Scroll down for full podcast transcript - click the ‘Show more’ arrow
While AI has the potential to augment the capabilities of workers by automating tasks, concerns persist regarding technological unemployment and the concentration of productivity gains.
As AI develops further, what are the trade-offs leaders will need to manage?
This is the full audio from a session at the Special Meeting on Global Collaboration, Growth and Energy for Development in Riyadh on 28 April, 2024. Watch it here:
Tiit Riisalo, Minister of Economic Affairs and Information Technology, Ministry of Economic Affairs and Information Technology of Estonia
Thomas L. Friedman, Columnist, Foreign Affairs, The New York Times
Paula Ingabire, Minister of Information Communication Technology and Innovation, Ministry of Information Communication Technology and Innovation of Rwanda
Øyvind Eriksen, President and Chief Executive Officer, Aker ASA
Hiroaki Kitano, Executive Deputy President; Chief Technical Officer; Chief Executive Officer, Sony Research, Sony Group Corporation
Abdullah AlSwaha, Minister of Communications and Information Technology, Ministry of Communications and Information Technology of Saudi Arabia
This episode is related to the Forum’s Special Meeting on Global Cooperation, Growth and Energy for Development held in Riyadh on 28-29 April 2024.
Check out all our podcasts on wef.ch/podcasts:
ポッドキャスト・トランスクリプト
This transcript has been generated using speech recognition software and may contain errors. Please check its accuracy against the audio.
Thomas L. Friedman, Foreign Affairs columnist, New York Time: Well welcome everybody. I'm I'm Tom Friedman. I'm the foreign affairs columnist for the New York Times. And I'm so honoured to be moderating this panel. You know, everybody's here. So, I'm going to do a quick introduction to the subject of AI and how it's going to be driving the future. I was telling the minister yesterday; I'm working on a new book right now and it's about, this moment in history, which I call a Promethean moment. If people could silence their phones, otherwise, there'll be a stun gun that hits you.
We're at a Promethean moment. Prometheus, the Greek deity who steals fire from a cabinet on Mount Olympus and gives it to humans to build civilization. And we know what these Promethean moments are. They're the printing press, the scientific revolution, the ag revolution, the industrial revolution and… this moment. Yeah. You are here for something really, really big. You know, when Gutenberg invented the printing press, some monk said to some priest: “Now, that is really cool. You mean I don't have to write this Bible out longhand anymore? I mean, we can just stamp them out?” Well, you are here at a similar Promethean moment, and the definition of a Promethean moment is that it requires you not just to change one thing. It actually requires you to change everything. How you learn, how you educate, how you do business, how you invest, how you govern, how you fight wars, how you commit crimes, how you stop crimes. We are in the middle of such a moment. But what's different about our Promethean moment is that it's not built on just a single technology, like a printing press or a combustion engine. It's built on actually two super cycles, and one is a technology supercycle. It's our ability to sense, digitize, connect, process, learn — now amplified with AI — share, connect. It's great. To sense, digitize, connect, process, learn with AI, share and act, and we're putting that loop into everything from your toaster to refrigerator, to your car, to your watch, to your F-35 fighter jet. And it is going to require us to change everything. And the countries and companies and communities that will do the best of taking advantage of our Promethean moment, which I call the age of acceleration, amplification and democratization. Never have more people had access to tools that amplify their individual power at an accelerating rate and are being democratized, with a small D to more people all the time. That's what this moment is about, and the countries and communities that will take advantage of this the best have a real similarity with ecosystems that survive when the climate changes, because we're in the middle of a giant climate change. The systems, the ecosystems that thrive when the climate changes have one thing in common: they have healthy interdependencies and are able to build complex, adaptive networks with all the elements of the ecosystem network together to maximize their productivity, resilience and adaptability. What nature does emergently — the deer eats the grass, the lion eats the deer and you end up with a healthy ecosystem — we need to do intentionally. In our societies build complex, adaptive coalitions between business, government, entrepreneurs, educators, the private sector. And that's what I'm always asking whenever I visit a community. So, to begin with, I want to ask Minister Swaha to kick it off. And I'm going to ask everyone the same first question to start with, and then we'll go another round. Minister, what's exciting you now about AI from Saudi Arabia's perspective? Then we're going to ask what's worrying you. But let's start with the excitement and the examples you can give. Thanks.
Abdullah AlSwaha, Minister of Communications and Information Technology, Government of Saudi Arabia: Optimism then realism. I love that.
Thomas L. Friedman: That's a good way to put it.
Abdullah AlSwaha: So why is this such an exciting time? Let's take a step back, Thomas. If we look at the broader context and look at global economic output, and I have to congratulate Rwanda for clocking 7% growth year over year last year when it comes to GDP growth. What's really driving the growth momentum? We have $120 [trillion] of economic output. $32 trillion is attributed to labour force output. Out of that $32 trillion, already, 1 trillion is being augmented, accelerated and democratized by AI/generative AI. Over the next 5 to 7 years, it's projected to go to 40%. That's 43% of the labour force productivity potentially could be accelerated, augmented or democratized and disrupted. And this is why we are not at a tipping point, but a turning point in humanity where we're moving from what I would coin, and I think the World Economic Forum soon is going to release it, shifting from the industrial revolutions to the intelligence revolution, how we're pivoting from digital to AI and maybe later on, quantum.
Thomas L. Friedman: It's interesting you make a distinction between sort of the information revolution, which we talked about to the intelligence revolution. I like that.
Abdullah AlSwaha: Absolutely. And right now it's all about achieving, first of all, co-intelligence — how humans and machine can coexist within this artificial intelligence era or AGI [Artificial General Intelligence] era. And this is where the Kingdom is very excited whether with Rwanda, with Sony, with Ericsson, with Estonia, with you. How we could create this complex coalition of adaptive forces to make sure that we push adoption. You spoke about the printing press. It was Germany that invented it. But the adoption took place among the British and the Dutch, and hence they enjoyed and they reaped the fruits. So we're pushing today an agenda that is inclusive, that is innovative and indisputably multi-stakeholder to make sure that we lead and leapfrog in this era. Yesterday I had the pleasure of introducing you to Doctor Ali Alhassan. Doctor Alhassan is a clear example of how this general-purpose technology can really make the difference for the most pressing challenges. I spoke about it last Davos and I shared with you the video.
Thomas L. Friedman: Yeah.
Abdullah AlSwaha: He used generative AI to look at the correlation of proteins and enzymes, to reduce the time for drug formulation, for sickle cell disease from ten years to two years. There's some great news. He's pivoting already to three other genetic diseases around vision, around respiratory systems and cholesterol level systems. These are remarkable examples of how this general purpose technology is making the difference for a disease that impacts 20 million folks, 75% in Africa, in the Middle East. And this is making the difference. We have learned a lot from this. The OECD principles of trustworthy AI. His Royal Highness led, under his leadership, the consensus among the G20 in 2020. This is the toughest year and this was three years ago, way ahead of the generative AI era and those OECD principles of trustworthy AI have paved the way for the AI act, for all of the different pieces of legislation that we're hearing today. And those principles today have not stifled innovation in the Kingdom. So we've pushed on direct formulation, and we have learned that on a use case basis, you have to do a risk based assessment out of which you still have to conform to clinical trials regulation. So right now, 15 patients in the National Guard hospital are undergoing this treatment, and they have to go through stage one examination for toxicity and for side effects. You then have to move to efficacy then to full approval. If I just give you one last example and then I can wrap up. Yeah, if you will shift to the other supercycle you spoke about, which is the climate change. We have Aramco that launched Last Leap, one of the largest tech events right now in the world. And I do invite you, Thomas, to come and join us for Leap next year. They have announced the Metabrain bring 250 billion parameters model, soon to be a trillion parameters, really helping them achieve the lowest uplift cost and carbon intensity for their downstream and upstream activities. And there's one particular use case that they're very excited about when it comes to drilling, and how they can reduce the numbers of runs as they correlate the porosity and the pressure points for drilling. And the savings that they have realised already is $1 billion per annum for the next ten years. And that's one use case.
Thomas L. Friedman: So just quickly, you also talked about how they've turned the drill from this direction also to this direction and taking stuff out of the atmosphere at scale, not just out of the sensors.
Abdullah AlSwaha: When it comes to metal organic frameworks for carbon sequestration and storage, how the circular carbon economy can work. So this is live examples of how we need to foster today in a multi-stakeholder fashion an inclusive and innovative AI that pushes the needle on this general purpose technology diffusion and adoption.
Thomas L. Friedman: And of course, what was different about when the printing press happened: yes, it diffused quickly to neighbouring states in Europe, but took a while to get to Saudi Arabia and to Rwanda. Now this happens at the exact same time.
Abdullah AlSwaha: Average is no longer the case today.
Thomas L. Friedman: Exactly, exactly. Mr. Erickson, what's exciting you at Erickson?
Øyvind Eriksen, President and Chief Executive Officer, Aker ASA: Well, I'm thinking about this for my role as industry leader. And, what's so fascinating just now is that we all realize that industries will change fundamentally. But we don't know exactly how and how fast. So, hence, it's so important to engage not only inside individual, business enterprises, but also with, other companies in the supply chain and in the network. The ultimate result will most likely be that more and more business enterprises will be fully automated. Helped by software, AI and robotics. But that will also not only impact individual companies, but entire value chains. And to understand how to set that up and what it takes to unlock the full potential of the technology is a huge and very important leadership task as we speak, which really excites me.
Thomas L. Friedman: You know, it reminds me of when electrification happened was introduced. There was actually no initial productivity boost. It actually took about 30 years, and there's a lot of study. Why is that? When we went, you know, from steam to electricity. And the reason was the full impact of productivity improvement from electrification actually required you to redesign everything from how the factory was organized, the workforce. And only then when you redesigned all the system is around it, then you got this explosion of productivity is the same you sort of are suggesting around AI?
Øyvind Eriksen: Sure. And in a number of, discussions like this it's easy to be excited about the potential of the technology. But we tend to forget about what it takes to realize the full potential and unlock, bottlenecks. Yes. So that's probably your next question.
Thomas L. Friedman: Yeah, that's what we will. We'll get to that. Minister Riisalo from Estonia. Give us your perspective.
Tiit Riisalo, Minister of Economic Affairs and Information Technology, Government of Estonia: Yes. I would say that I'm super excited and that's true. That's not just the exaggeration. And what is our plan? We plan to be super users. Super users of AI. And we have a plan in place. What we're going to build is, we call it a post-digital personalized state and we are convinced that we can do it because we just have not, you know, talked our talk, but walked our walk, last 25 years. We still have the most advanced digital society. And we see a huge amount of benefits becoming available from the from the latest technological developments.
Thomas L. Friedman: Pardon my ignorance, but how many people live in Estonia?
Tiit Riisalo: 1.5.
Thomas L. Friedman: 1.5 million people. You're making that up. I don't believe that.
Tiit Riisalo: You asked me to bring examples to describe what I mean. We have been working on AI issues already for some years. So it is not a surprise for us. And we have to remember always that before, you know, generative AI there was, you know, machine learning, deep learning. And these all are usable technologies. So just to bring you one example is that we have from 2020 up and running national clinical decision support tool, which means that it's a support tool for doctors to make better diagnosis. And what enables us is that we have our genome database established already in Tartu university in the 2000s, beginning of 2000s. So, it is a quarter of population having a full genome sample there. So we have digitalized the way for e-health systems already, you know, 15 years ago.
Thomas L. Friedman: For everyone in Estonia?
Tiit Riisalo: Absolutely everyone. We have a universal coverage. So there is in this e-health story, you have everything, you have your doctor visits, you have diagnosis, you have your analyses, you have do all the scans. And then from the third side, we have the so-called what, what in the industry call the life log — as you know all the service in Estonia are digital. There is a lot of data when you marry, when you divorce, when you get driving licenses and so on and so on. These are different data sets. But what's the beauty of Estonian system for the very beginning that we have this underlying protocol called X-road and data standards. So we can securely change the data, between different data sets. And if you put all this together and then now add the capacities of generative AI, you can only imagine what you can see in a in a field of personalized medicine.
Thomas L. Friedman: You have all this trusted data, you’re saying?
Tiit Riisalo: Yes, yes. And it has been a buzz word for some years already. But there is not really true results and results in a very specific group. But we are convinced, for example, in cancer diagnosis it is going to be completely different picture after two, three, four years. So this is what makes me really excited. And we have the possibility to create this world. And, you know, we are going to make mistakes also. And this is why it's important to have this kind of forum and talk with other countries and organizations who are like minded. To learn from each other's experiences and then you know, our purpose. You know, I'm not thinking so much about the trillions in the global GDP. We are thinking how to make it useful for our people. Yeah. I'm super excited. I see so many possibilities arising if we don't blow it up.
Thomas L. Friedman: We'll talk about that in a second. I still don't believe there are 1.5 million people only in Estonia. By Minister from Rwanda. What do you see that's exciting you?
Paula Ingabire, Minister of Information Communication Technology and Innovation, Government of Rwanda: Thank you, Tom. I think just like most of the other panellists, I share the optimism when it comes to AI, particularly for Africa or a country like Rwanda, for two reasons. One, the ability to be an equalizer. And especially for developing countries, the ability to leapfrog. One of the benefits that we have is that we don't have legacy infrastructure and systems. And so, if we're able to be very laser focussed on how we deploy AI solutions for the societal problems we're trying to solve for, then one, we gain the benefits, but also we're able to leapfrog when it comes to technological development. The second thing that excites me about AI is the uplifting effect that it has on the workforce, particularly for least-skilled professionals. You have the ability, and they stand to benefit the most when it comes to generative AI. What is challenging, though, is then to think about what's the wage gap that will persist? Yes, but at least what we continue to see is, for the least skilled professionals, you have the opportunity to leverage AI to also increase productivity in the different trades. And I'll give you a couple of examples in Rwanda. We have, similar to what the minister from Estonia was sharing. We have, universal healthcare insurance coverage for our citizens. And what we've also done, the structures that we've set up for delivering healthcare, primary healthcare services, have required that we deploy at least 100,000 community health workers. And these will go door to door in the communities providing primary healthcare services. Now, what we've done is to build a Kinyarwanda — this is our local language — large language model that allows these community health workers that when they're employed, they're not necessarily medical professionals, but they're able to draw information from a wide pool of data and be able to respond. Of course, it's very much linked with the medical record system that has already been put in place, but it still allows for quality healthcare services to be provided to the citizens. Now, when it comes to the enabling or equalizing effect. We are a population of close to 14 million people, and two years ago we were looking at the number of radiologists that we had were not more than 15. So in essence, you're looking at, you know, for one radiologist, 1 million population. So here is a great opportunity in how we leverage AI or AI models to really ensure that we can help these radiologists to be able to efficiently and productively deliver their work. We've done an economic impact assessment, and what we're seeing is that AI can contribute about 6% to our GDP growth. And we've looked at particular use cases in agriculture, where we're looking at early warning systems to support farmers. 70% of our population is within the agriculture sector. So, if we're able to build systems that allow them to have timely access to information and inputs.
Thomas L. Friedman: Can people turn off their phones or leave the room. Excuse me.
Paula Ingabire: No problem at all. So essentially, looking at all these different use cases in agriculture and delivering health care, in improving public service delivery, there's really immense opportunities on what AI can do to really help a country like Rwanda to leapfrog its development stages, but also at the same time, to really enhance the workforce.
Thomas L. Friedman: So, I was telling the minister yesterday, back in 2011, I wrote a book about America, it was called That Used to Be Us about how we were falling behind. A chapter in the book that was called Average Is Over. And at the time, the theme was that if you just have average skills now, that will not return an average lifestyle of income, of housing or whatever. And what excites me about AI right now, and for a country like Rwanda, is average is over in a different way. So those health care workers in the past, they never even had access to average. Now they'll have access to world class in their own dialect.
Paula Ingabire: Exactly.
Thomas L. Friedman: And I think as that scales it's going to be super, super exciting. From Sony.
Hiroaki Kitano, Executive Deputy President; Chief Technical Officer; Chief Executive Officer, Sony Research, Sony Group Corporation: I think this is a really exciting time. And I've been in this AI field for over 40 years, four decades. AI, robotics and, large scale computing and, you know, people are equidistance like industrial revolutions. And then, if you look at the, industrial revolutions they took two phases. One is steam engine, the other one is internal combustion engine. Right now, what I would say is a kind of excitement to get on AI is a steam engine. You know the best is yet to come. I mean we're going to see internal combustion engine quality of AI in coming years, over the decade. And right now, like, you know, gen AI is predominantly used for a productivity tool, particularly for like, you know, large scale language models and, you know, some partly like, individual models style. And, I think to have like a real impact on the GDP and the productivity and new industries, that's for sure. What I say, thing is the next phase of the AI development are going to be on the creativity tool, particularly for the AI, for scientific discovery. I think this is going to be a huge change, because our society is driven by the scientific discovery and then the technology is, you know, of course, a part of that. And, the way we do science is going to fundamentally change in the coming years or decades. Right now, like when you actually asked, you know, ChatGPT or whatever, a large language model, you get an answer. It is a decent answer. At the same time, what you need in a scientific discovery is, you know, for the bulk of our knowledge, we need to get, like, a number of hypotheses, right? Like, you know, if for this question, this could be the solution. This could be another one. This could be another one. I mean, you know, humans actually do- in science we do that. We actually — I'm a scientist myself — actually we’re not that good at that. Yeah. Like we can’t really come up with all possible hypotheses. We cannot do that. We come up with something we think is most likely. Yeah, but you know nature is more tricky. So we might have, like, something that that we haven't thought about. This could be the, you know, real thing, you know? So, we probably need to, move into the new kind of AI based on what we have to be able to generate a lot of hypotheses and prioritize it. And in the moving to the experiment for the verification, that means we need robotics. So, I think that’s going to be the loop of the new kind of AI combines with robotics. And it will change the way we do science on the biology, pharmacology. material science, engineering and design. And so a lot of things I think is going to be huge. And then I think there will be, you know, fundamental. Also like a similar principle applied to the more general creativity for like, you know, artists and musicians, because there, you know, creators will explore the new expressions, new music, new art form, new movie, right. So basically, it's exploring a hypothesis, not really getting the answer from the digital world. So as a web like, what we know, it's exploring new opportunities, new expressions. I think it both are common, but I think, that's something that excites me. And I think, I guess, that is the fundamental change.
Thomas L. Friedman: I mean, you're really describing the company that the minister actually introduced me to yesterday. It's exactly that kind of phenomena. Minister, what concerns you, though, now going forward. Whether it's about job loss or security authentication, what are the concerns and how are you addressing them?
Abdullah AlSwaha: So, let's talk a little bit about realism, but I'll still infuse a little bit of optimism, because I believe this general purpose technology, we spoke about this yesterday Thomas, is actually net positive on the long run. And if you look at the steam engine, the internet, they have been net positive on the long run. But the current model is overhyped because of a few limitations. The good news is that there are solutions that are being worked at, but these problematic and systematic issues are driving misinformation, bias, hallucination. And let me explain to you why this is coming from. First of all, let's appreciate the technological development. The current models are operating at around 4 trillion parameters. Just to explain this complex notion, in a very simplified way, we all have to be, by the way, if you want to appreciate artificial intelligence, you have to be a pseudo neuroscientist because we're trying to mimic neural networks within the brain. 4 trillion connections are the weights for how you discover, answers the human brain right now, approximately, because we still did not figure out everything about the human brain, is that 100 trillion parameters, 100 trillion connections. So that's 4%, right? But a few years ago, that's 175 billion parameters to 4 trillion. So that's 22 x in just less than three years at this rate, the most optimistic believe we can achieve general intelligence by five years. The pessimists believe by 14 years. But there is a catch. The current model is overhyped because there are three fundamental challenges. The first one from a great book, I’ll borrow from it, thinking fast and slow. We love that book. Humans have thinking two systems, thinking fast and slow. It's like teaching your kid math. If you want to teach him math, you will ask him what's one plus one? He will immediately answer two even without thinking. Yeah, but if you ask him, what's 22 times 70? He will take some rationale and some time. System two thinking. The current model has 4 trillion connections, wealth of data, but only has system one thinking it's only probabilistic, not deterministic. So that's an immature brain with close to general intelligence, a superintelligence, that can be problem number one. Problem number two: humans, neuroscientists they call perception and imagination constraint hallucination. So actually, for humans, when you go into sleep, the first phase of sleep you have something called rapid eye movement sleep, which is vivid dreaming. It's good for learning, for memory. So actually, constraint hallucinations is good. The current models for AI only have unconstrained hallucination, what neuroscientists called mental disorders. So, you have an immature brain, with close to general intelligence, with system one thinking probabilistic and with a mental disorder. Third problem is that if we want to oversimplify the human brain, it's really complex layers of limbic and cognition and survival systems and emotional systems. But really if we want to oversimplify it, it's three systems. The current AI model has only one reasoning inference. It lacks empathy. So, again, an immature brain with only system one thinking with a mental disorder called unconstrained hallucinations and biases and lacks empathy. The good news. There are solutions undergoing. But the call to action here is that we have some headroom to do some skilling and reskilling when it comes to transitioning to general intelligence. And let me reiterate this. This is something I shared in the AI summit during 2020. And I think I wasn't heard well. So let me make sure I repeat in this room for the folks in this room and over stream. Talents and leaders in this room and over stream. If you do not master AI within the next 5 to 7 years, you will become irrelevant to a talented leader that uses AI. In other words, you could be potentially but surely displaced or disrupted.
Thomas L. Friedman: Well said. Being a columnist is to do constrained hallucination. So I appreciate that. Mr Ericksen, from your point of view, what are your concerns these days, both for the, the general market of AI and also from your company in the employment and work, etcetera.
Øyvind Eriksen: So the multi-billion dollar question to me as a business leader is how to get value and how to get the returns from the significant investments in AI, both financial results but also environmental results. And when you do the root cause analysis, it's always a part of a question about people and how people are using or not using the technology available. And it starts with the leaders and AI and the potential of changing the way we operating our businesses can't be a separate initiative. It must be a part of the core strategy and the daily operation of the business, including how we're thinking about organizational development and changing of processes. But if people or our employees don't trust the technology they will never use it. And that leads me to the next fundamental enabler. And that's the quality of data. Yeah. And it was mentioned by His Excellency the risks related to the deployment of AI including hallucination. But if you're concerned about that in an industrial environment, the short answer is that you will not deploy the technology. So, a number of my colleagues, are not paying enough attention to the importance of a complete, high quality, verified set of data in order to get the results from the deployment of AI tools.
Thomas L. Friedman: Like, I know from the news business what trusted data is, you know, whether it's New York Times, Wall Street Journal, Financial Times. When you're an industrial company, what is the qualified, quality, trusted data you need to build a new product, when you talk about that?
Øyvind Eriksen: It starts with how we are designing and how we're building the facility and engineering data. It continues with how we operate the new facility including an enormous amount of real time data. And, you can add, the set of data from third party sources, in order to interact with the different parts of the value chain. And when we in my company started to think about this, we arrived at the conclusion that without solving the data problem, we would waste a lot of time and money on investment in digital tools, including AI. That's why we established a company called Cognite to build a data platform that can solve the data problem for industry. And today, that's one of the fastest growing industrial software companies in the world, with Aramco as a fellow shareholder and a joint venture called context here in the Kingdom.
Thomas L. Friedman: Thank you. So, Minister Riisalo, how does a country like Estonia with 1.5 million people, what are the challenges you have?
Tiit Riisalo: 1.50.
Thomas L. Friedman: I don't believe that. It's like we're just not going to believe that.
Tiit Riisalo: We have absolutely the same challenges and you know, as we all which is a consensus that, you know this whole new, beautiful intelligence revolution is based on data. And so, the core issue for us as a state or as a country is these two how to maintain the data set. We know how to do it. I think we- Just to give you a story, again. So, a little bit more than two years ago a nasty war started in Ukraine. And so, a group of people were sitting in Estonia in certain rooms and discussing how we could help. And then, you know, military people went on their track. And then I sit with the technology people. And then we discussed it a little bit and one of them said, or there was a discussion that what they go after and Estonians have the historic memory that we have been living in the same place approximately 4000 years, but in a geographically pretty volatile area. So, a lot of people have taken interests in our land. You know, at first the [unintelligible] and then the Danes, than Swedes, then [unintelligible], then the Russians. And when the when the new ones arrived, you know, what was the first place they always went? They went to the church not to pray there and not to burn it, but to get the church books, because church book was the source of data, source of truth. And then you destroy them and then you have a confusion. Now we are in the same situation. As a minister in IT my main concern there is the data, the data safety. But the story was that when the war started, we thought that this is what they are going after. And Ukraine and Estonia, we work a lot together, it is a digital country. And becoming really a major player after this war ends. We said that we have to save the data. So, we immediately started to, you know, bump over the data to Estonia and also move the servers. But it's like, you know, separate story. Now, now our concern is just how we can build our cybersecurity in a way and data safety that we really sort of maintain the trust of the people. This is the core issue in the in the coming years. And but of course, it's a little bit from the positive side also that what we can do yourself sort of to maintain the trust sort of fight, you know, the bad guys. But we already have up and running for some years an application called Data Tracer. Again, as our data is interchangeable between the different data sets, not only in the public sector, but private sector is in the same platform. So, Data Tracer is actually a system that when you log in into the state portal, you go to a sort of devoted room, a page, and you see what queries have been made for your data. And then you exactly know, who asked what. You know, why police asked this, why the people register has that. And so now we are creating a new version of our state portal mobile app. So, we're really, really focusing there to build a really good dashboard for people to follow their personal data. So, this is —some will opt in, some will opt out — so this is sort of building the trust in these two issues is a major challenge I guess. And yes, hallucination. You know, it happens. If you want to be in the forefront of technology, then you have to take certain risks.
Thomas L. Friedman: Wonderful. Minister from Rwanda, what are the biggest challenges and concerns you have?
Paula Ingabire: I would bucket them in three themes. One of them, and when we look at different countries and the AI policies and strategies everyone is putting in place, there are themes that are consistent across. There's talent, everyone is focussed on building the much needed AI literacy and talent that is required. The other one is compute infrastructure, making those investments. And then you also look at aspects of regulation. And I think this has always been a very tricky subject because the bigger question is should it even be regulated? And can you regulate what you don't understand? But I would say, starting with the one, I did allude to it earlier, which is really the wage gap that would persist largely because of the wedge depressing effects when you look at 90% of people globally depend on their income. And so, if we're talking about this up levelling, you know, effect, the question becomes then how do you create meaningful opportunities. A continent like Africa, where three quarters of people are under 30, where every policymaker across the board is thinking about creating jobs and meaningful jobs for these young people. When you're thinking about AI and the effect that it could have on those jobs, the question then becomes, what kind of intentional decisions do we need to make when we're upskilling and creating these opportunities so that we don't see a widening wage gap as we go forward. We just came out of the plenary session where the conversation also evolved around Africa having a growing middle class. But that's, again, it's also going to be something that we must think about as we think about how we are deploying AI. The second risk that I see is around the exacerbated digital divide when it comes to AI. 10% of countries globally own 100% of the compute capacity that we have. And so, if you have developing countries that don't own and have this compute infrastructure, it also means the benefits that we are all discussing here in many ways will not be able to achieve them fully and in an inclusive manner, as minister Swaha mentioned earlier. And so really, even thinking about the issue of data sets, how do we train models to have diverse and representative data sets so that we minimise the biases that could evidently come out, the insights that are coming out of these models that we are training. And then the last one for me, which again is discussed among different policymakers and regulators, is the issue of ethics. And the reality is that ethics and norms are not necessarily uniform across cultures. And ethics may not necessarily just be about what cannot be done but really, how do you enable human excellence? And so is it the standards that we are creating. Because ethics is really a moral obligation. And how do we weave that in in how we deploy responsible AI solutions. So I'd say those three things: the persisting wage gap, democratized access to infrastructure and ethics are some of the risks that I see. But also, to paint a picture that they are possible to solve for if we have, you know, the right intentions, the right people and thinking about it the same way in a matter that they're going to be inclusive because to the extent possible, you don't want certain parts of the world developing at the expense of others. And really thinking about this inclusivity agenda.
Thomas L. Friedman: Very helpful. From the nation of Sony?
Hiroaki Kitano: Yeah, I think AI have like, three issues. At the same time, I think AI has got opportunities as well. Two thing is in data. One is authenticity with data where the data come from. Okay. So, you know this gen AI, AI requires a lot of data. It’s driven by data. It's the fuel. That is then processed by the GPU cost. And then you know, if that data is- do they respect rights of the creators, the creators and writers and they are the ones who actually created the data. They need to be respected. And, you know, of course, that going gen AI a creator will use the tool to be even more creative. At the same time, creators need to be compensated properly and they need to be respected. Otherwise, like this cycle of high quality data and improvement in the AI capability will break down. So I think we need to have like a decent scheme how the creators need to be respected and compensated for that. Okay. The second part is that, I'm on the UN advisory board for this AI, and one of the big issues: unlike, you know, conventional software access to the AI doesn't really mean that you get the same benefit. English speaking countries in North America or, like going to Africa, Middle East and Southeast Asia — very different. You know, if the data you are interested in is not really there you don't get anything, right. You know, very low performance. So, like a fair and inclusive data set, also that respects the local culture and also the creators’ rights is mandatory to have like, you know, AI system that will be a benefit to everyone. So basically if you want to have a principle that no one is left behind, we need to establish that either in open source. Whatever it is needs to have like a, you know, a group of people, a group of governments and organizations, working together to create a common- this would be a global commons. You know, on the technology side, I think that's really important. Those two things in the data side and the third one, I fully agree with the Minister, what the Minister mentioned: current gen AI is actually taking only the fragment of the CNS (central neuro functions) like a re-engineer. Okay, so this is a really remarkable discovery for the group of Google scientists on this is transformer model. I mean this multi attention model. But at the same time, I think there are all the complex things, there other functions, all this architecture of the network, you know, or the brain. And we take you know, only a fragment of that. You know, when you talk about like all the scientific discovery, we need to be able to find the hypothesis. So, a hypothesis needs reasons and then the logic behind it. And in the last months I had a discussion with Yoshua Bengio. We fully agree that we need a new kind of architecture on top of what we have, or as an alternative an evolution of that. And I think, so that's actually the concern that at the same time means opportunity. I think we have like a much more powerful and a much more reliable AI in the coming decades and there is a lot of research going on in that direction.
Thomas L. Friedman: We have time for one question or maybe two, if anyone has, and for whoever, you'd like to pose it to. Any questions around? Yeah, please.
Audience member: How do we give the figurative microphone to the countries that don’t have the microphone to be heard for being included, and what are the practical steps to actually implement that?
Paula Ingabire: So I think even platforms like this is one way to make sure that we have representative voices where we are looking at those that, you know, may have some systems and foundations in place and those that are yet to even build those. That's one thing. The second thing is, who are the people that are at the forefront? Whether it's building talent, putting in place this infrastructure that is needed. Of course, governments need to be very open and agile enough to create the right environment. That will attract these kind of investments, whether it's for infrastructure and skills to happen. And so it's not a single player's responsibility, but one that requires multiple kinds of partnerships to happen. Now, the reality is that, and again I’ll reference the plenary session that we just came out where my president was speaking, he said one of the things that he highlighted is Africa getting out of this mentality where we feel like we're a victim. So, what are the things that we can also do for ourselves? And this is not just Africa. It's also every developing country, every developing part of the world. What are those things that you can do for yourself without necessarily waiting for help to come? And when help does come, you'll find when you've already put in place those kind of foundations. And so really, that's my take on how we make sure that every voice is heard. But also we have this inclusive and equitable benefits across the board.
Tiit Riisalo: If I could add a more optimism and positive note in this, if we do it right, you know, AI will be a great equalizer. We will adjust. You know, smart people have created for us a new interface between the men and the machine. You know, it's huge access to the global knowledge base and you don't need any more, you know, specific skills to be sort of the user, the high end user of this. And it's not any more science fiction. We just have to sort of work together to make it happen.
Thomas L. Friedman: Well, the clock just hit zero. You know, for me, the best indication of a panel is how many notes I take and my book is full. Thank you all very much. I learned so much, especially that Estonia with 150 million people. Thank you very much.
Daniel Dobrygowski and Bart Valkhof
2024年11月21日