The Urgent Need for Computational Thinking in an AI World with Conrad Wolfram

In this episode of Pulse, host Aldo de Pape sits down with Conrad Wolfram, Mathematician, Technologist and Author of 'The Math(s) Fix: An Education Blueprint for the AI Age'. 

Aldo and Conrad explore how AI is transforming education, computation, and society itself. From the failures of traditional maths curricula to the urgent need for “computational literacy,” Conrad explains why schools are still preparing students for a world that no longer exists.

The conversation dives into AI hype cycles, Wolfram Alpha’s early role in artificial intelligence, human versus machine thinking, and why adapting education may be the most important challenge of our time.


Find out more about Conrad Wolfram's book here: https://www.amazon.co.uk/Math-Fix-Education-Blueprint-Wolfram/dp/1579550290

Watch this episode on YouTube: https://www.youtube.com/@AlphaWireHQ

This episode was brought together by AlphaWire: https://alphawire.xyz/

Transcript

SPEAKER_01 0:00

If I want to start a company that will get somebody a better score in their maths test, pre-college maths test, I can probably get good funding if I get them 1% better score on average in their tests, right? The problem I have is that I think they're completely the wrong tests and they're testing the wrong things, and we're educating people in the wrong stuff. So I if if I want to change that, every single thing basically is against me.

SPEAKER_00 0:25

Welcome to Pulse by AlphaWire, the podcast where science and education meet cutting-edge technology and artificial intelligence. My name is Aldo the Pop, and each week I sit down with innovators, thinkers, and doers who are working to change our world for the better. Together we explore their journeys, uncover the lessons they've learned, and take the entrepreneurial pulse that drives them on their path to success. AI is changing almost everything. But are our schools, universities, and ways of thinking changing fast enough? In this pulse episode, I'm joined by Conrad Wolfram, mathematician, technologist, CEO and European co-founder of Wolfram, and one of the leading thinkers on computational education to explore how AI is transforming not just technology, but the very way we learn. We talk about why traditional maths education may be overdue for reinvention, what computational literacy actually means, and how tools like Wolfram Alpha helped pave the way for today's AI revolution. We also discuss why the future may be less about humans competing with machines and more about learning how to think alongside them. Sincerely hope that you enjoyed this episode with Conrad Wolfer. Yes, and we are live with yet another episode of Pulse. It's a Monday morning, and often Monday mornings are a bit tedious because we're recovering from beautiful weekends, but this Monday for me is beautiful because I get to speak to an absolute expert on mathematics and maths education, Conrad Wolfram. A very warm welcome to you. Thanks very much for having me. Conrad, we need to be honest with our audience. This is not our first interview. I had the great pleasure of speaking to you just when your book, The Maths Fix, was published, and was such a great interview because you dove into how AI is going to change the way we educate each other, specifically in mathematics. I remember us talking and really taking a deep dive, but this was all before Chat GPT or Claude or any of those tools, any of those mainstream AI tools. So I recall it coming across towards me like it's still a bit abstract. I get what you're saying, but it's still very abstract. Lo and behold, just four or five years later, we're in this completely new era almost. And I thought, well, this is a great opportunity for us to come back together again and talk about all these things that have happened in such a short period of time. So I guess that's going to be my first question. In your book, do you feel that things have come to fruition the way you you thought it would be?

SPEAKER_01 3:36

Yeah, I I agree with most of what I wrote still. So that's always a good start. And I do think it was to some extent prophetic of how, but I obviously had no idea at that moment of the exact way in which things would unfold in terms of you know new AI era and how that suddenly sort of hit, you know, in recent times. Has it come to fruition in terms of have countries changed their educational setup? No, they really haven't. And that's a little shocking. And I think the pressure has mounted yet more because of AI, the new AI we we just talked of. But in fact, the impetus for the change I talk about of fixing maths really goes back decades. And in a sense, what AI is doing is explaining that we'll have a similar kind of metamorphosis we need in many other subjects as well as maths itself. And maths should have been the forerunners that we still haven't changed that in terms of the subject. So we got an awful lot of work to do in education.

SPEAKER_00 4:40

Yeah. As of now, is our job to catch up with AI, or is it going to be to reject AI, to make sure it doesn't drip in into education and doesn't kind of take over from our brain? Because there are loads of scary notions. There's loads of, you know, room for optimism as well, but I also hear kind of the scary end of it. Where are you on that spectrum?

SPEAKER_01 5:02

Look, broadly optimistic, but it's not going to be a plain sailing. Okay. But there's a lot to unpack in in what you were asking there, because at one level, you know, uh it's like, do we want to you know not have AI in education? Tough luck. The world has changed. So that's not going to be an option.

SPEAKER_00 5:20

Yeah.

SPEAKER_01 5:21

Second thing is we got to really distinguish between the subject matter that we're covering in education and the pedagogy, how we deliver that subject matter. And this gets confused literally all the time. In every educational conference I go to or anything like that, people get confused between these two. If you take maths, so just go back to maths itself and forget AI for a second. My argument's been that for many decades now, the computer has taken over the calculating piece of mathematics. So in the outside world, basically people don't calculate by hand anymore. I mean, there are some edge cases you can cite, but fundamentally, and maths, by the way, allowed maths to move massively forward. We used maths in all sorts of areas we never used it in before. Medicine, biosciences, for data science, all these things which we could never have imagined before mechanized computing. So what happened there was the machinery changed fundamentally, and it changed so fundamentally that the subject of mathematics changed. And in fact, it became a much more general subject that was required across much of life, in a way it wasn't before. In education, we don't seem to have noticed that. So if you look at the curricula now across the world, they are pretty similar with mild adjustment to what they were 50 years ago, or even a hundred years ago in some cases. I mean, they've been adjusted. So that is a key that the subject matter is what you've first got to look at. What is it that people need to actually do in their lives? Now, we don't know the effect of AI yet on people's lives out there because it's only just hit. So it's all evolving. I mean, I just like this this week in my company, I've got meetings about okay, how are we going to change our workflows of what we're doing with AI? We don't know yet.

SPEAKER_00 7:14

Yeah.

SPEAKER_01 7:15

Right? So it's hard to figure out what we're trying to educate people for exactly. But what we do know is the mathematics has changed decades ago, and we haven't yet changed the subjects in education. So we've got to change how we evolve education. The other level of that, which we the pedagogy, it's also true that modern technology, including AI, changes how we learn on top of that. So when we have the new subjects, when we understand what it is we're learning, we then should evolve also how we do it. We don't necessarily need human teachers doing exactly the same as they were doing before. I do think we need human teachers because we have humans to educate, but I think there's a lot that can be changed in the process by which we do that. So that's kind of what I what I see. And will humans still be needed? I think so. But that really depends on making sure our education steps up to the next level. If we just go and educate people like their AIs, yep, well, that's what we'll end up with, and they'll they'll lose against the the uh computer-based AI.

SPEAKER_00 8:15

Yeah. Is it what I always wonder in these discussions, and I, you know, I I completely underline what you're saying, but I always wonder in these discussions that it sometimes sounds like, okay, humans versus the machines almost, right? So even though there's loads of work to be done in maths education, we've also made loads of strides. As you say, mathematics is this integral part of computer science that has allowed us to do so many more things just because we saw the scale and the potential of it, and we kind of knew how to grow certain things, which has also led to these machines being here now today with us. So it's almost like we've done well on a certain level, so it's not all doom and gloom. And for that matter, how do we kind of position ourselves from as human beings that need to get better in mathematics just because for our brain's sake versus the machines that have also been developed by human beings?

SPEAKER_01 9:10

Yeah, yeah. And as you correctly point out, maths has been the driver behind, you know, building machines that can do these fantastic things. And it's so it's been unbelievably successful and beyond all human imagination. You can look back to previous industrial revolutions, and it's quite instructive to do that a little bit. I mean, everyone is different. And this one is quintessentially human in the sense that it's previous industrial revolutions were, you know, machinery that did things, I mean, agricultural or in the 19th century, of course, machines to make things in all sorts of ways and transport us. Those, of course, were mostly physical in style, whereas now the machinery coming is mostly sort of mental. It feels very like that was the thing that marked humans out always against animals. So it feels very much like it's getting at us as humans. Again, if you look at some of the stuff in the 19th century, one of the things that really helped countries that did well there was having mass literacy. So the countries that were ahead on everybody reading and writing, or a reasonable fraction of the population doing that, that sort of backed that early, were better equipped for what came later because the people then could go up to the next level better. So Britain was reasonably well ahead on that, for example, uh, and did pretty well in the Industrial Revolution. You know, that was one of the reasons for that, I think. Now, one of the things I talk about now is the need to have computational literacy instead of just literacy.

SPEAKER_00 10:41

Yeah.

SPEAKER_01 10:41

And I feel like it's a very similar kind of game, which is that, as you say, we can have humans just compete with what AIs are going to do well, or what computers have been doing well in maths, but mostly they'll just do badly at that. So what we need is humans to be in charge. There, it's a bit like management versus employees, right? You want the human as the management and the AIs as the employees, not the other way around.

SPEAKER_00 11:06

Yeah.

SPEAKER_01 11:07

And the question is, how do we get people set up for that? Well, one of the ways is they need to understand much more about how computational thinking works, or as I put it, computational literacy, so that they can know what pattern of ways to make decisions and how they can help to get the machinery to get them to do that. And the machinery may help them more with that than it has in the past, but that's a pattern of thinking that we're not really backing in mainstream education. And that's partly what's caused, in my view, this bifurcation between the sort of digital haves and the digital have-nots is that we've got a small echelon at the top of society who kind of understand this stuff and can make it work for them. We've got most of the rest of society who are kind of completely confused about how all these decisions are made based on computational things, models and data science and so forth, and they're not educated in that. And so my view is that's a very important part of this picture to fix. In the same way, and just to briefly, you know, to go back to as I understand the history in the late 18th century, there was starting to be a movement that mass literacy was important. And there were people who said that, you know, that's nonsense. You know, you can't get most of the population to read and write, they're too dumb or whatever it is. We just have a few aristocrats and priests and people like that who know to tell everyone what to do. And in fact, of course, been incredibly empowering to have most people be able to read and write because they can actually look at stuff directly, do stuff directly. That's what's driven our economies forward. And so I feel the same can happen now. But countries and people in charge of the curricula really need to get their act together on this. And the AI has really driven that agenda up.

unknown 12:47

Yeah.

SPEAKER_00 12:47

I like what you said about computational literacy. Am I correct to understand this is something like knowing how to use AI? So knowing tools and just being familiar with them, or would you also mean kind of understanding the back end, understanding how to build such an AI model?

SPEAKER_01 13:06

So I would say between those two in the following sense. I think it's a pattern of thinking that uses what I would call the maths process or the computational thinking process by which you try to reason things. And you see this with people today who are technically trained. They often apply this sort of pattern of thinking even when it's not necessarily in their subject. So typically, what I mean, the way we think about this is it's sort of a four-step process. It's like you're defining a problem which you want to somehow get an answer to. Now, the problem could be anything. Sometimes it's very amenable to computational thinking, sometimes it isn't. But I mean, an example I've been give when I'm giving talks, you know, in a room, a lecture room, is I say, you know, imagine if we turn the air conditioning off and we're all in here and I talk for far too long. How long could we survive? Okay, so then that's the definitional question. The next step, step two, is we're abstracting that. We're saying, okay, can we, instead of just talking about it in English or whatever, can we talk about it abstractly, which will give us a much better answer and will allow us to apply hundreds of years of computational ability, so to speak, to get an answer. So we then pose the question. The question might be, let's say, some sort of expression or model that we're trying to write down. We would typically write that down today in computer code. Traditionally, you would have done it in mathematical sort of expressions. Then step three is we say, okay, that's the question. Can we work out the answer using various techniques like solving or whatever? And so then we might get x equals three, you know, is the answer, let's say. And that might refer to three hours we can survive, etc. So then step four is does that make sense? Did that actually answer the question? Is this a plausible answer? Have we missed something? Are we answering it for everyone? Or if somebody's, let's say, diabetic, they might have different constraints. Somebody's running out of oxygen, so forth. So you want to then go round and iterate. So, in a sense, that's you can have different versions of this, but fundamentally, that's how one thinks computationally. And that's a pattern of thinking we need people to understand. Now, do they need to understand what's being done, particularly in step three, to actually compute things? Yes and no. They need to not get fooled. But the main way to not get fooled is to get experience of lots of complex real situations, to see how they pan out and to make sure that you understood how we got to it. Mostly that does not involve knowing how the machine works inside. And it's a it's a slightly tricky thing because if you think of driving a car, most people who drive cars don't really know in any detail how the car works these days. They know how to drive. Now, driving is very distinct from how the car works because the car has a lot of automation in it. 150 years ago, you really needed to know how your car worked, otherwise it wouldn't go anywhere. Now you don't because there's automation in between. And that's the same thing with computation. If you're a Formula One driver, you tend to know much more about how your car works because you're absolutely on the limit of what it can do. And so it becomes more important when you're more specialized to know that. But that's later. It's not most people most of the time driving cars, and it's very much the same with computation.

SPEAKER_00 16:25

You brought up a very good example that really visualizes it because I think now there's, as you said, the digital haves and the digital have nots. I think what they're trying to do is what makes the difference? Is it my understanding of digital and therefore my computational literacy that's going to make me a digital have or a have not? And how far should I go in this process when everything is so new? And I like what you said, comparing like if you have a car, do you need to know your car and all the parts of it, or do you need to be able to drive it? Right? Those are two completely different things. I think we're trying to find our way, this equilibrium of okay, how how are we going to use it? What do we need to know? And in that respect, what do we need from our surroundings? And in case of education from our schools, to make sure that we we stay on par with everything. And I think that, yeah, it it makes for an interesting discussion. Do you have any ideas on that? Do you have any kind of okay, this is the benchmark, this is the minimum?

SPEAKER_01 17:26

Or I think we don't know in detail for AI, but we do know for computation. I mean, I should probably have said, as well as digital haves and have nots, computational haves and have nots, which perhaps is slightly different.

SPEAKER_00 17:38

Yeah.

SPEAKER_01 17:39

I mean, I suppose the way I look at it is the following. If you take maths, which is not really called maths in the outside world today, I mean nobody actually refers to maths when they're doing engineering, really. They think they're doing engineering and computation, things like that. But I I would generally think of maths as the bigger picture subject. I think it's instructive to understand what's failed in education with regard to that, because I think it will tell us something about what's coming. So for math, for not AI right now, but but you know, data science, modelling, the things most countries we were all talked about in the pandemic. Here's a big chart. Should you get a vaccine? Well, most of the population don't really know how to understand that because they haven't learnt it. So that's the kind of thing I think of as computational literacy. And really the process needs to be that we're looking very hard at what people are actually doing in the real world and what we would like them to be doing if only they were better educated. And we then need to take that subject back to education. This is explicitly what has not happened with maths and computation over the last 30 or 40 years. So what are we doing yet with AI? And what do we need to bring back? Well, it's pretty obvious that we are getting AI to, for example, very well in many cases, but take sort of amorphous material that we have and summarize it and pull pieces together from it that we would never have done efficiently as a human. So it's very obvious that that particular new facet is available to us. It's also pretty obvious that it often hallucinates and produces stuff that doesn't make any sense. And so we need to get experience pretty quickly about that. Now, I think the best way to get experience of that mostly in education is to use it as much as possible. Not to avoid using it. It's there, it's not going away. But we need to get skepticism in how we use it and what we do. And the other thing about AI, just to be clear, I mean, Wolf from Alpha, in a sense, which was launched in 2009, was an AI. We would now call it an AI, but it was an AI that ran on a very different system than LLMs. And so what it did was it basically tried to compute answers from curated data. So you'd ask it, you know, what's the population of, I don't know, London versus Paris? And it would go off and it would actually try to find the data, it would try to understand the question, it would compute to that, and it would come out with what it thought was relevant and draw graphs and everything else, which it did live. LLMs do a completely different process, mostly statistical, of going out and figure out what other people have said and done. By the way, putting these together is very exciting for many outcomes that we want. So that's one of the things we at Wolfham have done a lot of, both for pedagogy and education and also many other areas. But when we're talking about AI, I think we need to be very careful to distinguish the outcome of artificial intelligence from the current mechanism, mechanics by which we're getting there, which is at the moment overwhelmingly statistical style, which has just taken the world by storm. But that's not the only mechanism that will be there and it won't have the same characteristics as things that we've had before, like Wolf Malpha, and also things that are to come.

SPEAKER_00 20:59

There's there's loads to unwrap. One of the things that that came to mind as you were speaking about this, I I kind of sense like that it's also about democratizing knowledge, because we have made progress, and I think one of the fears, something that you read about now, is like, okay, it's gonna be very much the winner takes all, the AI winner takes all. There's the geopolitical spheres between China and the US. It's like who's gonna have better AI, who's gonna win this battle, and then there's the you know the internal struggle between companies like uh Claude and ChatGPT and kind of all these beautiful tools that are out there. Looking at it as an outsider, it also feels like, well, this is not knowledge that should just be with this CEO or that head of state. It should be of all of us in a certain way. And what's your kind of opinion about it? Also looking at it from an educational perspective. Should we be kind of engaging with closed AI models that we kind of we feed it? With data and we don't know what's happening with that information, is that a good thing moving forward? Or should we more say, like, listen, if I'm gonna feed this model, then I wanna be able to take a look under the hood and understand it slightly better so I can also educate others on it. Which model would you prefer?

SPEAKER_01 22:19

So just take your last point first. It is not possible to understand sort of how these are each operating. I don't think that direction is going to get one very far. I think you have to watch the effects of them more than try to unpick how the algorithms are working. It just doesn't really it's just too complicated and it's not really where one can get most benefit. Now, you know, the whole issue as to how power is distributed in the world, uh, I mean, both energy power, but but I also was meaning more generally power of power of governance, etc., is very complicated. Like I think it has been, and is, you know, in all industrial revolutions, and I think will be even more so this time. We're a vastly more connected world, things happen much faster. This is a more quintessentially human uh change, as I said. There certainly is, I mean, there's certainly risks all over the place, that's for sure. Um, and there's certainly a risk of the wealth not being well distributed. Do I believe that humans generally are going to be able to do much more? Yes. I mean, I think overall, you know, I mean it's a bit like all of these things, like, you know, can we produce more food than we did before? Yes, massively so. Uh, you know, we couldn't possibly feed the population we've got on the earth today without massive changes. So most of the predictions about how that wasn't gonna happen were wrong. However, of course, the food is not equally distributed. And there are all sorts of problems with that, and the governance of that has been very, very tricky. So I think you've got to take things to some extent once at a time. I think the it there will be a complete overhaul of many, many different things. I think we've all got to look at workflows that we can change now that we're able to, and be as aggressive as possible about changing those workflows where they make sense. I don't think AIs are immediately going to take over everything. A bit like we're still waiting for self-driving cars to actually take over everything and actually drive themselves. Now, on a Californian freeway, that's one thing. I think in the snow on country lanes in England round here, that's something a bit different. And I suspect it'll be quite long before that's really working properly. So I think these things have a longer tail than you might think, but clearly direction of travel is that. I think the main sort of I mean antidote, if I can put it that way, or I maybe more positively find a word, is getting humans experience an education that really is commensurate with the age that we're moving into. And the thing that most scares me about that is the stuck ecosystem of education we have in most countries. Where if I want to start a company that will get somebody a better score in their maths test, pre-college maths test, I can probably get good funding if I get them 1% better score on average in their tests, right? The problem I have is that I think they're completely the wrong tests and they're testing the wrong things, and we're educating people in the wrong stuff. So I if if I want to change that, every single thing basically is against me. It's very, very hard to get that kind of change. But yet, and the worry is that if we make a change, there's an increased risk. But actually, the the problem is the other way around. If we don't make the change, there's massive systemic risk across the population.

SPEAKER_00 25:42

Yeah.

SPEAKER_01 25:43

So that's the thing we've got to figure out how to change. These sort of things have happened in the past. An example I often give is, you know, if you wanted to start a company in most countries in the 1960s, it was really tough to get funding. And then started from the US, this all got unlocked. And actually now every country is falling over itself to say, start, start, you do your start up here, right? We're the best country for your start. Right? So we need to have a similar thing for subject change in education.

SPEAKER_00 26:12

Yeah. I like what you said about the word change, because I think there are some people say there are too many changes, it's too quick. We can't keep up. Do you recognize that feeling and I didn't know what to do?

SPEAKER_01 26:22

Oh yeah.

SPEAKER_00 26:23

Yeah.

SPEAKER_01 26:24

All the time. And I mean, you know, it's like I'm thinking that ourselves. I mean, look, we we build, you know, technical software and do consulting on the future of computation and AI, right? So stuff changes, changed incredibly fast. It's very dizzying. Yeah. But I'm afraid it's not an option.

unknown 26:41

Yeah.

SPEAKER_00 26:42

So, you know, it's not an option to what? It's not an option to not to change.

SPEAKER_01 26:47

I mean, you'll simply get left. Right? And that that that's the danger. That's to my mind, more of a danger than reasonably intelligent. I'm not talking about sort of frenzied change that has no control. But the mindset in education has got to be we got to get people ready for the AI age we're entering. If we don't do that, we're failing. And we will fail. It doesn't matter how well we teach the wrong stuff, how brilliant a teacher we are, it will be the wrong stuff if we don't rethink what what it is.

SPEAKER_00 27:19

Yeah.

SPEAKER_01 27:20

So, I mean, you know, by accident, some of it may be the right stuff, and you may be a fantastic teacher, and by accident, learn your students may learn all sorts of other things. But that's not really where we want to be. We we don't want that. What we want is a basic curriculum which has both actual things you learn, but also ways to do things that you learn that are what we would expect people to be doing. And we can't predict it all, but there's certain things we can predict, and we're way behind, and we need to then build the process by which we iterate that very rapidly. I mean, we can't have a situation in which it's like, okay, we're doing a review of subject X. I mean, obviously I know the math stuff, you know. So if we do a review now, in most, you know, then hopefully by 2033 we'll have a new exam in it. And then by 2035, we'll have students running that, and then hopefully by, you know, 2040 they'll be ingrained in this new curriculum.

SPEAKER_00 28:13

Yeah.

SPEAKER_01 28:14

That's not going to work.

SPEAKER_00 28:15

Yeah.

unknown 28:15

Yeah.

SPEAKER_00 28:16

Um, I want to go back where you said like Wolfram Alpha was actually AI before anyone called it that, but also the implications of all these the AI revolution, as you call it, on Wolfram Alpha and Wolfram research, which you had up. What changes have you kind of come across and how are you coping with those changes?

SPEAKER_01 28:37

Well, to go back into history a little bit, I mean, we were founded in 1980, my brother founded in the US. Uh, so I can do my arithmetic properly, we're we're we're coming up to 38 years. Back then, maths and computation were pretty specialized still. I mean, you know, you saw research departments, we were obviously doing the universities, research departments in companies and government and things like that. What I think we've seen, perhaps going back to, I don't know, um, early 2000s, is that now this is really everywhere. Every CEO of every company, every board of every organization is worrying about data science and now lately AI. And I think of data science and AI as applications of computation. I mean, computation is sort of the underlying structure that allows these to happen. Obviously, they're they're both massive applications of that. And it's, you know, so so one of the things we've seen as an organization is we were sort of a big fish in a in a fairly small pond, or it was a fairly large pond of research, technology, and now we are a smaller fish in a massive pond of basic what everyone wants to do to make decisions in every walk of life.

SPEAKER_00 29:49

Yeah.

SPEAKER_01 29:49

And that's gone through various phases. Wolfram Alpha was, in a sense, an interesting stepping stone because that was kind of, in a sense, democratizing the computation from people who are more specialist to people who are much less specialist, to be able to say we had a lot of journalists using it, a lot of people who would not have classified themselves as technical people. And by the way, that's when my Wolf uh the maths, the uh computer-based maths and rethinking education started for me, because we had an endless people saying, hey, it's great, you can do an integral uh just on the web, but symbolically, and uh calculus and all sorts of other things. And this is now possible. So maybe we should not be learning exactly what we learned before. And it's like, yeah, we've actually been doing this for like, you know, many years by then, just wasn't as open.

SPEAKER_00 30:35

Yeah.

SPEAKER_01 30:36

So that got me thinking about that. You know, what does it mean for Wolfram? It means that we are focused very much on sort of a mixture of deliberate computation that we've built really the ultimate platform for, and this sort of, in a sense, this AI layer that is able to communicate so fantastically and put different pieces together for humans and for itself in some cases. And so a lot of the last few years has been figuring out how to take sort of what we might call hybrid AI and optimize what you get out of that. And so there are many cases right now where the AI that's sort of been at the forefront for the last couple of years, or three, four years, hasn't quite got straight. Like, for example, if you want very traceable medical diagnostics, there's some great stuff it's doing, but there's quite a lot of hallucination. There are numbers and things that don't quite work efficiently with that form of doing it. If you put something like Wolfram Alpha or our Wolfram back end with that, to do the deliberate computations and then, in a sense, have the AI communicate with the human, that's a very powerful mixture. So those are the sorts of directions that we are exploring in all sorts of ways. I mean, you know, it goes back to the fourth. We we can be the back-end API, we we can be a front-end, you know, they're also we can get code of ours written by the AI, so there's so many different ways to do this. It's kind of exciting, but you know, as you say, it's also nerve-wracking.

SPEAKER_00 32:03

Yeah. It changes all the time, but it's good. I mean, you said that you were a big fish in a small pond, and now the pond's gotten so much bigger, but that also comes with more opportunities, I can imagine, of where you can collaborate and you're not you're not the only one anymore, and nor nor are you the only one kind of screaming that it is important.

SPEAKER_01 32:22

Well, well, it's actually strange in that, just to just to answer it. We are sort of the only one doing exactly what we're doing, which is we've built this sort of big structure of computation that we think we we believe computation is this very universal thing, and we've built sort of the ultimate representation of that, whether that's data, models, or even AIs, neural networks, whatever it is. The thing that's sort of changed recently is that if you take a problem that seemed like it was only amenable to very deliberate computation, now the question is, should you use that or should you just get a statistical inference to figure out whether that's the answer? And to what extent do we need this computation? Now, I certainly a strong believer we will need it in all sorts of ways, because you need to generate new things, and there's all sorts of crossovers. But so it's a slightly different it it's even a different picture that, because it's kind of like, yes, we we have our own thing that we're kind of unique in, but it's like, well, there are different ways, different mechanisms by which you might get to some of these, but not all of them, and then it's like which things do best to produce these answers.

SPEAKER_00 33:30

Yeah, that's true. But that was actually also going to be my next question. Is like there are so many AI tools, and and kind of new ones pop up every day. And so just a while back, I was uh um at the great pleasure of doing a week at MIT doing a training in entrepreneurship development. And one session was about someone who knew so many AI tools, and he was so he was really in love with all these tools, and he said, Oh, look, this is a multi-billion dollar company, and it just started last year and whatever. And skeptic as I was, I said, Well, that's what they are today, but the competitor's gonna take over tomorrow. So it feels like there's this gold rush that's gonna lead into a bubble of all these companies that are the hot stuff today, but where are they going to be tomorrow if kind of new companies like that keep on popping up? It's like, are we are we kind of in such a in such a moment that we're not playing any favors? Um you know. Well, we're we're building up this bubble and and it's gonna it's gonna implode. So how are we going to deal with that? Or am I now too much of a doom thinker and and actually should say no, just I don't I don't think it's binary.

SPEAKER_01 34:40

I think you're gonna have companies that come and go, you're gonna have some collapse in different directions, there's gonna be a tulip rush of some sort. But the thing is, there is reality to it too. And it's a bit like, you know, I mean, look, we the most recent history obviously is the dot com bubble, but there was reality to it, right? It's like our lives have all changed because of what happened then. Now, did every bizarre startup that was there and was covering that actually succeed? No, of course not. But some of them did. So I think, particularly led from the US, there is certainly hype. There's no question about the fact that there is hype. But there I think is a very strong reality. I mean, this new machinery we have is extraordinary in what it can do, and I think it will become more extraordinary, both as it is hybridized with other machinery, like the sort of things we've been building, and just as it's developed in its own rights, and just as you know, hopefully it gets more efficient, it burns less energy for, you know, each thing that's being done, new models get discovered. So I you know, would I say that every valuation is correct and everything? No, I wouldn't necessarily. I think there's all sorts of things going on there. Do I think in general trajectory this is really important? A critical moment? Yes.

SPEAKER_00 36:00

Well, we should stay awake, and it's definitely a critical moment. It's just that I'm I'm just a bit of a skeptic in the hype building and in kind of all the praise that is being given to this one thing, and you feel like you're missing out because you didn't understand this specific tool, and then the day after there's another one, and you know, no one speaks anymore about the one you've missed out on. I don't know.

SPEAKER_01 36:24

Yeah, as a human being, because I have human emotions around it, I don't always know how to it's very difficult to remember, you know, for example, to what extent you invest when you think that there's a bubble and you know, but there's reality, and I remember feeling this very strongly in a dot-com boom.

unknown 36:42

Yeah.

SPEAKER_01 36:43

It's like this isn't making sense. I mean, as in the underlying thing is there, but there's an element of it that doesn't make any sense, right? And it didn't at some point. But I think there'll be something similar that happens this time, I'm quite sure about it. But I think that one needs, I mean, the thing I try to do, and I'm not in any way saying I have anything figured out here, but I just try to look at my workflows over and over again. What am I doing each day? Why was I doing that? Was it predicated on the machinery I had 10 years ago? And actually, that's not really the right way to do it anymore, because there's much better machinery. And it's very, very tough to do that, right? And you know, it's very tough, by the way, to build technology. We have exactly this problem building technology all the time. And by the way, thinking about the whole results we're trying to deliver with computer-based maths, where we've tried to build a curriculum assuming computers exist, and we're asking all sorts of questions of ourselves there. You know, what do you really need to learn? Is the way in which you traditionally have learned it make any sense now, given that you're learning something different, and you also have computers available and you have nice things you can instruct with and sliders and so forth. And so if you look at our materials that we've been building, they're very different. And we're now asking the same question again. Well, we built the materials before these AIs were available. Now that AIs can help, let's say, do open-ended tutoring and also open-ended assessment, can we change the whole way in which we might deliver this? And I think the answer is to some extent yes. So I think these are I I was focused on workflows, and I think the best one can do is to understand what one's trying to achieve, whether that workflow is correct, and whether there's a fundamental change that one can see. And it's really tough to do that.

SPEAKER_00 38:34

Yeah. No, it's it's really tough. And but you need to be at it, right? You need to be involved because you need to see how it plays out. And as you said, like you're you're going on 38 years with Wolfram Alpha and Wolfram Research, so so that's quite a journey. So how do you stay relevant in such an ecosystem in such a thing? And you're obviously doing a great job at it, but it that's that's a question that that one one would ask, I guess. So yeah.

SPEAKER_01 39:03

Which we are asking every day. I mean, so I mean, I hope we're doing a good job at it. I you know, I can't, I you know, I've never been secure in that in the 38 years, yeah, but I'm probably less secure in knowing that we're doing a good job.

SPEAKER_00 39:17

Yeah.

SPEAKER_01 39:18

As you say, the gains are fantastic, and I actually think what can be achieved now with this mixture of computation and sort of AI versions of computation, if I could put it that way, or statistical AI version of computation, is fantastic. I mean, one way, by the way, I put this, uh I mean, another way to think about this, which is I sometimes say, is back at Wolfram Alpha type AI, back in 2009, we would have loved to deal with some of these issues about helping diagnostics in medicine better than than humans alone could do. And I think what the barrier there was we weren't that good at talking to the human. So we were very good at I mean, I I put this sometimes, that if on the human spectrum, Wolfram Alpha is a bit aspergic in nature. It's it's very good on its being very accurate and specific, but it's not always the most easy, it doesn't find itself very easy to communicate with the human. I think we've got the other end of the spectrum for LLMs, and we have something that is very poetic, but doesn't always get facts straight.

SPEAKER_00 40:22

Yeah.

SPEAKER_01 40:23

And so I think there's a huge opportunity now to really make sure, and this spans everything from education. I think it's really important. I know a lot of people tried LLMs, for example, for maths education specifically, and found that it's not that good at it most of the time. Now, as the LLMs you know mine more information, they can kind of fake it in a sense by knowing everything, but that doesn't really mean they're working it out. So so I think there are great ways to uh you know empower people in in all these areas that that need a mixture of information, straight data, and working things out and need to be able to communicate well with people. So I I think that'll come up. I think in the next year or two you'll see quite a bit of that.

SPEAKER_00 41:07

I think um I mean, next to doing this beautiful podcast, I'm involved with an AI initiative specifically in genomics, also uh healthcare adjacent. Uh so what we're trying to do is actually we're doubling down on the privacy and safety of people when it comes to their DNA information. And we're focused for that matter on one end as a kind of we've built an infrastructure where people can get to vault their information and make sure that when their data is used, it is done so with their full knowledge. That's on one end. And on the other end, because AI is, of course, now so on, you know, as we this entire conversation was about that, AI is so dominantly present. We are also making sure that if DNA information is used, it is the safest version of that. And simulating DNA information, making it synthetic, is one of the ways to go so you can feed these models at scale, right? Uh, because you don't just want to put in proprietary patient information in an in an LLM and just, you know, as a Hill Mary, just see what happens. Sure. You don't have the consent for that. So we're all focused on synthesizing information or relevant biodata so it can work in these bigger models without violating anyone's security or privacy. We identify ourselves a little bit as a as a bridge figure because we don't know how LLMs will deal with our information, like let's say five years down the road. We factually don't know, but we do think that synthesization is in this, at this moment, the best way to work with that. And I think that's yeah.

SPEAKER_01 42:50

And we've also been generating synthetic data, not so much in the medical area, but in all sorts of areas where traditionally we've been generating uh information, which is effectively what Wolfram Alpha was doing in all sorts of different areas. It was it was computing new information from existing data that it had. Look, I think all of these areas, look, whenever there's major new technology, there is an arms race to some extent between good and bad uses of that technology. Yeah. Bad maybe for several reasons. It may be incompetence or it may be malfeasance. That's going to happen this time. You know, there's no way around it, right? And you know, over time, societies, I hope, you know, unless there's a catastrophic mistake, societies will, after some pain, adjust to the mistakes and start to it'll always be slow relative to the technology. But I think that is the nature of progress through history. It just happens more in view of everything. We didn't live through the previous one, so we don't know. It feels like it would, because it feels like it's sort of several levels up.

SPEAKER_00 43:58

Yeah.

SPEAKER_01 43:59

And we need to makes us always set up for that properly is so that we as humans remain in charge.

SPEAKER_00 44:04

Yes. One last thing I was wondering is because you ended with humans being in charge. And I think that's just that's just such an interesting tension. Because there are people who are saying like what what do you mean in charge? Humans will always be in charge because we've developed the AI and the I will always do what we say it will do. Am I correct to understand that you say we should be be a tad careful here because that's not something we can just assume Yeah I don't think we can just assume it.

SPEAKER_01 44:32

I mean not not for everything in all that and you can argue the extent to which machinery of past revolutions has in a sense subjugated humans, not in quite the same way as we're discussing here, but where humans have become sort of, I don't know, automatons to the machines and the machines have ended up subjugating groups of humans. So whether in total or whether for some aspects, I think we do need to be a bit careful of that. In the end, I think the the only thing one can do, I mean apart from trying to persuade people to be as responsible as possible when they're developing the technology and you've also just got to make sure people do as well as they can to understand and get education what to do. And we see this in our democracies right now. Some of the problems with recent issues with democracies I think is because people are not involved in the decisions very much anymore. Because many of the decisions depend on computational literacy and they don't have that. And so if you're watching endless governments for example take decisions that you feel you really don't understand in any way at all and can't really connect with, you might not understand the detailed models, I don't understand the detailed models in many cases that are being used, but I have some idea of how they arrived at how to be skeptical of them, what I could believe, what I might not believe. If you don't have any of that, then you're going to start backing people, for example, who talk nonsense because you can't distinguish between people who are making decisions on a reasonable basis with this new era of computation and those who aren't. And I think the same thing will sort of happen with AR. So the best we can do is to step people up, get them experienced as quickly as possible and make sure that they're as prepared as possible for at least what we know is coming as well as we can. But I think there is a risk if we don't do that. And I think that the there's a risk to societies even right now from the failure to change maths to be this computational literacy. It's almost like we're in a pre-Enlightenment era where you believe anything because you can't really believe any expert because you didn't some of the experts talk nonsense but you didn't really know which ones they were.

SPEAKER_00 46:41

Yeah. And that's why we need a basis of computational literacy to make sure we can debunk those that are talking nonsense. Yeah. Correct. Yeah yeah this has been a wonderful conversation. I have two very last questions which I ask all my guests one is about and this is just me being an amateur researcher into morning routines. So my question is do you have a morning routine and if so would you care to share with our listeners?

SPEAKER_01 47:08

I don't have as good a morning routine as I ought to. My main morning routine is to get up quickly and make a plan for the day as to what I'm doing and make sure that I do that as actively as possible rather than just drifting into something. And the other thing I try to do is cycle to work mainly because it forces me to take some exercise. So I reduce the excuse tree about not doing that into when it's very icy or other other such problems or when I have to go somewhere in the middle of the day. So those are two very minor things that I I try to do. But I this is another workflow that needs to get improved.

SPEAKER_00 47:44

Yeah. I like that that you kind of say okay what am I going to do today that you kind of just plan ahead because that that takes away the chaos maybe that you get up maybe not the chaos but the apathy when I when you sort of get to the end of something think you'll do something I think what sometimes happens is you put off things that are harder if you haven't forced yourself to remember that you have to do them because you do slightly easier things first. At least that happens to me. Because you've got the most energy. And then the last thing is about reading other than the maths fix because I I'm afraid you can't plug your own book Conrad other than the maths fix. Has there been any reading throughout your you know beautiful career that has inspired you? So for those people listening who would say oh I would love a career like Conrad Wolfram has there been any book that has kind of inspired you I'm not a great book reader I have to say in general.

SPEAKER_01 48:39

I think there are characters that you know of which I've read books and seen doc you know I mean seen biographies and so forth or read biographies. I mean Steve Jobs is certainly a very interesting character in terms of sort of incisiveness and by the way he was very connected with our company or my brother in particular and he came up with the name Mathematica was our launch event and all sorts of things. Steve Jobs was was a very interesting character in in being able to just be exactly clear about different technological changes and fantastically good at changing these workflows. And another one also who was always very interesting is the physicist Richard Feynman, also associated with my brother's tutor at Caltech actually again another very incisive character and I always enjoyed and appreciated reading his insights into both physics and life.

SPEAKER_00 49:33

So I mean those would be two people rather than necessarily books specifically by Walter Isaacson about Steve Jobs indeed.

SPEAKER_01 49:40

Yeah that's that that could be like and we very rightly point out it doesn't have to be reading it could also be in what inspired you and and you've you've you've given a lot of Yeah I mean I do enjoy biographies I I should probably add that in general I find how people operated at different points in time whether as I say that's Jobs or Churchill or other people where they've had to make very definitive changes in what people thought and understand things clearer than other people I find that interesting wonderful conversation a lot to chew on.

SPEAKER_00 50:12

We should have you back in a year's time again to see where we are with our mathematical literacy or our computational literacy thank you so much for your time Conrad Wolfram. Thank you very much enjoyed the questions and the conversation You've listened to Pulse by AlphaWire produced by Natalie Piles and Amela Faisal with great music The Optimist written by Holly Hamill performed and produced by Alo. Episodes hosted weekly by me Aldo DeP