AI Twins: The Good, The Bad, The Ugly

Hi, I'm Emily Campbell. Welcome to another episode of Sounds Accurate. Our guest this week is Mo Gawdat, the former chief business officer at Google X, Google's legendary moonshot factory. He's an engineer who applies first principles logic to humanity's biggest challenges. His work offers a stark, urgent forecast on the future of AI. But to understand how he thinks, you first have to know how he treats even happiness like an algorithm. His core idea is that happiness is our default state. And the goal isn't to add things to get it, but to remove the stressors and flawed expectations that block it. It's a process of subtraction, not addition. This view is informed by his time at the cutting edge of technology, but is driven by a deeply human mission. And it gives us a unique window into how he sees the world. Mo, welcome to the show. Thank you so much. Thanks for the kind introduction and thanks for having me. Well, before we deep dive into AI, we love to start our conversations with a quick check-in. How did you start your day-to-day? Today, I'm in Dubai today, and we had floods of rain for a very unusual day at this part of the world, which basically means you stay home because you can't navigate the streets. I couldn't. I had quite a few appointments. So I rushed around, drove in the wet weather. And here I am, smiling and ready for another conversation. Well, I'm glad you were safe as you were running around town. I know those rains can be very torrential. What is one thing that's exciting you or capturing your attention right now? Exciting me is I have quite a few projects that are literally a few weeks away. Hopefully, things that will make the world a better place, which is something that is starting to feature as many more people are listening more and more to the warnings we say about AI. At the same time, what's capturing me is I woke up last Sunday with a very interesting idea of trying to analyze how the world came to where we are right now, which I have to say is probably not the best we've ever been. So when I tend to write, I tend to really, really, really dive very deep. And it does take a lot of concentration, 10 hours a day,

02:35

not really knowing if I'll arrive at a clear conclusion, but something that definitely is preoccupying me very heavily. What are the absolute exact roots of how we ended up here? Wow, that sounds like a deep thinking type of time that you're in right now. And, you know, certainly every day I wake up and wonder how we got here. So I'm sure it's very timely and relevant to a lot of people. We focus really on finance and accounting. And for finance and accounting leaders, a profession which is historically very risk adverse, the concept of an AI twin, which is something that you've spoken quite a bit about, can sound both exciting but also alarming. So can you define an AI twin? What is it? And how do you think about that in the context of a leader whose company depends on accuracy, control and predictability and using AI and the infrastructure and AI twins in your company? Wow. OK, that's a very multilayered question. So let's try start with the definition. The definition, the very, very common definition is sort of a clone of you in the form of an AI. For the current, for the time being, it would be a virtual clone. So an avatar on a screen, if you want, that would be able to replace you in some or all tasks. This is a narrow definition. You know, there could be an expanded definition in terms of twin in terms of the tasks, but not in the liking, if you want. And I think that's something that the industry is catching up to very, very quickly, because for a very long time, the industry, if you think about someone who is as geeky as I am, we were completely obsessed with something that was called the Turing test back in 1956, if I remember correctly. Yes, Alan Turing. Basically started to say machines can be as intelligent as humans. And, you know, here is a test that would let us know when they are. And the test was more or less all about not being able to distinguish. If you had two conversations, one with a machine and one with a human, you wouldn't be able to tell which is which. And I have to say the industry got really absorbed in this, perhaps because we like to win almost as if in a video game,

we're trying to create something that beats that test, if you want. Lots of players were. You could see that in the idea of a digital avatar and AI twin. You could also see it in humanoid robots where we're trying to create robots that are behaving as humans or looking like humans, you know, walking like humans, gripping like humans and so on. And, you know, you'd hear the top influencers in the AI world, you know, people like Elon Musk or whoever saying that this is really important because that means we don't have to redesign the environment in which we live. They can navigate our kitchens as much as they can navigate our factories. And this is a beautiful form factor to allow humanity to blend with AI and robotics in a seamless way. Of course, you have to imagine that the biggest benefits we've had so far came from robots that really didn't look anything like humans at all. You know, whether these are trained grippers or self-driving cars, when you really think about it is an AI robot. It doesn't look like a human driver, but it drives as good, if not mostly better than a human driver. And so this shift of a twin or a digital avatar having to have your liking might not last for very long. What would definitely last is that you would have an AI twin that can perform a task as a human or better than a human, which is the measure that everyone is obsessed with right now. It's the idea that sooner or later, we'll get to something called artificial general intelligence. And general intelligence is a moment in history where AIs can do anything, any task that a human can perform better than a human. So some people predict that to be in a year or two. Some people predict that it will take a little longer, but most of the experts say before 2030. So that's the topic we want to talk about today, like the good, the bad and the ugly associated with these AI twins. So let's start with the good. You've spoken about a utopian future where AI handles all of the tedious, repetitive work, kind of freeing up humans to explore more creativity or allowing people to do different types of tasks. For a finance and accounting professional, what is this positive vision of the future look like? Does it shift their role from being a doer to more of a director, more of a strategist? How can we think

07:47

about something like that now? I think I would like to sort of frame the conversation around a very important point, which is there's absolutely nothing good or bad about intelligence, right? There is nothing, you know, benevolent or evil about intelligence. Intelligence is a force with no polarity, right? And so the way you deploy intelligence is the way our future is going to be written. So if we deploy it for the, you know, for the betterment of humanity, for the, you know, increased production, productivity of work for whatever it is, that is a good objective, it will yield that objective. What does that mean? It means that if we deploy this properly, humans will be freed up to do more things that are more interesting when those things are available. But more interestingly, so this is what most of the industry promises that, you know, if closing the books is a tedious task, maybe you can give that to an AI and the AI will do it instead of you bothering yourself with doing it. Of course, as the industry thinks about that, it doesn't discuss the fact that I like bothering myself doing it, because I'm an accountant, I studied this, and I do it for a living. So something needs to be happening in that space, where where an accountant can be upskilled to do the next level job, or, you know, basically, we find a way for us to, to compensate the work environment in a way where that productivity gain is distributed across everyone. Now, what is possible, which was your question, everything's possible. So, you know, the way the way we have created intelligence today, is going to go through two stages. One stage is what I call the era of augmented intelligence. And the next era is the era of machine mastery is what I call it augmented intelligence is going to last us 10, 15, maybe 20 years, where humans are not fully replaced, but they're paired up with AIs to do things together. Right. And as we do things together, we can do things that we've never, ever dreamt of doing before. We can do them faster, we can do them more efficiently, you can do them more accurately, we can do them in ways we have never even imagined possible. And all of this is not just a promise that the geeks are, you know, saying is going to happen in 50 years time. These are things that we already started to see evidence of, like, for example,

AI inventing new mathematics, all together, new ways of us multiplying matrices, for example, right now. So if we do it right, that means the jobs are more accurate, they're easier, we save hours, we don't have to work as many extra hours, and so on and so forth. There is productivity gain for everyone, there is more profitability for everyone, and accordingly, an easier life, hopefully for everyone, right? That's on the bottom end of the of the job scale on the top end of the job scale. You know, augmented intelligence allows someone like me, you know, I mentioned I woke up Sunday before last, so nine days ago, or 10 days ago, with an idea of a book, I'm now 80 pages into that book already, much faster than my typical writing speed. Because in my past, you know, as an author, I needed to do a lot of research that would take me, say, 14 days to get to a chapter's knowledge, and then write it, I can now do that research in three days, right? And so that kind of productivity gain is not only allowing me to work better as an author, but it's also allowing me to produce more so I can write more often, and also produce much higher quality research and much higher quality ideas. So you're going to see that at the top end, you're going to see ways for us to understand the markets that we didn't before, but ways for us to look at economics in ways that we didn't understand before, look ways for us to, you know, to achieve cost savings that we didn't see before, simply by allowing an AI to look at all of the numbers and all of the records in a trusted way and allowing them to, to basically do the jobs that we couldn't fit within our human brains, because of the limitations that the human brains have. That's awesome. You know, you talked at the beginning around the Turing test and people wanting to prove something. I think there's an element of a human ego, maybe where as you're paired up with these AI partners, if you will, people might start to feel inferior or try to compete with AI. As a leader, how do you take that and create an environment of collaboration with AI? How do you move your teams or how do you move people past that competitive nature and really into that collaboration state? Yeah, it's quite an interesting positioning because, you know, I never really

12:58

wanted to compete with my scientific calculator when I was in engineering university, never really wanted to compete with my Excel, right? You know, interestingly, of course, AI is a lot more autonomous than that. But we found since deep blue IBM's very, very early, very primitive AI beat Garry Kasparov at chess, which was the world champion at the time. No one has ever beaten a machine in chess. And it's it's naive to expect that you ever would be. OK, but but but the thing is, you know, humans did not continue to play chess against machines for very long after we've realized, you know, the machine has won. We started to play in teams of AIs and humans together, playing in, you know, competitions against each other. And we could see an incredible increase in the quality of the game if you want in in the quality of the moves when an AI and a human work together as a team. And and and it's very interesting that we think of AI as something we're jealous of, right? When when the idea for the next 10 to 15, 20 years is augmented intelligence, augmented intelligence is can I work together with this thing to be better? Now, the challenge in my mind is that some of us will choose to do this like I did. There was a point in my life as an author when I said, I'm never going to write again. I is going to be writing better than me. And since then, I've written three books. Right. You know, in in reality, there is a shift in my mind of, yeah, they're better than me in certain things. I'm better than them in other things. And I think this collaboration can create amazing, amazing results. And it it really is proving to be quite an interesting collaboration so far. Now, imagine the amount of progress an organization could have if instead of having a team that is at 110 IQ average, that is at 230 IQ average or at 300 IQ average. I can't grasp how CEOs don't see that. Right. I can't grasp how, you know, how we're so stuck in the in the way we used to do things that we can't see that I personally, as a thinker, as an author, as a even as a mathematician, I'm probably twice as smart as I was in 2023, not because any upgrades happen to my brain, but because now I can outsource something to to to a collaborator that is way smarter

than my scientific calculator or Excel. So how do you bring this change about? And I think that this is quite a challenge for established organization for many reasons. One of them is, you know, we're not really sure about security and where data is going and what happens with AI learns from it and so on and so forth. That's a task that is the responsibility of your, you know, IT team. And there are AIs that are behind your firewall and that you can protect. But then there are the other challenges, which I have to say in finance organizations specifically are really more around the barriers and the resistances. To be in finance, you have to comply to a lot of regulation. You have to comply to a lot of rules. You have to comply to a lot of this is the way things are done. This is how the books are closed. This is right. So there is a bit of rigidity around those things. And my approach, believe it or not, which we built at Google is there was a time in Google's life, which I know most most companies in finance will will find, you know, shocking. But around 2008, 2009, we felt that we were becoming old and boring. Right. And we were becoming too bureaucratic. We were becoming too rigid. We were becoming very slow. We started to say, hey, we're interested in those new technologies. Right. If anyone can talk to me about something that can enable a car to to drive itself or, you know, autonomy in general in transportation or whatever. And and we would allow the bottom up to come up with the ideas. And I think that needs to come from the top in terms of interest, you know, in terms of saying, I'm you know, one of the things I used to do is to say, OK, every Thursday at 11 a.m., for example, my my one hour is booked for people who come up with ideas. Right. And so people who wanted to be on my radar screen would actually book an appointment for 10 minutes and talk about their ideas. And that was the way that innovation started to come in. And when we found an idea that was that had potential, we funded it with enough investment to make it an experiment. OK, so, you know, if you're the CEO of a finance company and someone says, hey, I found a way to understand derivatives a little better and we you know, there is this tool out there that can do this or that and so on. Right. Perfect. Can you go try it in a shielded environment? Show us the result. Come, you know, every two weeks and show us the progress. And here is your maximum investment. Don't go outside

18:10

that. And we'll give you a month to work on it. Right. And if you start to do that, something very unlike finance will happen where you might actually stumble upon something that increases your productivity by 25 percent. Right. And if you don't, you know, you have a limited investment of time and expenses. And in that way, you know, you don't really lose much, but the upside is quite massive. I think that that would resonate a lot with the finance and accounting executives or in team members out there. You know, it's like, how do you make A.I. work for you or you instead of making it compete with you? I think that that's a really critical point. As an author, I know that sooner or later, the majority of all writing will be done by machines. Right. But remember, there is an element of trust between the reader and the author. OK, that element of trust comes from the fact that if I write about happiness, it's because I'm a human who felt happiness and unhappiness. It doesn't matter how intelligent the machine is. That element of trust cannot be established between the author and the reader unless they know I'm a human. Similarly, with accounting, with finance. I mean, think about the number of years that your client dependent on you. OK, and how much how how much value is that in the value chain of that relationship? Because in a very interesting way, your client is also looking at A.I. But if you come in and you say, I'm using A.I. OK, I'm going to do whatever you want for, you know, with A.I., but I'll do it more professionally at scale because I know what I'm doing. You know, actually, that element of trust of I care about you is what they will be interested to buy from an accounting firm. More interestingly, I think that the reality that most people need to understand is that. Whether you like it or not, this has happened, this this change is upon us, it's not something that is going to, you know, that hiding our heads in the sand is going to, you know, make it go away, it's not going to go away. And so if you really, really acknowledge that, then what you want to do is to say, OK, if A.I. is here to stay, I'm going to be the best in the world at it. Right. I'm going to be using it. I'm going to be utilizing it. I'm going to be understanding it. I'm going to keep up. Yes, it's not a comfortable thing to learn a new skill after all of those years. But hey, I'd rather the discomfort of learning a new skill than actual, you know, threat of not having the skill and being replaced by it.

Something that I talk to my team quite a bit about is it's more important for us to be the best utilization, utilizers of A.I. and to really embrace the technology, because today is probably the worst that it will ever be. And it's going to continue to get better every single day. So learn it, love it, embrace it. It can only make us better, smarter, faster in our in our work on a day to day basis. But I think that one of the biggest challenges that I see when I talk to CFOs is trust is critical. And they're very, very concerned about this idea of a multi-million dollar black box mistake, an outcome that was not anticipated or an answer that was given by an A.I. agent that was a hallucination or incorrect. How do you protect against that? How do you create accountability? What are the safety channels that we need to have in place in order to make sure that we don't see those kind of mistakes in this very numbers focused finance and accounting space? I do not know of any kind of way of saying this, so I'm going to say it exactly as it is. Please, please don't shoot me. Oracle CRM or, you know, Oracle ERP or whatever was also a black box. Intuit was a black box. Every one of them was a black box. That over the years, you as the accounting professional did not really understand the back end of any of that. OK, and, you know, there are many, many stories where that black box could have made a mistake. And in many ways, some of them did. And we figured them out and we worked with them. If you if you deployed an Oracle ERP at a point in your life, what you did is you kept your old system and the new system running for a while. You compared the numbers. And then when you felt that they were accurate enough, you discontinued one of them. And I think this is very, very doable in the current environment. Now, the interesting challenge is most of us are fishing for an AI mistake. So so, you know, we don't measure the number of human mistakes. OK, and, you know, and in a very interesting way, what is the ratio? How likely? I mean, Tesla did have, you know, a few accidents and a few fatalities for as a result of self-driving cars over the last few years. So AI caused the fatality. Right. But how many human fatalities happened at that time when, you know, we lose around one point two million lives a year on the roads around the world because of human error?

23:22

Are you being fair to hold on to that headline that your newspaper, local newspaper, wanted to excite you and keep you stuck to by saying AI hallucinates back in 2023? Or are you able to say, well, that has actually improved quite a bit? Right. Or that you can actually say, and by the way, I'm not about using a language model here. I'm going to use an AI system that actually works and is tested. And I'm going to make sure that things run in parallel for a while. It's an investment that I can make. Right. But that transition sooner or later is going to happen. So are you going to be ahead of that or are you going to be behind it? In the old school technology until 2023, a senior leader had the job of a chess master. OK, you know, the best leaders were the ones that could see where the markets are going, could plan and forecast a year ahead, could hire the resources for it, could train them, could, you know, basically steer the Titanic, the ship a little bit to the left. Right. So that they can capture that big opportunity. I can guarantee you the chessboard is off the table. OK, we are now in the age of AI playing a game of squash. Right. You know, you're literally on your tiptoes in the middle of the court and you have no idea in the next couple of seconds where the ball is going to be. Right. And you have to look for it. And if it goes to the right, you're going to take two steps to the right, hit the ball back and then go back to the center of the court on your tiptoes again, waiting for the where the next ball is going to be. This is the current speed of this environment. And and once again, I you know, I know there will be some CEOs that that say, I hate squash. I don't want to play that game. OK, that's fine. But there is no longer a chessboard that determines your advantage. You know, if you're not in the squash court, right, the likelihood that you're going to miss the next ball is quite high. OK, and so the skill is no longer vision and and long term strategy. The skill in software development is what we call A.B. testing to try a couple of things and see what works and continue. But the real, real, real skill in this new environment, believe it or not, is not a I. The real skill is agility, is the ability to look at things and say, you know what, I'm going to dedicate an hour a week.

That's not too much to ask. Go to your language model, whichever one you prefer to use and and just explain, say, I am this kind of professional with those many years of the experience in this industry with this background. I'm interested in those five financial topics. What do I need to be aware of? OK, and many of our eyes today can even set an alert where, you know, I use Gemini and say, update me if there is something new on this. OK, I you know, I go to my YouTube and and and basically tell my YouTube AI engine without really telling it that there are certain topics I want you to show me very frequently every time you get something new about them. And I think an hour a week for a finance professional to be dedicated to understanding how the rules of the new game are being done is not too much to ask. You know, it's basically all about how can you move fast now? It's not how can you move accurately every time? That's great. I think this has been an incredible conversation and there are many takeaways here that I certainly plan on applying to my own team and my own organization. To finish off, we we try to do a quick speed round. So I have three questions. Just first thing that comes to your mind. We'll keep it quick and then we'll let you get out of here. What's one thing that's keeping you up at night right now? Maybe that's not a quick question, but what's one thing that's keeping you up at night right now? I think as we started, it's the it's the book I'm working on. You know, it's called Evil and it is an attempt to understand the very complex web of reasons that go way beyond capitalism, way beyond, you know, debt of why the world is what it is right now. And I think if I figure that out, whether I whether I publish it or not, that would be a worthwhile cause. I would love to read it. I think it sounds like a very worthwhile proposition. What is the best piece of career advice that you've ever received? The career comes in seasons. Most people don't recognize that we think of it as linear. It's not at all. And most of us who are old enough would look back at at the you know, at at the seesaw pattern, if you know what I mean. You know, your career seems to stagnate for

28:33

a while. In physics, we think of that as potential energy. You're acquiring energy that's not moving you up, but it's basically building your momentum in a way that allows you to pop up. And for most of us, our careers had moments where things actually happened. Right. And I really think that most people are impatient around that. Your biggest real investment is your potential. OK. And for anyone who believes that that's staying a year longer in a specific job will increase their skill set. You know, that's definitely a very positive move on your career. Awesome. And finally, what is a skill that every leader should be developing today? It's funny when people ask me this because we're in the age of the rise of the machines. And I continue to say that because we're going to outsource so much to the machines, the top skill a leader will have today is human connection. You know, again, I go back to my career as an author, for example. It is the one reason why people read my work. Yeah, of course, there are interesting gold nuggets of thoughts that you have. You may go like, wow, I didn't think of that before. But I promise you, I will find those better than me in a year's time. OK. The only thing that becomes really irreplaceable is how we connect to each other as humans. Right. And if we can connect to our clients, to our readers, in my case, to our followers, in your case, to, you know, to our employees, to our teams, to our partners, to our peers in ways that that really make them trust us and relate to us in a world where opportunities for humans will become less and less, your opportunity will remain or will grow because you're the one that they want to deal with. Right. And human connection is by far underrated, was by far underrated in the information technology age. It's going to be the top skill in artificial intelligence age. Yeah, that that's very insightful. I love that. Mo, this has been absolutely fascinating. Thank you so much for taking the time to join us today and join us in the evening on a rainy day there. We really, really appreciate it. It's been a pleasure. Thank you so much for having me. And to our listeners, thank you for tuning in to Sounds Accurate. You can follow the show on Spotify, Apple Podcasts, or watch full episodes on YouTube so you never miss a conversation. We'll see you next time.