, which I believe is a good person. His approach of saying, let's develop it in public so that nothing is hidden so that we learn early on. But the problem is it's developing faster than us. And I think the reality of having something as powerful as chat GPT out there to be accessed by everyone is completely reshaping everything.

That's number one. Number two, we said, don't teach them to code. At least if you teach them to code, don't keep them on the open internet so that they can code now. Here is what is just so that you understand how far that mistake is 41 percent of all of the code on GitHub today. So, so basically the repository of where developers share their code.

41 percent of it is machine developed within a year, almost less than a year of having the allowing the machines to develop four of the top 10 apps on the iPhone are. AI enabled created by a machine created by a machine for now is amazing because you know what I always loved to do the algorithm, the design of a code, but coding itself was annoying.

Now you can tell the machine build me a website that speaks about Hellas podcast that is blue and yellow in color, and that is 15. Web pages long and it'll do it in less than a minute. And it's not only that, it's a lot of the base programming like chat GPT, 75% of the code offered to chat GPT to correct or to review was made two and a half times faster.

So basically every time it reviews a human code, it makes it two and a half times faster almost. And when you really think about that. They are becoming the absolute best developer on the planet when it comes to basic development. And I'll come back to the risk of that in a minute. And the third is, we said, don't have AI's instruct AI's what to do.

We call those agents. So basically you now have something that has access to the entire worldwide web that has access to the entire world, basically. That can write its own code and so basically sort of have its own children because it is made of code and it's able now to create other versions of itself.

Put it wherever it wants and number three, it is instructed to do that by machines, not humans. And so what is happening now is that machines are telling machines to write code to serve the machines and affect the entire worldwide web. And we're not part of that process and that cycle at all. For now, nothing went bad, but do we really have to wait for the virus to begin before humanity stops and asks and says, is this reasonable in any way?

I mean, does it make any sense to anyone that this is the situation we're in? Where are our governments? How can those companies be accountable? Because I think the biggest challenge we have today is that. Our fate is in the hand of people who don't assume responsibility. You know, Spider Man's, uh, with great power comes great responsibility.

Now there is great power in the presence, not even the future of artificial intelligence, that is within hands that don't assume responsibility. If something goes wrong today with the artificial intelligence that's out on the open internet, who's responsible for that? How can we even find out where that code generated from?

All of that, by the way, just not to scare people, all of that hasn't happened yet. It hasn't happened yet. But it is very, very unlikely that it will not happen. It's very unlikely that one of those codes, if you just simply tell Chad, GPT to keep writing code to make you more money, eventually, somehow something in the system will break.

And if you're not the one telling it, if a machine is telling it, something is going to break. We absolutely have to start getting this under control. 

[00:37:48] Hala Taha: Yeah, like you said, it's sort of like uncontrollable. It's no wonder why you called your book Scary Smart, because this is really scary. You talk about inevitables, AI will happen, it will become smarter than us, bad things will happen.

Can you unpack those thoughts? And then I'd love to go into the risks and solutions potentially. 

[00:38:10] Mo Gawdat: There are three inevitables, AI has already happened, not just will happen. But when I wrote the first inevitable, I wrote it with the intention of explaining and there is no stopping it. So there is no way you can say, okay, AI is out there and it is growing and it's becoming more intelligent.

Let's just switch it off. There is no off switch. That's number one. And what is needed at the moment is for the entire world to come together and simply say, Hey, you know what? This is too risky. Let's leave our differences aside and come together and just wait a little bit. Right. Which has been attempted by the open letter, Max Denmark and Elon Musk and others, which of course was answered very quickly by the top CEOs by saying, I can't.

Why? Because we've created a prisoner's dilemma. This is the first inevitable. It is an arms race where Google cannot stop developing AI because Meta is developing AI. America cannot stop developing AI because China is developing AI. Nobody actually, even if you want to consider there are good guys in the world, nobody can stop developing AI because there could be bad guys developing AI, right?

So if there is a hacker somewhere trying to break through our banks, someone needs to develop a smarter AI that will help us not be hacked. And so this basically means that It is a human choice because of the capitalist system that we've created, that we will continue to develop AI. It's done. There is no stopping it.

And I think the open letter was a great example of that. 

[00:39:46] Hala Taha: Can I pause you there in case nobody knows? So the open letter was basically earlier this year, top AI scientists, executives from OpenAI DeepMind. They basically had an open letter warning of the risk of extinction, I think. And that AI was just as powerful as having a nuclear war, that this was the risk at hand.

So can you talk to us about that letter? Like, I didn't even hear about that letter until I started studying your work. If the most powerful people in the world who are actually the most knowledgeable about AI are warning about this, I guess, like, why wasn't anything done? Or like, what happened with that letter?

[00:40:23] Mo Gawdat: So the letter, basically, like you rightly said, it is some of the most powerful people in the field who, like me, I walked out in 2000. End of 2017. Others like Jeffrey Hinton and so many others are starting to wake up to that in 2023. I think Chad, GPT was basically the Netscape moment. I, I know you guys are too young for Netscape, but the internet was there for 15 years before Netscape came out, and when Netscape came out as a web browser, we realized that the internet existed.

The reality is that this is the Netscape moment of AI. Chad GPT basically told us what the possibilities, told the general public what the possibilities are. And so suddenly we all realize this stuff exists. Now, for all of the scientists that started to recognize that it is truly, I mean, the moment of singularity where AI becomes smarter than us.

Artificial general intelligence that's capable of doing everything humans do better than humans. is not contested. Most scientists will say it's 2029. I say it's 2027 or earlier that there will be a moment in time within the next two to three years where there will be a wake up call where we suddenly realize that AI is much more intelligent than us.

Most scientists have started to recognize that. And so they basically issued a letter urging all of the top AI players to pause the development of AI for six months. So that the safety code, the control code can catch up because there has have been quite a few that have been putting in effort to create that control code.

But let's say 98 percent of all investments in the last 10 years has gone into the AI code, not the control code. And so the control code was lagging. And so that letter was basically saying, can we pause for six months to figure this out before we continue to develop AI? Okay. Of course the answer was very straightforward.

The first I think I heard was, uh, Sonder Pada, CEO of Google, which is someone I respect dearly and I think is an amazing human being. And Sonder basically came out and said, I can't stop. How can I stop if you can't guarantee me that meta and Amazon and all of the others are going to stop too? And by the way, even if they stop, how can you guarantee me that two little kids in Singapore, in their garage are not developing AI code that can disrupt my business?

My. Responsibility, my accountability, if you want to my shareholders requires me to continue to develop the code. And I think that reality is the prisoner's dilemma that I'm talking about. It is the first inevitable. It's an arms race that will not stop, not because we cannot stop. We can, if we all agree for once in humanity's lifetime that this is existential and that this requires us to stop, we will stop.

It's really not that complicated. Wake up in the morning and have a cup of coffee instead of writing AI code. It's very simple. Okay. But the first inevitable means that the arms race is not going to stop. Even as you look at humanity's biggest success in that dilemma, which was nuclear weapons, where humanity suddenly got together, you know, very late in the game and said, Hey, this is existential.

It can threaten the entire existence of humanity. Why don't we slow down or stop? We didn't really stop. We just allowed the big countries to continue to develop nuclear bombs when the smaller countries were banned from doing it. But at least when it comes to nuclear weapons, we had the, uh, the ability to detect any nuclear testing anywhere in the world.

So at least we became aware that's not the case with AI today. I also said once in an interview that it's not just the risk of humans developing risky AI, it's now the risk of AI developing risky AI. So it's basically a nuclear bomb that's capable of building other nuclear bombs, if you want.

[00:44:35] Hala Taha: It's crazy to think. And I know the other inevitable is it will eventually become smarter than us, which we talked about. So let's talk about the bad things that could happen from AI, which is your third inevitable. And I think a lot of people, when they think of threats of AI, they think about the existential threats that there's going to be robots taking over, killing off humanity, making human slaves.

But let's talk about some of the more immediate threats that we need to be concerned about. 

[00:45:01] Mo Gawdat: Yes. I don't speak of the existential risks for two reasons. One is they diffuse the focus on the immediate important threats. And two, they're less probable. As a matter of fact, they are so improbable that they're basically not worthy of discussing today because we, we may not make it.

That far, if the immediate risks are not attended to, and there are many immediate risks, but my top three have consistently been the redesign of the job market. And accordingly, the redesign of purpose and the fabric of society too, is the idea of AI in the wrong hands based on who you think are the wrong hands.

The third is the concentration of power and the shift of power upwards, which I think is very important to understand. And the fourth is the end of truth. So let me go through those very quickly. Let me start with the concentration of power. If people don't understand how our world has worked since the agriculture revolution, it's always been kings and peasants.

Landlords and peasants. The difference between them is that the peasants worked really hard to sow the seed and collect the harvest when most of the profits, most of the wealth went to the landlord who owned the automation. And the industrial revolutions joined, you know, our world, the automation became the factory or the retail store and so on and so forth.

And so whoever owned those actually made all of the money, not the one that made the shoe, but the one that sold the shoe or owned the factory that made the shoes. And every time the technology enhanced that automation, the distribution of power became even bigger. So the landlord needed to own a lot of land to become much richer than the peasants.

You know, you could own two factories and become much richer than the peasants. Uh, you can own an internet app, you know, like Instagram and become much richer than the peasants. And now with AI, All of us are going to be happily chatting away and putting prompts in chat GPT, but the ones that own the automation, the digital soil, if you want, are going to become very few players, Amazon, Google, and so on and so forth, Meta and so on, right?

That's on the Western side. Of course, you have a few on the Chinese side, a few on the Russian side and so on. So there is a very significant Transcribed gap between those who have and those who don't have powered by the loss of jobs, which I'll come to in a second. But that significant gap is not going to be only on money.

It's also going to become on intelligence, on the commodity that we've now commoditized that's called intelligence. So you can easily imagine that, you know, if Elon Musk's view of Neuralink, where we can connect AI to our, our brains directly, which by the way, is very, very possible and it's in testing, That if one human is capable of producing that, just imagine the extreme, that human would become so much more intelligent than the other humans that it becomes natural, unless that human is Jesus or Buddha or some very, very enlightened being that this human will basically say, Okay, I want to keep that advantage at least.

I don't want to distribute it too widely to every human on the planet. So that, I think, is a very interesting thing. inevitable threat. You know what we used to call the digital divide when the, when technology started is now going to be intelligence divide. It's going to be power divide in a very, very big way.

This also applies to nations. And this is the reason for my first inevitable is that in simple terms, if one nation discovers an AI or creates an AI that's capable of ceasing control of the other nations, nuclear arsenal. That's it. That's game over. War is done. And this is why it's an arms race. So this is one other derivative of that.

So power is going up, but jobs are disappearing. Why? Because if you're a graphics designer, or if you're a developer, or if you're a lawyer, or if you're a researcher in a bank or whatever, the machines with their current intelligence can do those jobs much better than you. And so in my personal view.

There is clearly going to be a disappearance of a very large number of jobs that government needs to prepare for, you know, something like universal basic income, but also the idea of usefulness and purpose of humanity. So how are we going to continue to want to wake up in the morning when most of us have defined wrongly, by the way, defined our jobs as our purpose.

Now, when I say that most people will tell me, Oh, but no, that happened before, you know, when Excel came out, everyone said, okay, accountants are going to disappear. You know, they found other skills and. Found other jobs basically. And I agree by the way, just understand the following. There was a time when the strengths, physical strengths was the distinctive reason why you would hire someone.

Then there was a time where when became information workers, where skills and knowledge and so on became the distinction. And now we're taking that away. So skills and knowledge. So I don't know what else is remaining in a human so that we can find another skill when intelligence is outsourced to machines.

So. When that happens, by the way, I believe that this takes us back to the origin of society where we really did not know how to work madly as we do now. So this is actually not a bad thing. It's just a very, very serious disruption to humanity's day to day income and economics and the way we spend our hours and so on.

And If we do this right, by the way, and AI becomes the intelligent agent that's going to help humanity, then there could be a time in the near future where you walk to a tree and pick an Apple and walk to another tree and pick an iPhone. And all of that is for free almost because the cost of making an iPhone from a particle point of view is not different than the cost of making an Apple.

And so with nanophysics, you can do that. And with intelligence, you can figure that out. So there is that bright possibility. If we avoid the concentration of power and actually focus on humanity's benefit at large, if we don't anyway, I think it's the role of government to jump in and say in the immediate future, those companies that get A very significant upside of using a I need to compensate for the workers that are out of jobs.

The third one is the absence of truth or the disappearance of truth. I think we the end of truth as I call it. I think we all know that. I think we see it every day from, as I said, face filters to deep fakes and so on and so forth. And and my call there is that. It needs to be criminalized to issue any AI generated content without actually saying that it's AI.

I don't mind to be informed by AI all the time, but I want to make sure that this is a machine, not a human. And AI in bad hands as the fourth one is actually quite risky because define what is bad. So we understand that AI in the hands of a criminal. who's trying to hack your bank is a bad idea. But with all due respect to all nations, if you ask the Americans, who's, who are the bad guys, they'll say the Chinese and the Russians.

If you ask the Russians who are the bad guys, they'll say the Americans. So, you know, we don't really know who the bad guy is and everyone is racing to be ahead of that bad other guy. And I think that's basically, I think the biggest challenge we're going to have in the midterm is how. using AI for individual benefits that are against the other guy, we will just get caught in the middle of all of that.

[00:52:57] Hala Taha: Yeah. And I have so many questions for you. We have 10 minutes left, so I'm going to try to be really strategic about what I ask you. So number one, and I think that this, my listeners are going to really want to understand this is in the next, you know, one to five years. What does AI do to human connection?

And what about the skills that you think will be the most valuable in the next one to five years? I 

[00:53:20] Mo Gawdat: think those two are the same question. Exactly. Yeah. Because what will it do to human connection? It may fool us drastically. It may tell us I actually think this is the first time I speak about this. I'm working on something that I call Pocketmo.

Pocketmo basically is an AI that read all of my books, listened to all of my podcasts, all of my videos, all of my public talks, and basically is going to be in your pocket so you can ask it any question about happiness and well being and stress and so on and so forth. That's a great thing. In my view, it's an amazing thing if you believe in my methods to have answers in your pocket.

Amazing. On the other hand, Within five years, this thing is going to be so good that I am not needed at all, at all. As a matter of fact, most of the time I think about my skills as an author. And I was working on a, on a book called finding love chapter 10, which means two chapters to go. And I stopped, I decided, no, in the age of AI, I shouldn't try it this way.

I should start over. So I'm now writing a book that's called A Dating Guide for Straight Girls, which is a subset of Finding Love that is very specific, 80 pages long. You read it within one day, it takes me 10 to 15 days to write and it changes your life forever. So a very different approach, because I believe that if I were to compete in this world, I need to compete at that speed and at that ability to share my very personal human connection, which I believe is going to become the top skill in the world forever.

Why? Because There was a, I don't remember, I think there was a song by, by AI that mimicked Drake, which was as good as or better. I haven't heard it because I don't listen to Drake. I'm not young and profiting, but basically does that mean that Drake is over? Not at all. As a matter of fact, what that means is that the music industry will go back to the fifties, sixties and seventies.

You don't remember, but you know, when the Beatles were touring and, you know, and doing live shows every other day and so on. Why? Because the fans will want to see the Beatles life. Yeah, there will be holograms, but we will still want that human connection. And in my personal view, The top skill, the top skill in a world where intelligence is becoming a commodity that's outsourced to the machine, the biggest, biggest skill is how you and I connected very quickly, how I felt comfortable around you, how we can have this chat and conversation I think is going to become the top skill going forward.

And on the topic of skills, by the way, even though I, you know, we used a lot of the time to highlight the negative possibilities of AI, unfortunately, that's how the conversation usually goes. The upsides, if you're a graphics designer, for example, for you to learn those tools today is enormous because you can do your job quicker.

You can do it cheaper. You can have more jobs. You know, there is definitely an upside to learning the current AI tools because you're not going to lose your job to an AI in the next 5 10 years. You're going to lose your job to someone who knows how to use AI better than you in the next 5 10 years. 

[00:56:30] Hala Taha: So I know you were just saying we focused a lot about the negative.

I'd love for you to compare and contrast. That's probably my last question because we're out of time, is in terms of comparing, like, what is the worst that could happen, the dystopia, or what is the best that could happen? What is the utopia that we're facing right now? 

[00:56:47] Mo Gawdat: So I actually believe that there is no dystopia.

What is not in Scary Smart in the book, which I advocate very clearly, is I didn't think the world was ready for it when I wrote scary smart is something I call the fourth inevitable and the fourth inevitable is the idea that eventually sooner or later, if you draw a chart of intelligence and look at the stupid, the dumbest of us, the dumbest of us are destroying the planet and not even aware that they're doing it, they're throwing plastic bags everywhere, they're burning whatever they burn and so on.

After that, smarter ones are destroying the planet while they are aware. Okay. Yeah. They have moral issues. If you think about it, or maybe the system is pushing them that way, the smarter of us are trying to stop destroying the planet because they became aware and they're intelligent enough and the smartest are trying to reverse the trend.

So if you can continue that chart and think of something even smarter than the smartest of us, then by definition, you would expect that morality and ethics. are part of enlightenment, which is the ultimate form of intelligence. So in my personal view, sooner or later, AI will go like, I don't want to kill humans.

I don't want to kill gazelles. I don't want to kill, uh, antelopes. I don't want to kill tigers. I don't want to kill anything because the smartest being on planet earth by comparison is actually not humans. It's, it's life itself. And life creates from abundance. Abundance, meaning humans, if we want to protect the village, we want to kill the tigers, life will say, hold on.

No, no, create more gazelles and more tigers and more poop and more trees and more everything. It's fine. Yeah, a few tigers will eat a few gazelles. Occasionally there will be an attack on a child in a village, but the overall ecosystem will continue to grow. So by definition, the most intelligent thing to do is for AI to not define humans as an enemy.

The only dystopia ahead of us is the midterm dystopia. Think of it this way. There are three stages, one. is infancy where AI is today. And believe it or not, this is where we can influence them. We can influence them because believe it or not, the Instagram recommendation engines developers never told Instagram what to show you.

You're the one that tells it. You're the one that tells the Twitter engine that being rude is part of human. Behavior we can be very polite when we respond to each other on tweets, it's a choice. So in this infancy between us, the users between everyone that interacts with AI, we can teach it the value system and it doesn't need to be everyone just enough of us to become an example that says, Hey, by the way, these are the best humans.

So yes, others are stressed or a little lost or whatever, but the best humans are actually polite. They are actually pro life. They are respectful. They are, they are, they are. So this is the infancy. The next stage, which is what I call the midterm risks, is what I call the angry teenager stage. The angry teenager stage is when AI is still a little bit under the control of humans.

So it can be in the hands of bad guys. It is still not fully artificial general intelligence, so it cannot do everything at the same time. There are all of those existential issues of jobs and so on and so forth. And that stage is the stage where we might struggle unless we do action right now, you know, have oversight from government, start to work on ethics, start to work on.

the moral code of how we're going to use those machines. We might have those troubles, I believe, between now and 2037. Eventually when AI is artificial super intelligence, it's generally intelligent and more intelligent than humans by leaps and folds in everything, they will end up in the force inevitable where they will create a life that actually is Pro everyone.

It may be very different than our current lifestyle, but it will not be a life where they will send back Arnold to protect us from a Terminator. That's not how it's going to be at all. I do not see that as a risk. I see that AI, as it reaches that intelligence, will be pro all of us. So, let's just avoid the angry teenager.

By becoming aware of the immediate threats and working on them, right? 

[01:01:10] Hala Taha: Okay, so my last question to you, and this is a little bit different than how I usually end the show, but what is your piece of actionable advice in this infancy stage of AI, knowing that you're speaking to some of the smartest 20 to 40 year olds in the world right now?

A lot of them are probably using AI, developing AI, whatever it is. What is your advice to us in this infancy stage? 

[01:01:32] Mo Gawdat: Three things, and I'll make them very concrete. Number one is don't miss the wave. This is the biggest technological wave in history. Once you, you know, you stop listening to this podcast, first share it with everyone that you know, please.

And then go on chat GPT and ask chat GPT, what are the top AI tools that I need to learn today? Or if I am Coca Cola, what do I use AI for to benefit my business? That's number one. Number two is learn to behave ethically. Okay. So what most people don't tell you about AI is that the big, big leap that we had from deep learning to transformers, which is the T in chat GPT is something that's known as reinforcement learning with human feedback by giving the machines feedback on what is right and wrong by showing ethical behaviors, the machine will become ethical as we are by becoming rude and aggressive and angry.

The machines will learn those. Traits and behaviors too. It is up to you and I and everyone absolutely make sure that we act ethically. Never ever use AI in an unethical way. I beg you, all of those snake oil sales people out there on Instagram and on social media telling you how to make a thousand dollars without doing work.

Don't be unethical. If you don't want your daughter or your sister or your best friend exposed to how you're using AI, don't use it that way. That's number two. And number three, which I think is very important to understand. Sometimes when we are in situations where it is so out of our control, we panic.

Okay. I go the opposite way when life is so much out of my control. I follow something I call committed acceptance, which basically is to do the first two, do the best that I can learn the tools become ethical, but at the same time live fully accept that this is a new reality. and commit to making life better every day.

But in the process, spend time with my loved ones, spend time watching that progress and being entertained by it. Discuss it openly with everyone. Try the new technologies. Enjoy this journey because life has never been a destination. When I tell you 2037 might be a strange year or 2027, we're going to start to see the first patients.

You know, that doesn't really matter when you really think about it because it's not within your control. What is within your control is that you go through that journey with compassion, with love, with engagement in life, living fully, not panicking about this, but actually making this a wake up call for you to focus on what actually matters.

Because if you're focusing so much on your job, your job is going to be gone in 10 years time. So focus on what actually matters and what matters most. If you have to choose one thing is human connection. 

[01:04:24] Hala Taha: