Siliconsciousness: The Great and Good Disruptive Power of World Models

right here on the DSR network.

“And there's no better way to enjoy the DSR network than by becoming a member.”

Members enjoying ad-free listening experience, access to our discord community, exclusive content, early episode access, and more. Who's code at DSR 26 for a 25% off discount on sign up at the DSR network.com. That's code at DSR 26 at the DSR network.com/buy.

Thank you, and enjoy the show.

Welcome to Silicon Consciousness, the DSR network podcast focusing on the artificial intelligence revolution, politics, and policy. Hello and welcome to a special edition of Silicon Consciousness. As you know, each and every week, we take a look at the world of AI and related technologies. We've been doing this for a couple of years.

And throughout that time, I've been wanting to speak to Eric Sching.

Eric Sching is a professor and first president of the Mohammed bin Zayed University of Artificial Intelligence.

Professor Sching is a recipient of the National Science Foundation Career Award, the Alfred Sloan Research Fellowship and Computer Science, United States Air Force, Office of Scientific Research, Young Investigator Award, and the IBM Open Collaborative Research Faculty Award. He's also entrepreneur and business leader. He's a board member of the International Machine Learning Society. And we are delighted to have you here. Welcome, Eric.

Thank you very much, David, for the introduction. It's great to have you here. Well, it's great to have you here. Welcome to the U.S. I thought I'd divide our conversation into two halves, both of which deal with different interpretations of the term world model.

“I think in this, we've been talking a lot here at Silicon Consciousness about world models.”

So you know, we do a lot of work in partnership with the MIT Technology Review. I was speaking to the editor there a week or so ago, and he was reporting out of their big MTECH discussion on AI. And world models were a big source of conversation there. But before I get to the applied AI side of world models and some of what you're doing at the intersection of AI and bioscience, I'd like to talk a little bit about a different kind of world model, which is the model for countries as they go forward and deal with the new world of AI.

Because certainly the UAE has taken a different, very forward leaning approach on this. And the creation of the MBZ University of AI is right at the heart of that. And so I thought I would start with that as the origin question.

“I think you just graduated your fifth cohort of postgraduate students there, and the university is five years old.”

What was the original rationale and how was the rationale involving? Well, I like your analogy of discussing world model toward our countries scale and the MBZ global geopolitical scale. I think the university at its founding was out of a very simple ambition and vision, you know, from the leadership of UAE. That they want to diversify their economy beyond, you know, fields of oil economy, energy economy. They want to build the next foundation of their country and the society and the economy, not a driven economy.

And of course at that time we don't have tragedy yet, we don't have what we see the most fancy contemporary version of generative models. It's actually quite amazing, you know, for us now to retrospectively, you know, appreciate, you know, how come and the leader of UAE, the leaders of the country of UAE were able to identify AI as the future, you know, of society and the economy, you know, for the years to come. When I was asked to build this university, the vision was rather simple, there needs to be capacity both in R&D and also, you know, in talent, you know, to deploy and also to develop, you know, AI in this country.

Here are the mindset is that not only we want to adopt the best technology and the stand-in, but also to be more than a consumer, you know, of the best technology, but also a country future and a innovator, maybe even a leader in this technology because AI, unlike other things, it's all like you buy a car or buy a piece of hardware and you learn how to drive it and steer it and then it worked. AI connects to society to people much more closely than any other technology. Therefore, in order for AI, you know, or information science as a whole, to be truly integrated into a society's economy and a culture.

“You need to have the domestic capability of fine-tune adapt and also, you know, evolve, you know, ways the regional and local societal needs.”

I think that's probably in my understanding the incentive and also the early motivation, you know, of the UA leaders wanting to build this university.

Of course, you know, as the time go by, we also keep evolving this vision, we understand that, you know, AI is more than just AI itself as a discipline, say building models and the inventing algorithms. In fact, you can use AI to do many other things such as all the science management, you know, and maybe even healthcare, you know, can also be transformed in a ways new technology. Therefore, the current new vision of the university is to go beyond just a university of AI, but a comprehensive university, you know, taking a new look into how science and technology should evolve and innovate.

Where's the presence of AI as your daily tools? At now, of course, the situation that the UA was in five years ago, or even that it's in now, which is transitioning from being an economy that was dominated by the production of fossil fuels to one that is a leader and a technology like AI is unique.

But the situation that many countries find themselves in now is recognizing that we are entering a new era technologically and preparing for it, developing capacity to deal with it.

And in that respect, creating a university like this, which is unusual, unusual sort of center of excellence, kind of approach to this kind of thing, seems like a critical step if you want to lead and not just follow, but it also seems to have another dimension to it. Something when I talk, I spent a lot of my career dealing in emerging markets and when I talk to people in emerging markets about AI, they're afraid of being dominated by the US or by China, by a couple of big countries, and of having those countries own their data, own the technology, drive the future of their labor force, etc.

And so in that respect, this idea, and a central idea in the new world model using it in the geopolitical sense, is digital sovereignty.

“Can you talk a little bit about how creating universities and centers of excellence like yours, play into the idea of digital sovereignty?”

Yeah, I see sovereignty in two cents. When is this capability in the form of making products and coming up with solutions? Yeah, by using AI from abroad, you do need to submit your data, your sensitive information and other things across over the board, and may have less control over its security and so forth. That's a legitimate concern, all countries will have, and then you may want to build it by yourself, and therefore people can set up institutions, companies, or maybe relocate to foreign companies into the country and make that sovereignty production taking place.

“I think many countries, not including just UAE, is doing that, but maybe more profound, and the long-term vision of the sovereignty is this cultural transformation.”

That you want your society and your people to be able to master new technology to the point that you can reinvent it and push it forward in your own hands.

So, this is more than just a sovereignty in terms of a specific technology or product. It's the mindset, the confidence, and also that the power near in spirit, what I see from the people in UAE.

“There's another component of this, which I'd like to touch upon just briefly, because people obviously are aware of the news and they watch this conflict that has been unfolding in the region since February 28.”

Clearly, there's some components of developing AI capacity that are related to security and defense, but there are also components of developing AI capacity that create new needs in terms of what it is that's defended.

You know, data centers were potential target in this where they're main potential target, and I'm wondering how all that affects what you're doing and how you're thinking.

“Obviously, AI does play a role also in improving and upgrading defense technology. In fact, from the world that we see, we already see that the modern worlds are not fought in the same way as we see in ancient times.”

It's highly, highly, you know, effective fast and also information driven. So, no wonder, you know, everybody probably appreciate now that, you know, technology is like AI, you know, is going to play a more important role to make the country safer, to make also your defense more effective and powerful. In terms of the impact to the university, it's more at the level of an environmental change. Of course, as a public university, our goal is to do fundamental research and producing, you know, both, you know, fundamental technology and knowledge in AI and also to train AI specialists.

But, we are now, you know, entering a new era where this capability and this technology are getting more and more public appreciation and understanding. And the many technologies, you know, can, you know, have multiple ways of being adopted and disseminated, and that is up to, you know, the user and the developer, you know, in different organizations, including the defense organizations. So, university indirectly promoted the awareness and competence, you know, of such adoption by producing the talents and also that confidence of being able to catch up and the master and lead, you know, even the latest most difficult technology in the country of UAE.

So, from my observation in the past several weeks, the morale, you know, even within the campus has been super high because people really find their work, you know, to be at a frontier of, you know, global science and technology advancement. People get to see, you know, technology is now playing a important role, you know, in everyday life, including, you know, safety and the sovereignty.

“I think that kind of, you know, environmental, you know, change is positively impact the university's operation and also to build, you know, more unity and the commitment among the students and the faculty, you know, in their work.”

So, that's the most important impact. Secondly, you know, as a matter of fact, UAE is a very small country with a very limited size of population.

The needs of, you know, improving efficiency, you know, in government, you know, even corporate community to get better, you know, outcome and the productivity out of the people is becoming more obvious, you know, during the difficult time and the stress. I, you know, you know, is obviously one of these driving force, such as using the recent AI agent technology, war models, large language models and so to speak.

And this is much better appreciated in the country like UAE, where we are always in shortage, you know, of talented workers, or even, you know, all kinds of workers.

“So, that's another impact, you know, having an university, we are obviously under a higher expectation of delivering more and more, you know, well-trained people.”

So, I was very happy yesterday, you know, to, again, you know, graduate our fifth cohort of 150 or 140 graduate students. Most of which actually chose to stay in the country of UAE, in the past we have 80% of our students. After the graduate remain in the UAE for their career development.

I expect a number can only go higher in the future.

This podcast is underwritten in part by the US Embassy of the United Arab Emirates. It's editorial content is completely independent, and the views expressed are exclusively those of participating experts. It is presented live without editing. For further information about the UAE's efforts in the areas of artificial intelligence and technology, go to the website of the Embassy at www.ue-emBC.org and search for UAE-US Tech Cooperation. We thank them for their support, we thank everybody who is supporting this podcast for their support, and we look forward to it developing and growing.

“Over time, because the issue is so important.”

Well, I wanted to, I've just two more questions actually associated with this sort of geopolitical world model, and they're both related. They pick up on where you are.

Because AI is a kind of game changer, as one looks at economic potential, and the overall potential strength or influence of a country.

For the first couple thousand years of human history, Chinese economy was the biggest economy in the world, because the principle unit of economic production was a human being. And China was the most populous country in the world. What happened with the Industrial Revolution is it became people plus certain kinds of technology. As we move into the AI phase of technological revolution, the role of people contracts a bit. The AI becomes a big force multiplier. Small countries that have significant resources countries like the UAE places like Singapore.

Other countries that might have at one point been identified as middle powers in the global scheme of things. With the right kind of investment, AI technology, can see that not just as a force multiplier and defense sense, but as an economic potential multiplier, where they can compete with much larger countries or much more populous countries in new ways.

“Do you see that? Do you think that's a fair analysis?”

You are right on spot. I was about to say that yes indeed. We are in a transformation where there's an opportunity for a global reset of the leading force in the future economy and maybe even in knowledge creation. As you reflected in the past thousands of years, you will see the switch from the population driven economy by China towards an industrial driven economy led by the grip rhythm. Maybe earlier in other than small countries, in the outpower in the older countries with the new technology, but also on the other side.

The country who wants to break away from our traditions and take the opportunity to try new ideas to experiment in a new technology. Not only requires the availability of technology, but also a very very open and daring mindset to allow this technology to be experimented and also to be iterated fast. I think Yui is one of this unique in this position, that it is a very, very forward-looking and very open and also very, very focused. Yesterday in my speech at the graduation commencement ceremony, I actually highlighted in two points facing or two challenges in a modern university facing.

But if you look at it globally of high education and maybe culture as a whole, such a very, very candid and straightforward statement of the function for university and how it does its job is not actually easily communicated and accepted.

“The university is not as worldwide at doing many, many other things which distract them from focusing on doing the most important work which is not as creation.”

In Yui, we actually have a culture that is a very, very result driven, very, very meritocratic and also very, very agile and forward-looking.

And that creates more than just what a big capital and also a good geopolitical position is able to offer. It is actually giving you that kind of a culture proponent of trying new ideas and also deploying new technologies. And that's why not everywhere, it is so easy to build a new university and also get it on track and deliver results in very short period of time. I'm not saying that our job in Yui is anything easy, but it is at least possible. And I want to highlight to that difference, compared to many established economy, established environment where people tend to be less aggressive, more conservative in hanging on to the tradition.

“You know, maybe restricted by legacy and not able to move fast.”

So as I understand world models in a technological sense, they take us beyond the limitations of AI processing built on, say, language.

Because language can only take you so far in simulating an environment. And in a moment, I want to talk about some of what you're doing, for example, in biotech in world models in that regard, they illustrate it. Before we get to that, sticking with the initial theme, I'll give you an example of something that often doesn't factor into, you know, political scientists or economists world models.

But is analogous. And that is culture. And, you know, you made the point that this is a university that was started a couple years before people had ever heard a jet GPT.

But in the couple of years since we've heard of jet GPT, particularly in the United States, but also to some considerable degree in Europe, particularly in the sort of under 40 or under 30 generation. Something new and kind of unexpected has come up, and we talk about it a lot here. And that is AI skepticism. People are afraid of the change. They're afraid of labor dislocations. They're afraid of losing control of their data. They're afraid of, you know, how concentration of processing or compute a processing power or compute could lead to over concentrations of power.

And so there is a tension, and we see it, by the way, in the US, politically, you know, going into our midterm elections, AI is an issue, you know, nobody predicted it.

“But it's a real issue, whether it's, do you want a data center round or, how's it going to be regulated?”

And when you look at the world, different cultures are adopting AI more quickly than others. China is leader in this, but the UAE is a leader in this, and I think AI adoption in the UAE is, I mean, even for a small country, it's noteworthy, is twice the rate of the United States. And I'm just wondering, as you deal with new cohorts, are you discovering skepticism from international students that are there about AI? Is anything changing, do you feel at all what I'm talking about here? I kind of appreciate, you know, many of your points in terms of the AS skepticism.

In fact, you know, every new, disruptive technology, you know, come to the world, you have the same skepticism.

That people, you know, could have their own books, could people could wear their own way of their bibles, and therefore, of course, also job displacement, because the clinical first could not tell it and it find the jobs and so forth, right?

“But at the end of the day, you know, it may take longer, but eventually people realize that the technological environment advancement in, you know, have the ultimate benefit of a moving civilization forward.”

As long as you treat it as a technology, rather than as a religion or as a doctrine.

In here, I think AI is after a tool. If you view that as a tool, without submitting it, you know, to be a slave to a tool, you actually could write on top of it and make use of it the way you want, or maybe you just put it off and not using it.

What's happening or what is really, you know, underpinning in my opinion, the AI skepticism, is the fear that people may submit to AI.

And then letting go AI to be our master and also to come back to hummus, you know, to the point that we cannot even control it. Without realizing that AI is not basically, you know, a agentif creature that is not driven by a human stewardship before that, behind that.

In countries like China and maybe in UAE, I don't see that kind of sentiment viewing AI as a all-powerful supernatural kind of force.

“I believe the reason there is because the general public, the media and also the narrative is a relatively rational one.”

You know, people, you know, are able to see AI function positive and negative from all angle, and people are able to experiment with it, you know, and having their personal experiences and develop maybe a very diverse, but still, you know, evolving view about AI as a technology. In my own observation at NBC and also my engagement with government officials and the policy makers and business leaders in UAE, I have to say that they're understanding their literacy about AI as a tool as a technology is better and more rational than many of the people that are met in the West.

I don't know the reason behind that, but they're understanding about AI as a algorithm and as a model maybe very naive, very superficial, but it doesn't matter. Because they don't see AI as a monster or a god figure that is coming to us enough because you know, you can different people, you know, take a car differently, some you know it, some really view it as a tool, but it never, we're worried the car will come to hear them by itself, right.

“So I think in AI in UAE education of AI and also education of AI literacy at different levels is really in a place at a high priority.”

One of the first things I was asked to after I arrived at the UAE directly from Sheikh Ahmad, the president of UAE, then not the president yet, then the prompting of UAE was that I was asked to develop a executive training program for the ministers and for the sea level officers of the country. So that we can tell them what really AI really is and how to, you know, use an understanding methodology. So then I was also asked to develop similar programs to, you know, middle level officers and the leaders, then all the way down to even K-12 kids and elementary school education.

So there is an emphasize, you know, of AI literacy education at a country level to all the population. I don't see that kind of appreciation and emphasize of education anywhere else in the world. So I was talking to, well, I'm talking about really leaders themselves who play with AI robots themselves who ask for our progress in the university directly in terms of performance of models and egg was and so forth. I was actually pretty amazed to be honest by the level of understanding and also the changing of understanding, you know, along the way as technology evolved from various various senior leaders in the country to work this technology.

the secondary, tertiary kind of information, second-hand information, you know, about what AI is.

So I think the way to really reconcile, you know, this kind of rather irrational and maybe all attitude towards AI technology is education.

“You need to let people understand AI and also to use AI, the more you use it, the more you become comfortable and adapt it to it.”

And then you also realize that it is not that great yet. In fact, there are a lot of things that AI cannot even do.

It's definitely not at a point that you need to worry that you are already threatened, you know, by such technology regarding your life and your job. Yeah, it's interesting, you know, culturally and, you know, being involved in this field for a few years. In the United States, there is a, I don't know, there is a cultural predisposition to want to focus on whatever is biggest and fastest.

“If something is being developed, the first question becomes, how do we win? And how do we win that race? And it's see it as zero sum. And with AI, that has led to what I think is a over emphasis on things like AI.”

So when I look elsewhere, what I look at China, for example, the focus is on applied AI. And where I see the rapid growth, it's on applied AI. And when, and that's not to say something less than it is very creative. And that brings me to this second use of world models, because you have training both in biological sciences as well as computer sciences, you're running university or also developing some ideas in a business context. One of the ones that I find most interesting is this idea of a world model at a cellular level that allows one to create essentially a simulation that not only has is much richer in terms of how it shows the interactions chemically and otherwise to take place, but also over time and other factors.

“And this is an example of what I see as the genius and promise of applied AI and biological sciences. And I was hoping that before we wrapped up here, you could talk a little bit about that.”

So maybe I should also start from what is a world model at its core, because even the world model itself has very, very many different definitions and incarnations around the world.

In fact, many people who are building video games and building video generation engines are calling their assistant a world model. And obviously, this is way off the mark, I have to say. Recently, I had a debate with the Hong Kong about how to build world models. One thing we do agree is that the world model is not about video generation, or doing a little bit better than a large language model in terms of a predicting the next. Right, world model fundamentally is different from an LGBT type of approach in a sense that it is now a generator of possibilities of your environment and your world.

And we call it physical intelligence, we call it embody intelligence, but the manifestation of all these intelligence is not about giving you a single next token prediction. It is about simulating what would happen after you take action on it. So therefore, we emphasize on action conditioning and also long term simulation, meaning simulating for not just a few minutes, simulating for a few weeks, days, years. Imagine that you cannot simulate a few years with videos, it's impossible to do that kind of simulation, not only computationally, but also the consistency everything will be lost if you follow the other definition lacking L and therefore, world model is fundamentally a different beast.

It's like reproducing our mental world, you know, in our brain, we have a world model which allows us to do thought experiment, to do simulation or thinking, therefore, when we plan and then reason, we're not trying to do calculus to solve a optimization using linear algebra. But in different ideas, in the thought experiment, what if this happens, what if I do multiple steps and approximately come up with multiple outcomes and then score them separately and then jump to a solution.

And then, you talk about what world you want to simulate. When you simulate a go player in a go board, you actually have a go. It's actually a world model playing the go.

“But if you say, my data is going to be biological data, say, the cell morphology in the form of image, but also it's a molecular content in terms of protein structure and the molecular interaction.”

And the molecular sequence are in a DNA protein. All of a sudden, you have a biological word encapsulated in such a world model. And then what it does is to simulate the behavior of a cell in response to your perturbation and action.

In that case, you apply a medicine, you change the temperature of the environment. All you have some shock in the cell in the form of chemicals and other physical beings.

All of these can be really taken into consideration as a form of a prompt as we use in a world model space. And then you're a cell would do this virtual simulation of all the possibilities that could happen to that, including, for example, the change of the shape, the change of the longevity, the healthiness of the conditions and so forth.

“That's why I see world model as a perfect, you know, an implementation of what people are looking for as a virtual cell because why you first of all need a virtual cell.”

It's because one of the words of virtual cell to replace the real cell to do experiments. In real cell, you inject a piece of virus or you inject a piece of antibody and see whether your design worked, which in my opinion is very expensive, but also very risky.

But if you know, this physical, you know, a material escape your laboratory situation and make create an epidemic and make cause harm to the experimental surface.

And now we say that we use the world model to allow virtual trial and virtual experimentation.

“So that's why, you know, waste the virtual cell, you can turn what was originally a very expensive, slow and dangerous physical experimentation.”

That biology is purely due into a digital experimentation. And this is actually not a new idea, by the way. You know, if you look at nowadays, how people manufacture silicon and the chips, how people manufacture, you know, or build atomic, you know, power plants and so forth, they all started from simulation and computational experimentations. Right. So we basically just want to bring this idea not into the more complex situation of biological science, where hopefully by training the world model without even knowing all the rules, you know, governing the biological world.

You can use a data driven approach to begin with a simulation, you know, of the phenomenon. This is like without knowing the rule of physics, you can already understand how, you know, flew it, you know, behave in some environment. You can actually, you know, you know, steer in aircraft, you know, by, you know, playing ways that the, the, the, that's for I simulate her, you know, in a sense it is a more data driven experience based kind of, you know, outlet, you know, of many capabilities in real world without, you know, knowing all the physics and the theories behind it.

And creating models like this has a lot of possibilities, one is generalized models. But as we have increasing compute and we have increasing amount of data, it also becomes possible as I understand it to create personalized models, so that one could create a world model for my cellular struck my DNA and test how I personally might react to a medicine or to a virus or to some other change in conditions. And so you, you know, doctors in the future will be better able to tailor treatments to individual people.

And, you know, everybody, you know, the average person is listening to this, watch as a TV and there's, there's a commercial, there's at the end of a commercial for a drug, it says, and here are 20 different side effects that could exist. Because we don't know, but if you can gather the data, you can build good models.

Because, you know, nowadays we have the human genome project becoming popular, it is possible not to sequence every individual genome and store them in a database.

“And why would we do that? Because every individual is different and therefore, by having their personal genome, you know, in a database, you could go preemptively check for genetic risks,”

predisposations of diseases and so on. And it is doable because you actually can afford to take a personal genome and put them in a database.

Now, coming to the cell, it is very difficult to collect every individual cell and in different tissue and organ situation.

And then build a physical copy of that in a physical environment, you know, throw them into a, you know, a liquid and then store them somewhere and later to retrieve them and do experimentation.

“It's a physically impossible to have such a vast kind of a collection of materials for every individual, not alone, the logistic difficulties and all that.”

Now, but you clearly specify a needle, right? It is nice to have because what if, you know, you have a certain situation that requires the testing of a medicine, you know, on you, we should begin with, you know, a cellular level, maybe tissue level experimentation. Having a virtual cell, if available in the digital form, you actually can legitimately and practically build a what we call a virtual cell bank for all the populations in the ward without even, you know, you know, you know, stressing too much about the resource requirement. In fact, one of the, I'm going to publish a paper in fact in a few days, which is called the virtual cell word model where we actually outlined a vision of what we call a virtual cell bank.

So that allow, you know, a personalized copy, you know, of a virtual cell to be quickly trained and then stored in a digital form. But of course, when not there, there are a number of steps, you know, requiring technological innovation. First of all, even the base virtual cell itself is just a beginning. We need to make sure that the word model of virtual cell should be good enough to first cover what is already in common for other populations. We are probably 99% similar. And then, you know, on top of that, how to forge from it and then personalize that, you know, base model, base virtual cell model into a personal cell requires a foundry of the virtual cell so that they can be industrialized mass produced into many populations rather than every individual take a month almost like a new scientific project to be reviewed, that's actually what's happening now.

So there are new technologies now in the ASB called AIHannis where we use AI to build new AI models, basically much of the reproduction and the extrapolation of base model into personal model can also be automated if you have AI2 that can do the optimal work.

All these are right now still in the early stage. I would say this is a whole new paradigm of war models, not just in physical intelligence, but also in biological simulation.

It's just at its starting point. I see a lot of new opportunities, including startups and also academic research to be to be coming up in the next couple of years. Yeah, you know, I don't want to be naive. We're not naive. We, you know, we are immersed in this world and our listeners are immersed in this world.

“And when we talk about these things, we have to acknowledge that they raise issues about privacy, issues about security, issues about regulation, but that's what happens when change comes.”

And rather than thing where we can't have change because we need to address new kinds of laws and so forth. I think we need to say, well, these changes create great opportunities. Now let's grapple with the appropriate way to deal with these other concerns. But, you know, one of the concerns that a lot of people have is labor dislocation because they can't envision what the next generation job is. Young lawyers, as well, I would have spent seven years researching and now AI already can do a lot of that.

My, my future is dark, but in listening to you talk about this, and this is a good point to end this conversation on as any, although I'd be honest with you. I don't really want to end it. I'm really finding it fascinating.

And I've always felt to be honest with you that calling it political science was, you know, a bit of a laugh because, you know, what a social scientist did in that field and any others is that they would try to come up with an eight variable model to, to recreate the world. And, of course, it was absurd. And, and so, you know, I mean, having an eight variable model is better than not having any model at all, but we are entering an era with ubiquitous sensing with massive growth in big data with massive amounts of compute with AI driven processing, where there is no field that will not be altered just.

“By, by this issue of simulations alone, that, that having the ability to envision conceived developed models and apply them in every aspect of life is, is a field that didn't exist before, but whether you're dealing with.”

If you're playing or planning a battle or a financial strategy or trying to, you know, cure, contain disease. This is, this is, this is new. Nobody did this before, at least not in this way. And so, all you have to do is extrapolate it out to, you know, dozens of other applications and you do see some promise.

This is a good place thing. First of all, tell me, I'm wrong. If you think I'm wrong.

“No, no, I like to chime in. You made the action points. I think that two points I want to really, really echo and maybe also a dive in a bit deeper.”

You mentioned about this social scientist using a parameter model, you know, to describe the universe, and so forth. Yeah, this is indeed, you know, a, a, in shrime the tradition, you know, in science. You know, the countest, you know, always believe that, you know, we can abstract a way, you know, a principles and the laws from complex words and being rational and using raising as a first in his and two basically, you know, govern the entire universe. Um, I think, you know, this may not be, you know, matching what is happening. First of all, um, if you look at the philosophy debate between countries and the humanism, even back in three or 400 years ago, there was already objection, you know, to this so called law based way of dealing with complex words because, you know, data itself has chosen it.

And different people may read the data or choose, you know, differently from the data, but why we rely on laws and we are so obsessed with, you know, very simple and the beautiful principles and laws.

But it's because I can be honest with you, we didn't have a computer, we have a small brain that we cannot remember all the data. Therefore, we want to have abstraction. Then we say, you know, those who can abstract better as matter and they really are becoming, you know, great. But more than what is very different, storage of data is actually very, very cheap. In fact, you could imagine store all data you ever have in the world, you know, on a chip very easily. I once asked my colleague, if you can store all the data, you have to derive your law versus your store, just the law you derive, which one do you prefer, you want to keep the law you bring or you want to keep all the data you bring.

Many people still like the law. I say, I like the data because maybe one day I would extract new laws all of it. Right. So we're in this era that data itself is the knowledge. I would love to interject when they hear it. The other thing is most analysis is based on analysis of correlation. The more data you have, the more correlations you will see, you will see correlations that didn't exist exactly in your limited model before exactly.

“I think a correlation is a better term than causation, even by the way, who, you know, to describe, basically the dependencies between phenomena and facts and other things.”

I think we are now re-entering an era of kineticism, where you have large data, your build models out of large data that can simulate basically what is in the data without even telling the law inside it.

So we have to basically embrace this period and take advantage of the fact that we can really get things out of a large data. That's one point one to make on your aid parameter analogy.

Another thing I really want to bring up about the job displacements is a following.

But if he's actually bad for that person and also for society, if all people are getting mentally mutilated.

“Therefore, you know, education still is very important for people not to just focus on the skill you need to survive or find a job, but actually the ability to learn.”

And this is more important now in the era because skill definition will be changing. Many of the skills that we can certificate in the past, like calculations and the wrapping complex math and now programming, you know, I will tell you it will be easily handled by AI systems, even now another long version in the future.

So many of the older skills that we, you know, cherish, you know, and spend time to learn. In fact, we'll be, you know, pretty trivialized by machines.

“But human agency is a different thing, you know, there are a lot of other dimensions that we just didn't emphasize or play up in the past, for example, you know, asking the right question and defining new spaces for exploration.”

Also, you know, how shall I say, you know, collaborating and, you know, you know, you know, you know, you know, how to communicate the world and idea to people and how to break down complex problems, you know, into partitions of connected, you know, units that can be, you know, you know, used to incentivize different teamwork and so forth. None of this actually is built in the machine intelligence. It's actually closer to how people, you know, can appreciate the complexity of nature.

And that part of the capabilities right now is not even what taught, you know, in university.

So, in fact, at my university, what trying to bring humanity science and the social science in a new way back to the STEM area, so that today, what they see is not just the equations, you know, or documents, you know, or programs. But also, you know, the deeper human agency behind it, you know, so that, you know, you can, you know, you know, be comforted by at least to learn new things and then to try new ideas out of it. I love that, and I really think, you know, as I listen to you, that one thing you might want to do, somebody needs to do, is to bring together the non-AI community and help them simulate the implications of AI for their worlds, not to be afraid of it, but to see it.

But to imagine how it's different, because in many ways, we are at one of those massive intellectual thresholds akin to when people tried to practice medicine before they knew anatomy, or people tried to understand the universe before they could see the stars. And what we are doing is we are taking multiple leaps like that into an era in which we now can begin to combine multiple sources of knowledge in multiple ways. The power is of world models is it gets us beyond LLMs, and it allows us to begin to take in the other kinds of data that we've got and understand the intersection, the disconnect, the correlations between all of those areas.

That's brand new, but I guarantee you that if you go to the sociology department at the University of wherever, all they are right now is afraid, they don't see how this opens things up for them. And so to my mind, a university like yours and frankly, all of us who are doing this need to help people see that.

“So I like what you said, I think the afraid is always due to uncertainty or unknownness, right?”

So I think having a simulation or engine that is able to display and explore all different possibilities, just in front of your eye, is one way of hopefully mitigating some of these anxieties.

Well, we'll keep doing what we're doing, you keep doing what you're doing, it's so exciting, something I hope to be able to go and see the university.

“But for now, I love to have you on campus and we kind of have, I can show a couple of other ideas, but I love to have you out here and see what's unfolding in front of your eye.”

I would like, I would like that very much for now.

I know you've been traveling all night, get some rest, congratulations on the latest class.

“And don't be surprised if in a couple of months we try to invite you back because I found this absolutely fascinating.”

Thank you very much. Thank you. Thank you.

“All right, folks, next week, we'll be more of this, although I'm not sure we'll get quite to this level, so please join us again for that until then, bye bye.”

This was Siliconjustness, a production of the DSR network.