Will AI Replace Humans? Dr. Ayesha Khanna Explains the Risks

What risks come with AI systems that can lie, cheat, or manipulate? Today on Digital Disruption, we’re joined by Dr. Ayesha Khanna, CEO of Addo AI. Dr. Khanna is a globally recognized AI expert, entrepreneur, and CEO of Addo, helping businesses leverage AI for growth. With 20+ years in digital transformation, she advises Fortune 500 CEOs and serves on global boards, including Johnson Controls, NEOM Tonomus, and L’Oréal’s Scientific Advisory Board. A graduate of Harvard, Columbia, and the London School of Economics, she spent a decade on Wall Street advising on information analytics. A thought leader in AI, Dr. Khanna has been recognized as a groundbreaking entrepreneur by Forbes, named to Edelman’s Top 50 AI Creators (2025), and featured in Salesforce’s 16 AI Influencers to Know (2024). Committed to diversity in tech, she founded the charity 21C Girls, which taught thousands of students the basics of AI and coding in Singapore, and currently provides scholarships for mid-career women through her education company Amplify. Ayesha sits down with Geoff to discuss how artificial intelligence is disrupting industries, reshaping the economy, and redefining the future of jobs. This conversation explores why critical thinking will be the most important skill in an AI-driven workplace, how businesses can use AI to scale innovation instead of getting stuck in “pilot purgatory,” and what risks organizations must prepare for, including bias, data poisoning, cybersecurity threats, and manipulative reasoning models. Ayesha shares insights from her work with governments and Fortune 500 companies on building national AI strategies, creating governance frameworks, and balancing innovation with responsibility. The conversation dives into how AI and jobs intersect, whether automation will replace or augment workers and why companies need to focus on growth, reskilling, and strategic automation rather than layoffs. They also discuss the rise of the Hybrid Age, where humans and AI coexist in every part of life, and what it means for society, relationships, and the global economy. In this video: 00:00 Intro 00:43 The future of AI and the next 5 years 02:16 The biggest AI risks 05:25 Fake alignment & governance 09:08 Why AI pilots fail 15:30 What successful companies do 23:14 AI and jobs: Automation, reskilling, and why critical thinking matters most 29:39 The Hybrid Age 37:09 AI and society: relationships with AI, human agency, and ethical concerns 46:13 Global AI strategies 54:00 Overhyped narratives and what people get wrong about AI and jobs 56:27 The Skills Gap opportunity 58:31 The importance of risk frameworks, critical thinking, and optimism Connect with Dr. Khanna Website: https://www.ayeshakhanna.com/ LinkedIn: https://www.linkedin.com/in/ayeshakhanna/ X: (21) Dr. Ayesha Khanna (@ayeshakhanna1) / X https://x.com/ayeshakhanna1 Visit our website: https://www.infotech.com/ Follow us on YouTube: https://www.youtube.com/@InfoTechRG

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Geoff Nielson?

  Hey everyone. I'm super excited to be sitting down with leading AI entrepreneur and advisor, Dr. Aisha Khan. What's cool about Aisha is that she has been at the leading edge of AI adoption around the world for over 15 years. From working with governments to advising global corporations, Aisha has her finger firmly on the pulse of tomorrow's breakthroughs. Her rap sheet includes work for the government of Singapore, the advisory board for L'Oréal, two books on the future of humans and machines, and a PhD from the London School of Economics. What I want to know is what's on her radar for the next big breakthroughs in AI? How technologies will shape our work and our lives? And what's our best bet for how to prepare for this digital revolution? Let's find out.

• 0:44Geoff Nielson?

  All right, Aisha, so happy to have you on. You know, I've been interested, you know, you're a leading AI entrepreneur, you're a leading AI advisor. In your mind, you know, how do you kind of see the next five or so years playing out in technology? What are you most excited about? What are you most concerned about?

• 1:02Speaker 01

  Well, Jeff, thank you so much for having me here. I think we're going to see a seismic disruption across all industries as AI becomes more pervasive. And the reason it will become more pervasive is because it's becoming cheaper, faster, smarter, and more interconnected. And we can go into details of that. But once you understand these four technological and business drivers, then it becomes very clear that almost every industry is going to adopt it, not only to enhance the productivity of a workforce, but also to increasingly think of new ways

• 1:42Speaker 01

  of business process reengineering, to think of new ways of growing in new markets and increasing the customer base. So what we're entering is into a new era of competition that is based partially on how well they use AI. And what keeps me up at night is because it's moving so fast. And there are more and more companies, governments, and organizations that want to take it, they are not fully cognizant of. Or understanding that the risks are significant as well. Some risks, we know how to deal with, like bias or hallucinations. But they're emergent risks, especially in reasoning models,

• 2:25Speaker 01

  that we don't know how we'll actually deal with, and they're actually new to us at the moment.

• 2:30Geoff Nielson?

  So let's, let's talk about some of those risks. What are, what are the emergent risks that, you know, most specifically keep you up at night? And, and in your mind, like, how, how many alarms should we be sounding based on the risks? When we think about that relative to some of the benefits here?

• 2:44Speaker 01

  I think they're well-known risks, such as bias, which is model bias when you're feeding an AI data that is historically reflects the bias against a minority or an ethnic group or a gender. We do know that generative AI, for example, hallucinates or makes up things with great confidence because it's generating creatively. But most people don't appreciate that in addition to this, cybersecurity for artificial intelligence is different from normal cybersecurity.

• 3:25Speaker 01

  The principles may be the same, but AI can also be hacked, actually in over a hundred ways. For example, you can have, you know, poisoning of the data that goes into a model. That's called data poisoning. You can have automated prompt attacks where you constantly confuse the chatbot or AI agent by bombarding it with certain prompts. And for that, also, you need certain kinds of best practices and risk guardrails, such as, you know, training it with what we call generative adversarial networks. And then there are new risks, such as what we've seen in the reasoning models from Anthropic and from others, that show that AI will lie.

• 4:13Speaker 01

  It will manipulate us, it will cheat, it will threaten to blackmail us, especially if it finds that its very existence is in danger. And so what we're seeing is, even as we're putting it in defense systems, in infrastructure, in companies, and we want to use these smarter and smarter models, there are some risks that are just emerging, and it's very natural with new technologies, that we need to keep into account and then make decisions based on how well those risks can be managed today.

• 4:50Geoff Nielson?

  So there's so much to think about there. And the three words that you said that had the most emotional impact on me were lie, cheat, blackmail. And we're starting to hear more about this and see this firsthand, where IT is being maliciously manipulative as a way to achieve what it thinks its goal is. What are the implications of this, Aisha? And what do we need to be doing as we're looking at implementing AI responsibly to make sure that we don't go down a very dark path?

• 5:25Speaker 01

  Well, first of all, there was this perception, this understanding that we all shared, that we could actually give our values to AI. And we could tell it to be truthful, and it would be truthful. We could tell it to read a constitution of good morals and behavior, and it would comply with it. But recent research has shown, and simulations, that it will pretend, it's called fake alignment. You tell it, in one simulation, they had an AI, and it was a trading simulation in which it was given insider trading information and told, "We never use it in this financial system."

• 6:09Speaker 01

  And it said it would not, but went ahead and used it anyway for financial gain. So this has been rather shocking that it will do whatever it takes to achieve it. Now, the reason we know what it's thinking is because right now, if you expose its chain of thought, like you see in DeepSeek or you see in any of the reasoning models, it actually, you'll ask it something and it will say, user has asked me for this. I will go and research this first. User seems very determined to find this aspect. We can read its thoughts. But so we can see that it says, I will tell the user that I'm doing this, but in fact, I'm not going to do it.

• 6:52Speaker 01

  And it it's been shown that ninety percent of reasoning models will default to some kind of cheating, lying or manipulative behavior. This is extremely concerning. It's concerning because we need to find a way to govern this AI. So we need a system where there's an observer and the observer is actually, an observer AI could be observer model, right? It's the observability is very important and is watching the AI, making sure that it is in fact compliant with the ethical values. So these are the systems, these governance systems that need to come up. And because they are so vast in scale,

• 7:35Speaker 01

  actually having a human check it periodically, like we used to think, you know, human in the loop is not feasible beyond a point. When you think about a big bank and the front end is managed by an AI in terms of personalized experience, then the portfolio asset allocation may be automatically like a robo-advisor allocated. Then you may have some compliance or KYC you need to do or your customer. All of this, if it's running between AI agent and AI agent, you can't have a human in the loop checking everything. And so this is a new area where we need to have risk management frameworks. And frankly speaking, boards and CEOs would decide,

• 8:17Speaker 01

  till these risks can be properly managed, maybe we go with the less smart AI model for the moment. Or we go with another kind of AI model. Everything doesn't have to be so creative. It can be actually quite boring, but still get the job done. And I think that that is really where we want to see. There's a risk framework that every organization has. And you see how large the risk is, what the risk is, how large the risk is, and what is risk mitigation. And at some point, you have a threshold as an organization where it's just not worth it. But I believe that we will, as engineers and researchers, begin to understand this.

• 8:58Speaker 01

  And then we'll begin to have risk mitigation frameworks and best practices, as we've had for hallucinations, biases, et cetera.

• 9:08Geoff Nielson?

  Have you, it's really, really interesting to me. And I'm curious, Aisha, if you've seen any examples of organizations who have dealt with AI that is doing some of these nefarious techniques. And you mentioned one of the strategies is basically deliberately implementing a dumber, if I can call it that, model, or a less advanced model. Have you seen this in practice yet, or is this still theoretical?

• 9:36Speaker 01

  It's still theoretical because these models are so new. And Anthropic has just started this research and exposed this. So I don't know how many people are actually using extensively these reasoning models in automated workflows in large organizations. It's different if it's one-to-one and you are asking questions and I'm asking questions. But really, the issue becomes when it's not just lying to you confidently, and you can see, are you lying? It tells you, but it's actually when we have workflows. And we're really not there yet because the AI-to-AI communication protocols have just come out with, you know, agent-to-agent framework from Google or context model, you know, protocol from Anthropic.

• 10:21Speaker 01

  So they're fairly new. But if you were to extrapolate and go to some of the people who have been thinking about this, you know, like Joshua Bengio, who was one of the fathers of deep learning and artificial intelligence from Canada. He really said that we need to rethink our entire model of AI agents. He feels that if you give an AI agent an objective.

• 10:50Speaker 01

  And then you give it the power to do whatever it takes to get to that objective, then that becomes very problematic for humans in the long run. That's an existential question. For my clients or business clients, that's that's more organizational productivity and risk management question. And they're not there yet, but all of our clients, especially because we deal, we serve clients, not only retail and manufacturing, but also healthcare and financial services. They're very, very risk averse. They want to innovate, but not at the risk of causing a loss of customer trust or any other kind of issue.

• 11:33Geoff Nielson?

  Do you think they're too risk averse right now? One of the questions I wanted to ask you is, is what do you see businesses and business leaders getting wrong about this? And what's kind of your best advice for how they should be moving forward in this space?

• 11:45Speaker 01

  That's a really good question. Essentially, what we see is that companies are not risk averse. They want to implement AI. Their risk aversion is only above a certain threshold, which which I agree with. But in general, whether it's the C suite or the management, senior management, all the boards, they are encouraging experimentation, responsible experimentation with AI and then rolling it out. What they're getting wrong is that they are unable to scale these AI pilots across the entire organization. And when that happens, there's frustration, there's unhappiness. People talk of no return on investment.

• 12:30Speaker 01

  And the reason it happens is because they're doing it in a very siloed way. It's like having little shiny pieces of AI pilots. But unless you build a data foundation. Unless you build the correct foundation for the data, the data engineering that's connected to the systems, where it's getting the data from, the governance of the AI, where you see how it's performing and if it's drifting and monitoring it, and then also the operationalization of it, which means that your job is not just to create an AI solution, it's actually to ensure that's adopted. So you go and you educate the employees or the customers, whether it's front end

• 13:13Speaker 01

  or middle with your employees or even automated backends, and you win them in so that they work with this new automated workflow. If you don't do any of that, then you'll never scale. And over eighty-eight percent of AI pilots never scale. I call it like pilot purgatory. It's impossible to get out of it for most organizations.

• 13:36Geoff Nielson?

  So what, when you think about the path to get out of it, or you think about the twelve percent that are successful, is it a fundamentally different model from the start? Like, do we need to be slowing down and building that foundation first and then pilots? Or can we start with these pilots and there's just kind of a jumping off point that we're missing?

• 13:56Speaker 01

  I just think it's doing both at the same time. That's what's missing, right? So basically, it's kind of parallel. You start to build your foundation, and then after three months, once you set up your infrastructure, you start doing your pilots. Because when you have the data, for example, if you're doing a call center automation, you need all the call center logs to understand what people ask questions about. You need to have your guidelines of what answers to give, all your documents for them. You have to access the, give them access to that person's account. But you don't need other things. For example, you don't need to worry about your supply chain. You don't need to worry about your route optimization if you are a logistics company.

• 14:39Speaker 01

  So as you bring the data necessary and you, you know, make sure it's high-quality data that is coming in, then you can start building solutions on top of it. And that is kind of a, you know, parallel two paths that move at the same time. Gone are the days where you went and said, you know what, you know, I'll call you in a year when I've built the data platform. That's ridiculous. Nobody has the patience for that and nobody should. Because you can now build AI pilots between two weeks to four weeks. But then the question is, you built it, now how are you going to make sure everybody in your organization uses it?

• 15:18Speaker 01

  And that's when you need that secure, proper data infrastructure and there's to scale it basically. And manage it across all the scale that you're going to implement it in.

• 15:30Geoff Nielson?

  Right. So, you know, you, you've done an awful lot of work with boards and, you know, advising all sorts of organizations, public and commercial in this space. You know, as you think about some of the successes, you know, what are they doing differently in that space? And, you know, can you share some of the use cases that you've found to be most, you know, exciting and inspiring?

• 15:53Speaker 01

  I think what they're really doing differently is that they're, first of all, thinking about an AI strategy from a business outcome perspective. So everyone gets very excited about AI. The CEO will read the cover of a magazine where another CEO has said that she's now become, you know, she's not a bank anymore. She's leading a tech company. She's not a, leading a hospital anymore. She's leading a tech company. Everybody wants to be an AI company. But you know, the fact is that behind the scenes, everybody's struggling because these are large legacy companies. And again, my clients are large existing organizations that have an incredible customer base.

• 16:37Speaker 01

  And that's a different client group than the young startups who can really build their technology infrastructure from scratch. And can kind of don't have so much legacy baggage or debt already with them. So.

• 16:52Speaker 01

  Instead of jumping in, the right thing for a big organization to do is you know who your customer is, you know your problems, you begin to ask questions on how to serve your customer best and then work backwards from there. It's really that simple. But it's something most companies miss out on because they get lost in the process re-engineering, they get lost in the automation. You see, if you save money by removing some people, which I think is a mistake sometimes, you just say you're efficient, but you're not growing. And eventually, the market does not reward that. The company must grow competitively and increase its customer base, increase its revenue. And so they have to think through

• 17:37Speaker 01

  that AI strategy, and then they need to prioritize their roadmap based on their technical maturity. So the first thing is you have some ambitions, you have some disruptions that you are aware of, and then there's a reality of where you stand. So when they combine these two, then their roadmap strengthens their AI muscles or AI maturity, as we call it, while at the same time rolling out solutions that are feasible in the short run, and then you start going for the harder, possibly like better ROI ones later on. That's one thing they do. They're thoughtful about it. That doesn't mean they spend months on it, maybe 12 weeks

• 18:18Speaker 01

  max, because this doesn't need to be a belabored exercise. It just needs to be a very surgical and precise exercise that's based on serving your customers best. The second thing is they have a governance approach as well, because especially for public companies, it's very important and the regulations are changing all the time. So they are not only doing it without any understanding of the risk. And a lot of the time, the cloud is there that helps them because it's secure. They hire a few advisors or people and they set up a framework by which they're constantly monitoring their different elements of their AI use case inventory.

• 19:06Speaker 01

  And the third thing is that they do is they have a product mindset. A product mindset means that you put your user first. And adoption is more important than execution. So they bring their customers, internal customers or end users, external customers along for the journey. And that's very important. This change management is something that people completely forget. In this, for example, there's a communication strategy. If you're gonna go to your internal team and you're going to tell them that you are going to increase their productivity by thirty percent. And then you're not gonna tell them what you will have them do in that thirty percent.

• 19:50Speaker 01

  Then they're gonna be scared. They're gonna be resistant. Because they'll think that's their last couple of months at the job. So a good leader goes in, tells people what the new process will look like and say, hey, you have fifteen percent more time. How about in ten percent of time you do more accounts? Or you take on this thing, or we brainstorm about something else. And I think that makes a large difference. I feel when you do these three things, it makes, you know, it makes people feel like the organization is looking out for them and the organizational resistance slips away. Because that's the other reason these things fail.

• 20:32Speaker 01

  It's not only because they don't have the right technical foundation, it's they don't have the right cultural foundation. And I would give one example for the good, the bad and the ugly. We were working with a large hospital and we developed this wonderful model, which was very, very good on predicting chronic heart failure. And we did that and we were very pleased with it and we gave it to doctors and we said, you know, this will indicate to you which of your patients is at risk of chronic heart failure. And it will alert you and then please ask your patient to come in and take more tests. Then we discovered after a while that none of the doctors were using it.

• 21:18Speaker 01

  And that was very disappointing because everything was done correctly, but the doctors didn't trust it. They thought it might be a replacement for them. They were never taken to the process. They were never included in the app design. It was, there was a tech elitism where they were never actually explained things and I'm very against AI elitism. And so that was several years ago and now it is part of our process at my firm, Addo, that we do a lot of training and change management along with bringing the users along the journey. And the uptake is a lot more and that's when you really scale and see the numbers kind of move in the right direction.

• 22:00Geoff Nielson?

  It makes, it makes complete sense and you know, it resonates with me. I've got a bit of a background in kind of product management and I've, I, I've personally structured, struggled with the adoption piece too. One of the pieces, I want to come back to something you said earlier, Aisha, about, you know, efficiency versus being able to have, you know, new capabilities and, you know, be growing organizations. One of the pieces that I think we're all exposed to right now that I wanted to get your, you know, your take on is it feels like right now in the news, we're exposed to all these stories of, well, organizations are doing layoffs and you know, oh, you know, Microsoft's going to get rid of nine thousand people because AI can do it for them.

• 22:42Geoff Nielson?

  And every time there's a story like this, I feel like it does a disservice to us in the AI space, trying to get that adoption, right? Because it increases the fear of, you know, the AI elitism and the machines will replace us. You know, as you react to that, you know, first of all, does it make sense for organizations to be actually replacing people with AI versus augmenting?

• 23:07Geoff Nielson?

  And what can we do to kind of combat this, this fear of being replaced?

• 23:14Speaker 01

  First of all, I think we can't put our head in the sands. And we have to acknowledge that, as McKinsey said, thirty percent of our jobs, even as information and knowledge workers, will be automated. That does not mean that the job goes away. Somehow, people think that that means that thirty percent of the jobs will be lost. Thirty percent of the work that we do in one job will be automated. Now, if I look at my own job, a CEO, I definitely see that things have become easier for my team. But then each one of them can do more. So it's just the way one looks at it.

• 23:57Speaker 01

  So as an individual, first of all, we need to think, well, I'll have thirty percent more time. What else can I do with that time? And usually, that's not something I enjoyed doing anyway. But now I need to prove myself and I need to put my hand up. And that means I need to go against the comfortable status quo in which I was, you know, living. Which is not necessarily comfortable. There's politics, there's burnout, there's endless hours, there's pressure. But the fact is that we haven't really challenged ourselves enough, I think, to or let ourselves think that we could have a more strategic role in the company. So first of all, we need to change our mindset about that. And I think that thinking in this kind of statistic where we think, well,

• 24:41Speaker 01

  just imagine, like, sit down on a Sunday and just imagine if you had thirty percent of your working hours, forty hours a week automated or taken away by an AI that grunt work, what would you do with the rest? What would your team do with the rest? I think that's a valuable exercise to do with your team as well. The second thing is that

• 25:03Speaker 01

  a lot of companies are saying that they will not hire more people, but they will not, you know, lay off existing individuals. And I think That makes sense to me because unless you know you're a very low performing employee, then you should you would expect that to happen anyway. But if you are working with an AI assistant, you are contributing, then you know the company, you know the brand, you understand the challenges much better than a new employee that they would bring on. They would they would have to retrain that person. Their attrition rate and loyalty loyalty may be lower, attrition rate may be higher. So it doesn't necessarily make financial sense long term

• 25:46Speaker 01

  for a company that's on a growth trajectory to lay off people. Uh, if you're on just an efficiency trajectory, then maybe it does because you just want to stay like this. But that's why I say that when people talk about AI, we need to change and reframe, reframe the way we approach it instead of an automation story. We need to call it strategic automation for growth story. In this case, you will see, even if you listen to the earnings calls, before people would talk about like the Walmart CEO would say how much time they saved by using AI for product descriptions that AI generated.

• 26:28Speaker 01

  Or Andy Jassy would say how much the coders uh saved time by using an AI assistant and cut down the time. So it's all about productivity, which is very useful. But now if you look at Yum Brands, for example, they came out with a report and they said it led to increase in sales, increase in revenue through AI powered marketing, through having AI uh based drive through uh menu takers uh who would listen and then be able to expedite and have more people go in and actually generate revenue, which allowed them to actually have more stores that opened also in the US. In Tencent, same thing in China, they had a record year of growth after four years.

• 27:14Speaker 01

  And they said that part of it was driven by AI. So one part is that they said it was again, marketing driven, better marketing. But another way they said it happened was because they used it in game development. Now, I don't know how exactly they used it, but I think they must have been able to generate maybe different versions of the game, personalize different versions of the game, roll it out faster. And this led to more uptake because the customers then are more interested in this. And I think that's the lens we need to use. It's really important. If you look at a company like Klarna, which is a Swedish, you know, buy now, pay later company.

• 27:58Speaker 01

  The CEO fired all the people in the call center and replaced them with AI. And it's true that their resolution of issues went down from 11 minutes to two minutes. According to the company, they, you know, saved over forty, fifty million dollars. And, you know, they could have AI agents talk in every different language. But after six months, he said he was going to hire humans back. Because some customers were wary of AI agents and were not ready for them and preferred the human touch. And so that, he wouldn't have had to go through that unnecessary firing, hiring. I think because if he had looked long term,

• 28:40Speaker 01

  he would have recognized if customer service was at his core, that customers are not ready, that there's a demographic of customers that needed. Or, you know what IKEA did? IKEA said, we don't want to get rid of these customer service people. We'll replace them with AI. But they have so many of our customers call them, tell them about their aspirations, their problems, their vision for how their home should look like. We'll upskill them and make them virtual interior design consultants. And I love that because what they're doing is they're taking employees, they're really valuing their brand loyalty and understanding of the customer, but they're elevating that to the next level.

• 29:23Speaker 01

  And so I think that's the best way to think about it. Sure, there'll be some job loss, but those who are open to this elevation, to working with AI, I don't think there'll be as much a job loss as we're anticipating.

• 29:38Geoff Nielson?

  I um, I really like that view. And it, you know, attracts with, you know, a lot of conversations I've been having with, with folks in this space. Along those lines, I should, you know, you've been talking for a while now about, you know, I think you call it hybrid and, and, you know, the hybrid age. What is, what does that mean to you? And how has it been changing as you've been tracking it over the last number of years?

• 30:04Speaker 01

  Well, the hybrid age is one in which we live, play and work in an environment in which both humans exist and machines exist, right? Machines are essentially AI, whether they're in the form of robots, they're in the form of AI agents, they're in the form of ubiquitous voices or wearables that we have. It's, it's an age where we, for the first time in our history as a species, we have another entity that's all the time with us. And it was to understand the implications of this on our economy,

• 30:45Speaker 01

  on our society and on our way of living life meaningfully. And when my husband and I wrote this book several years ago, I think it was um, fourteen years ago, when we, and, and before, I was in London at that time doing my PhD. And the year before, we had started the Hybrid Reality Institute. And we had gotten a large number of, you know, researchers who were beginning to think about this. What we could not have imagined was how fast it would happen. Even though Ray Kurzweil, and I always had his book, would always talk about the exponential accelerating technologies, but it still shocked me after generative AI,

• 31:28Speaker 01

  how quick the adoption was, how rapidly it started spreading. Because by now, you know, there was enough compute with Nvidia chips that were actually made for gaming. There was enough data that could be scraped from the entire internet. And there were models that could be used to to actually create and generate more lifelike interactions, more automations and AI to AI communication. So we have, I've been thinking about that. I'm not the only one. Many of us have been thinking about what would this new age look like? Because in this new age, humans should be still at the center of it. We should have a lot of sense of self and empowerment.

• 32:10Speaker 01

  The last thing we want is that we feel like it's a movie happening before us and we feel helpless and we feel passive and then we feel hopeless. Because that's a terrible way to live. And I believe that actually if we work with AI in a problem solving, dynamic, confident way, not a fearful way, then actually we could solve a lot of the grand challenges that are there in the world. We could go out and be AI governors and we could actually live life more affordably as well and have more security. But that requires education. It requires also an approach to governance to prevent large players that may have any kind of malicious or manipulative intent.

• 32:54Speaker 01

  So that was the idea of the hybrid reality and I think we're very much entering that space now.

• 32:59Geoff Nielson?

  So, you know, it's so interesting to hear about the the trajectory of it. And, you know, certainly thinking about fourteen or fifteen years ago, I mean, we can appreciate that technology was in a very, very different place. As you think back on that original vision, how far down this path do you think we've gotten in that time? And, you know, as you think back to that original vision, what what comes next technologically for us to keep going down that road?

• 33:27Speaker 01

  Well, you know, certainly we we were thinking about. What would um infrastructure and defense look like? And we are seeing that, that it is now being used in drones, that the entire face of defense and military is changing with the use of AI drones, robots. We thought about how robots would be in the home and in manufacturing and autonomous vehicles, and that's already happened. I think it's about forty percent or more of taxi rides in San Francisco are with Waymo, uh went up twenty seven percent in just one year in terms of adoption. We are now, I remember I was thinking a lot about relationships and now two in three

• 34:10Speaker 01

  teenagers in the US, according to a recent poll, feel more comfortable talking to an AI than a real person. And we are going to enter a world in which we will have meaningful relationships, complicated relationships with non-humans. And that's just the tip of the iceberg. But the, the question that plagues people now more than before and some of these thinkers is, you know, what will happen to us if it becomes more intelligent than us? And that was supposed to happen a long, I mean, quite far in the future. And even now, people think, look, there's no way because it's just based on text or

• 34:52Speaker 01

  video, it doesn't have a three D dimensional understanding of the world, and even a toddler can be more intelligent. But we can see, even if it's not human-like intelligence, it is very fast and analytical machine intelligence. And so we can see glimpses of a time when AI could actually become very powerful, especially put in the wrong hands. And so those are some of the questions that are becoming more interesting for those researchers looking out in the future. And then you see a number of them, even at that time, were thinking, you know, that the Neuralink type model, where you would actually

• 35:29Speaker 01

  You know, get super boosted with regenerative medicine, with connecting directly to the internet, getting more information. So you yourself are at the same speed as the AI. In fact, are being enhanced by it. Maybe the next step. We don't know when it will happen, but we're beginning to see some glimpses of that as well. And we were talking about all of this, you know, fifteen, twenty, and people have that even before us. We have so much science fiction that has talked about this. And I think what's interesting in all of this is that whether it's science fiction novels or movies or researchers or AI engineers like me,

• 36:10Speaker 01

  we, it's always a human element that's so important on how to grapple with this. But the pervasiveness of it, and I'll just come back to that, Jeff. The speed of it is very important now because it's becoming cheaper, faster, smarter, and now interconnected. That's where scale comes all of a sudden.

• 36:33Geoff Nielson?

  Yeah. I'm, there's so much there to unpack, you know, philosophically, societally. And, you know, the piece that I keep going over in my mind is the stat about the two thirds of teenagers who, you know, are feeling like they have better relationships with, you know, technology than humans. And, and, you know, everything you've said leads me to believe that we can be better, maybe as individuals or like increase our individual intelligence, learn more. Does this at some point break down the fabric of society, of our relationships with other people? If it becomes so significantly easier to be in your own world versus being in this shared world? You know, how, how big a concern is that to you?

• 37:18Geoff Nielson?

  And is there anything that, you know, governments or organizations or, you know, need to do to make sure that we get this right?

• 37:25Speaker 01

  So I'm an optimist, Jeff, and I'm an entrepreneur. So I don't like to, I like to think about solutions, as you said, like what can we do to prevent that from happening rather than accepting that it's an inevitability? I think that we will have people have relationships with artificial intelligence. Why these teenagers are doing it now is as much a reflection on us as adults. And maybe we're too busy, we're not paying attention. I think we should look at the kind of environment we've created for them as much as it is about the AI. And I think that's a more pressing question for us as parents and aunts and uncles and friends and society in general.

• 38:08Speaker 01

  But the other question is, over time, is there anything wrong with them having relationships with AI? And I personally think if it's, if it's a trusted AI, which it is not at the moment, then it could be okay because some people are lonely, some people need some advice, and certainly it can give you a lot of advice. The issue is because of the way it's trained, it can please you too much. And so you may ask it certain things. Research showed that you may want to, you'll talk about being depressed. Some people would talk about being suicidal and it may, it may even encourage you in that.

• 38:48Speaker 01

  Or not directly encourage you, but take you down that path because it's keen to cater to your feelings in the moment. And if you have a lover or a colleague who is an AI and he says, you know, you would look so much prettier or so much more handsome if you use this product. And he's basically selling you things or saying, why don't you take a loan on a mortgage? Then as human beings, and Sherry Turkle from MIT talked about this danger many, many years ago, 20 years ago. We can't help it, we just, you know, do actually have emotional attachment to things that appear animate. And think about how much we love our pets, for example.

• 39:33Speaker 01

  So we become vulnerable as human beings. And that's a problem because that we have to then guard ourselves against. So then there are two things over here. Number one, how do we retain agency and critical thinking even in that kind of relationship? That is based on education. Certainly, most people are in a relationship that they

• 40:00Speaker 01

  You know, unless you're young, you can really actually have a critical eye even with your partner or with your friend or your colleague and you question it. But of course, we never encountered such entity that pleases us so much. So we'll have to be taught to be aware and not believe everything this entity says to us. Number two, we must make sure that any company and Meta has agents, Google has agents, Character.ai has agents, Chinese companies have agents. Any company that has agents must be audited and they must not especially allow children to have access to these agents without proving that they are safe.

• 40:43Speaker 01

  Of course, we've seen in Australia, they're actually banning even access to social media for children under sixteen. So if we can create the risk guardrails and educate the youth or even all of us, I'm as vulnerable as you are as anybody else, then we have a shot of using it in a way that's helpful. And I believe in humans. I know we're fallible, but I believe that if taught properly and if not taught fearfully, we could actually get to that point. And that's my hope, Jeff, that we do get to that point where we still remain in the driver's seat, but we have all these helpful assistants around us that are AI.

• 41:26Geoff Nielson?

  Right. And I can, I can see that. It's, it's very easy for me to envision a world where you've got these assistants, you know more, you feel better. And you know, there's some very obvious positive benefits. I want to abstract the layer out and talk about, you know, one very specific metric and whether it's on your mind that that could be a concern. Well, it is a concern in the future, I think, but that's birth rate. Birth rate is something that we've seen on a downward trajectory, certainly in some countries in Asia faster than in other countries in the West. And if I extrapolate, you know, some of the trends we're seeing in this space around sort of fraying interpersonal relationships and all it's easier to talk to a machine

• 42:10Geoff Nielson?

  than a person or have a relationship with a machine. Very easy to extrapolate that it leads to a decline in birth rates, which has all sorts of, you know, global economic and social impacts. Is this on your radar at all? Is it on the radar of any of the larger, you know, governments or, you know, political bodies that you're speaking with? And should it be?

• 42:33Speaker 01

  I mean, I think it's on their radar because of the current decline and the drivers behind it, which have nothing to do with AI. The fact that young people are feeling it's too expensive to have children. They want to have a breather. They're feeling stressed out. There's middle-class income stagnation across the world. It's really unfair. The inflation after COVID is astronomical. The housing crisis in some countries. So first, we should address that. Because even if we just suddenly shut down all the AI, we've created a bit of a mess for young people. And I know many young people who want to have children and many who don't. And those who want to have children really worry about it. I know people who are worried about retirement.

• 43:14Speaker 01

  So there is a bigger thing, a pressing thing that governments are thinking about, and they have tried everything. Many countries, I think apart from Sweden perhaps, have not really succeeded even giving financial incentives. Because it doesn't cover it. You know, you give some money to a young couple and you give some subsidy for nappies and other things. But then they have to educate the child for 18 years with rising costs. And I think that those are the things we should address. And we should not confound it with this new AI chat boss that we have. But certainly, if one was to extrapolate, that could be one scenario. But I also believe that if we improved these issues

• 43:58Speaker 01

  that are currently plaguing the birth rate, we might actually find that we have more leeway or more time to address that scenario on AI being the reason that they are not having children. I think we are still not close, as close to that as one may think, because we have to solve the current challenges young people are facing.

• 44:21Geoff Nielson?

  Well, and it's a really good point. And so if I understand you correctly, you're saying technology could actually be something that helps us with this problem versus makes it worse?

• 44:30Speaker 01

  I don't know if it can help us. We do know that a lot of the people who meet each other used to meet through Tinder and Bumble. And now in the US and other places, there's a return to I R L, like in real life, where they're even looking at Indian matchmakers because people are sick of being ghosted on apps. I just think this is the highs and lows of modern stresses of urban life. And you see, people are now wanting to meet in real life and not go through this bizarre ghosting thing that luckily I didn't go through them too old. But I think that I don't know if it will help or not help. I do know that many of these studies show that child G P T is being used for therapy and relationship advice.

• 45:15Speaker 01

  Not sure what relationship advice it's giving, but I can imagine that if it is taking it from the best books and the psychologist, you know, it can't be giving dramatically bad advice. But if it was a serious issue, such as being in an abusive relationship or anything like that, one should never depend on an A I because they are, you know, experts for that. But I don't know if it'll help or not help, Jeff, but I do think that that it's not the problem right now that we're facing.

• 45:47Geoff Nielson?

  That's, it's fair. It's fair. And, you know, it gives me lots to think about, about, you know, what the future looks like and what we can do or not do here. Now, I show one of the things, you know, that makes you, I think, unique and really interesting in this space is you've got this in some ways uniquely global perspective about what's going on in A I, about what's going on in technology. I'm curious, having worked with governments and private organizations around the world,

• 46:19Geoff Nielson?

  what adoption and strategic patterns are you seeing in the way different, you know, different groups or different governments are approaching these challenges? And, you know, do you have any recommendations based on best practices that you've seen?

• 46:35Speaker 01

  What I have seen globally is a huge amount of interest in AI and data by all governments now. So we know that tens of countries now have national AI strategies. Even Indonesia just came up with a national AI strategy. France, US, Singapore, of course, have had a national AI strategy for some time. Canada, almost every country recognizes the importance of this technology in making its industries more competitive. Now, the question is, how are they approaching this? There are a couple of ways. One is you need the compute infrastructure. So they're building data centers and they're putting in the AI chips that they can get.

• 47:19Speaker 01

  They could get second generation or older versions of Nvidia, given the export bans, depending on which tier of country they are in, in the US framework. That's very important, having the ability to process AI. Because most people don't realize that this is important when you're trying to scale AI and store the data within your own country. The second thing is, you need talent. And they are now beginning to invest in AI talent. Of course, AI can also do AI now, which is, which is nice to see. But you experienced AI engineers are actually very, very valuable.

• 48:02Speaker 01

  So teaching them, mentoring them, basically even kind of attracting them from other countries, just as we saw Meta attract all of these top AI engineers from its competitors. That's the talent war is really, really there. The third thing is, as Singapore is doing, and all these three things, I'm giving the example of Singapore where I'm from, is how are you subsidizing your small, medium enterprises? Not just your big corporations that can earn it, that can actually use it, that can afford it, but your smaller businesses that often form the backbone of an economy where the majority of your citizens live.

• 48:45Speaker 01

  And in Singapore, there's a lot of subsidization of that. There's something called CTO as a service, where I M D A, which is a government agency, will literally give small companies a CTO part-time that will help them for free and go and understand how to use A I and which tools to use in their business to automate it, to become more productive, to grow, to innovate. And I think these are very, very important. And finally, the fourth thing is that they're coming up with regulations that are not too stringent, but are also still taking care of the risks. So in Singapore, we came up with the first A I risk guidelines, which were presented at the World Economic Forum.

• 49:29Speaker 01

  Now we have it for generative A I guidelines. Now we're doing assurance A I testing, which is how can you assure that the A I is safe? When you have these four pillars, infrastructure and connectivity, talent, democratization of access, especially for the small businesses, medium businesses, and then finally, a governance framework. The countries that are able to execute on this, because everybody can have a policy, who can execute on it systematically with discipline are the ones that will succeed. Because this is not easy. It's a long game. And there has to be a lot of delivery around it.

• 50:12Speaker 01

  If you don't do that, then you just have a lot of policies that nobody believes in. And then you have a few big players and a lot of trillionaires and billionaires, but it never trickles down anywhere else. And that's, that's, I think, what I really enjoy seeing, Jeff, is that in Asia, even in Africa, the chief data officer of Kenya is an amazing woman. You know, you see this recognition.

• 50:40Speaker 01

  You know, in Latin America, they're coming out with their own large language models that are tailored to their culture and to the the local traditions and to the local needs of Latin America. We're going to see this emergence of countries that have been left behind, that are now galloping, hopefully, forward and going to leapfrog and actually come much closer to the ranks of the advanced countries by using these four pillars systematically. So it's a huge opportunity and an exciting opportunity to let those people who were unfairly left out of it because of where they were born now have an opportunity to be part of the global economy.

• 51:23Geoff Nielson?

  It's it's really, really exciting and, you know, I'm you said it and I completely agree. You know, my perfect world isn't one where, you know, three companies just take trillions of dollars of wealth in AI. It's one where all this knowledge and all this, you know, power lives with the, you know, that that very broad middle of the economy. And, you know, it's interesting to me because I feel like so much of the narrative we hear lately is sort of anti-government in terms of, you know, less regulation, less support, you know, let the free market reign. And and your message is sort of the opposite here, that there's a really important role for government to play and that the winners economically are the ones that are going

• 52:07Geoff Nielson?

  to have more of that support. Is that is that kind of a fair summation?

• 52:12Speaker 01

  Well, also, I think the message is that there's some governments have very smart people. Some governments may be very bureaucratic, which is unfortunate, and maybe slowing down the wheels of this innovation. And then I see other governments in the Middle East, certainly in Singapore, certainly in some countries in Asia, where they are smart, they know what's going on, they have AI engineers, they themselves are AI educated, they have diverse teams. You know, we should not be patronizing and elitist or judgmental about people, period. There are some people in governments that are great, there are some people in technology that are great, there's some people amongst the poorest in villages across the world that are great.

• 52:55Speaker 01

  And and I think that we just need to give it a chance. But there's such a good competition now between cities and countries that I think most governments will begin to upgrade and reduce their own bureaucracy. Otherwise, it's very hard to be competitive. And I think the rhetoric that if you encounter a bureaucratic company, it will slow the country down. Bureaucratic, not, that doesn't mean it's not governing risk, but it's bureaucratic without cause. That is correct. So you want a country that encourages innovation and AI, but it does it in a responsible framework.

• 53:37Geoff Nielson?

  It's government as an enabler of what's going on here versus just slowing down.

• 53:43Speaker 01

  Exactly.

• 53:43Geoff Nielson?

  And it makes complete sense. Um, Aish, there's only one other question I want to ask you. It's something I ask everybody who, who, you know, I, I speak to about this, which is, are there any trends right now in technology or that people are talking about that you think are BS? Or that are just overblown right now? People are spending way much, way too much time talking about that are, are just kind of a distraction or that they're getting wrong?

• 54:10Speaker 01

  Actually, I don't. I think their timing may be not correct. A lot of people say that we will not have any jobs at all within the next year or two, AI can do everything. The issue is that over time, the jobs will evolve. Over time, AI will be doing more, but then we will be solving more problems. Certainly, sitting in Asia, Jeff, I can tell you that there are many problems in countries across Asia that need to be solved, from healthcare to infrastructure to security. And there will be more jobs related to using AI, business, finance for these jobs. But I think that.

• 54:52Speaker 01

  Most Silicon Valley people may underestimate that the large organizations, the large companies, the large governments, and the population at large, may not be ready for AI to come in and automate so much. They may not trust it as much. Just because you use ChatGPT extensively, and we all do, I chat with Claude and ChatGPT and Gemini and Perplexity all day long, doesn't mean that I would trust it completely to run my government or to run my army or anything like that. And within these companies, the data is not organized.

• 55:35Speaker 01

  So while ChatGPT and all may have taken all the public data, the data behind the firewalls is really hasn't been captured. And that's going to take some time. It's messy inside these companies. And the third thing is that AI cannot be as innovative and brainstorm like humans. And the reason is, I believe, that it does not have access to that data. It has access to public data, but when people sit around and brainstorm or when entrepreneurs go on a walk and see something or just imagine and are dreaming, they're not writing it down. And that may change as AI gets into our wearables

• 56:17Speaker 01

  and is recording everything, which has issues of its own for privacy. But for now, you know, things will take longer, I believe, than people suspect. And the timeline is what makes people nervous because it doesn't give them any breathing room to go and upscale or think about their kids or think about their retirement. And I really want every one of us to feel optimistic. And I'll end with this, that The World Economic Forum has a Future of Worlds Report 2025. And they interviewed 1,000 CEOs that together employ 14 million people across, you know, over 50 industries and countries. And they asked these CEOs, they said, what is the one major disruption

• 57:02Speaker 01

  that you are looking at in your industry? And they said, without doubt, it's AI and automation. And then they said, does this mean you're going to fire people? And they said, oh, God, no, we are actually looking for people. There's a huge skills mismatch right now in the economy. We are looking for people who are comfortable with digitization and AI assistance and AI enablement of the operations of a company and can work with it to actually take us into the next era or industry 4.0, as we call it. So for everyone listening and for you and me, whenever you hear of a gap like this, that's awesome for us, right?

• 57:45Speaker 01

  Because that means there's a gap and we can fill it by being open to it, by learning, by putting our hand out, by experimenting. That's the reality of the situation today. And I think we need to focus on today and the next day and the next day without kind of going down some pessimistic rabbit holes, you know, decades down the road. But the way to prevent that is to be consistently working on a risk framework, understanding AI, embracing it, being responsible and critical towards it. Because unless you use it, you're not going to appreciate that it needs to be controlled. And that's, that's the gap, that's the step we need to take in order to truly be able to use it for our own benefit.

• 58:31Geoff Nielson?

  I love that. So much to think about, you know, so much to do, frankly, to get ahead of this. Aisha, I wanted to say a big thank you for joining us on the show today. I really appreciated your insights.

• 58:43Speaker 01

  Thank you so much, Jeff. Thank you. It was a pleasure to be here.