If you go from the base assumption that we are very bad at predicting the future and you start to look at the practices of anticipating futures, it becomes surprisingly awful. The horrible truth at the center of all of that is that experts are weak predictors of the future. They typically underperform the general population. And so in scenarios rather than say, I know what will happen or we because we're so bloody smart, we know what will happen, you say, no. How could this happen otherwise? What could surprise us here? And all models are wrong. The the first model is wrong. It's the model inside your head. That's the problem, not anything else. It's the model right inside there. It's... That's the thing. You are the problem. And when everybody says the economy is going to grow, they're preparing for a global financial crisis and to give them the courage to act in new ways that are almost unimaginable, prepare for events that are almost unthinkable. Right. Because the future is fundamentally uncertain. And anybody that tells you they know the future is a charlatan.

So, a way of framing it in in the model terms is to say, models fail you. So the scenario viewpoint on that is models fail you when you need the most. And,

And so, at its heart, the assumptions of scenarios are saying that, you know, some of the most powerful people on Earth with the greatest resources, you know, in in the world, often make mistakes on what they anticipate the future to be. And we we sort of offloaded that and say, well, AI can function in that, and there's certainly gonna be successes, and I can talk to many of those successes. But if you go from the base assumption that we are very bad at predicting the future, and you start to look at the practices of anticipating futures, it becomes surprisingly awful. So experts... So the usual thing is we don't know what the future is going to be, so we're going to ask experts. And you go around. So whether it's AI, we'll find model builders, or whether it's economics, we'll find economics, or, which... Whichever discipline there is. And we get the best experts because this is serious decisions, and we're serious people, and so we're gonna spend real time at finding that. The horrible truth at the center of all of that is that experts are weak predictors of the future. They typically underperform the general population. And the reason for that is is kind of obvious if you if you look at that, but it still hasn't translated to practice. The reason for that is much of the future is dominated by how things interconnect. It's how the economy relates to politics, how politics relates to mathematics, how mathematics relates to technology, and it's those interconnections. So if I'm an expert in the economy, I forget about all the interconnections because that's where my expertise are before I'm an expert in this. And so because they don't really talk about the lines or the ways that they interrelate with each other, we get very bad at prediction. And so I'll give an example. So in 2006, we were working on, what is the future of financial services? What is the future of sort of banking, basically? A great time to be predicting that. Exactly. Because, you know, the world had had a significant, expansion since 2001, since the dot com crash, which been bought. And then slowly, slowly, slowly, things were were were moving forward, and people thought they were very safe. And so we spoke with an... And and so when you do scenarios, it's a very simple practice. You go and find kind of a network of... Pier Wock used to call it a network of heretics. And so a thousand heretics. And you find people... So these are ideas which are not yet written down. And you go speak with them, and you say, well, what do you think is going to happen? And you think, what is this uncertain? What what is the underlying uncertainty? So normally, the question, what do we know about? But then you ask them, well, what is the underlying uncertainty? What don't you know? And what do you think would change things if that one thing changes? And so speaking with an evolutionary biologist, which is unusual when you're looking at finance, the person said, I don't understand why the financial system is designed this way. Because we know in evolutionary biology, if things are coupled closely to each other, so there's lots of feedback and very fast paced feedback. If one thing breaks, then everything breaks. And... But the way we treat the risk, equations now is that that reduces risk. It doesn't increase risk. And so you go ask the economist, like, why is this done? Why is this equation interpreted this way that, because they're closely coupled, that'll reduce risk. So the idea was if it's closely coupled, market will understand what those risks are and the risk will be better priced. I said, well, actually, you can interpret it both way, but we chose to interpret it as if it does risk reduction. And so in scenarios, rather than say, I know what will happen or we because we're so bloody smart, we know what will happen, you say, no. How could this happen otherwise? What could surprise us here? And so the first thing you say is, well, what would happen if there is a breakdown of the world's financial system? And it came from... Ancillary to that was a question is we've had five years of growth after this crisis. I can talk about anticipating sort of terrorist actions and all of these in 09/11. You know, I can talk about that as well. But, we've had five years of growth. What in the world could make things go badly? And then we said, well, one of four things, could lead to this, could lead to a kind of shift in the global financial system, and that would cascade throughout. So we said, abrupt climate change. So we kind of assume that climate is, the climate change is a slow and steady process. But there is good science that say, actually, there could be some very abrupt and big changes there. And we know that from geological time, there can be abrupt change, but we don't really know exactly what the speed is of those changes. Could be ten years, could be six months, could be a hundred years. We just... It's a very hard thing, and I'm... Maybe the science has developed since we last looked at that. So abrupt climate change, levels of US debt, an oil shock, and, I've forgotten the fourth one now. It's been a while since then. Let's say, one or any... Oh, global pandemic. So one or any of these four things, if that happens, could lead to, well, it could lead to nothing because we solve it, or it could lead to a global financial collapse. And so we wrote a scenario about that and, actually wrote it around a global pandemic. And, and then by 2006, the the party that we were working with started preparing for that. They started doing things anticipating it. And when the crisis hit in, 2008, they were one of two banks globally that had asset accumulation. Two. In all of the world, two banks. And the bank had, at the end of that year, I think 2,800,000,000, dollars profit, in that it's most profitable year in a hundred and fifty years of existence. So when people say, nobody could have guessed. We could never anticipated it. There's no need. Things happen. No. It's looking at models and saying, how could this... Our assumptions on this model be wrong? How could we be wrong? How could the assumptions around what this model is built? And when we need the most at a times of change, yes, models are wrong. So in your your... What was the phrasing of this web?

ways of looking at things. And you get to tell... So rather than tell one story about the future, you get to tell three or four stories about the future, which you work... Together you spend maybe a year. Scenarios are typically a thousand to 2,000 pages of actual documentation and work that goes into them around a year, and then pretty much half of that time to to 60% of that time is spent building those pages, and the other half is bringing it down to one page. So it's communicable. And that's really more or less the time that you would spend on that. Yeah. So, I mean, was was your

Interesting. Yeah. So it doesn't really... I mean, when you look at that, it doesn't really matter what the causes of the crash. So he said there are these four vulnerabilities, and it doesn't really matter which of those four. But if one or any of them were to happen, that will lead, to this and said, well, which of them are the most interesting to explore? So we chose the pandemic, but it it could have been something else. The interconnection is the problem there. So in in this case, you made a wrong but useful model.

it's... The the point isn't necessarily the narrative structure by which you build it. The point is that that interconnectivity could lead to a... Not a not a Asia crisis because we had the Asia crisis. Before that, we had the Mexico financial crisis. So there's indications of that. But this time, it will scale from a region to become global, and that's the insight. The actual cause is not that important. Okay. So we try to look at these things like archaeology, and we should look at them as systems. And in systems, how those systems connect, going back to my expertise point, how those systems connect changes things.

show us how that how that transition happened. So I'll I'll I'll answer the question with with three different hats, and they're kind of opposing viewpoints. So, the traditional scenario perspective, and then I talk as a violin maker. You know, we've made the violin this way. We've made it forever. This is what make... Goes into a good violin, etcetera. As someone who played violin for ten years. My my daughter just finished playing in a concert in Austria, so I'm very, very proud father here. And, so, so if one approaches it this way, so Pier Wach back in the day, so this is, you know, good sixties, seventies, the work, which was done then. He said... And it had a lovely phrase. It says, scenarios change is the microcosm of the mind of the manager. So in our minds, we have some type of mental model, what is happening in the world, some type of model which is there, and all models are wrong. The the first model is wrong. It's the model inside your head. That's the problem. Not anything else. It's the model right inside there. It's... That's the thing. You are the problem. And many of these things are done in organizations, but it doesn't have to be. So I'll talk from the perspective of organizations or governments or these type of things. So there's a small group of people sitting in a room, and it's how they talk and how they think about this. And so you're you're shifting that. And so when when one presents and say there's an alternative future coming, so 2006, you say, there's a good chance of a global financial crisis coming. What's the likelihood of somebody making a decision based upon that? So even if you were to travel back as a time travel, knowing the future and you just get into the room and you say, guys, because it's typically guys, but increasingly it's a bit more mixed, this is going to happen. The likelihood of people taking decisions on that is nothing. So we tell each other it's about the models, it's about the... It's nothing to do with that. It's about the decision making process that follows that. So you you take people. So in scenario practice, I would take an executive, a very, very senior executive or somebody, governor of a reserve bank or, politicians or... I'll say, you are coming with me. We're flying around the world finding some of the most interesting, remarkable people you'll meet in your entire life, and we're gonna listen to them. And and, typically, we would film it as well, so those things exist as records now. It was very, very similar to this kind of discussion.

Once it was in the forest. In one of my favorite interviews, somebody sat in a tree, and we looked at the future of learning. Okay. So, the organization was spending about 1,000,000,000 per year on learning systems. And so they say, can we do it differently? Can we do it better? And that also had huge societal impact. But you talk with a ranger... I mean, from, you know, one of the sushi masters of the emperor to Japan, How did they learn it all the way to the undersecretary of education for this process? So you go find people who are remarkable people in their own right, and you listen. And something changes in the person because they sit there and they listen to somebody. I remember, having an executive next to me with a billion dollar IT spend. So that's literally... They're sitting right there. And, and, Jaron Lanier, who's a who's a lovely, exciting, kind of controversial person, invented virtual reality. And he's... And he was interviewed, similar format, and, he said the way IT decisions are being made is exceedingly dumb. You could feel this person bristle up. You know? You know, we're doing our best and this and all of those things. But in the shift there, suddenly, they were forced to listen to an entirely different viewpoint than their own. And if you're a senior executive, people pitch you things all the time. So they tell you a story like, well, we're going to do this, etcetera, etcetera. But sudden... And you make a judgment call, but suddenly, you're in a position where you're not judging and you're listening to an opinion which probably disagrees with you but which you respect. And you can see how that mental model in in the in the mind of the manager, that kind of microcosm there is shifting and changing. So as a violin maker, you have to bring people into situations where they're confronted with other futures, other realities, and the ones that they think are doing. And you have to give them the space where they socialize it amongst each other so they can take dramatically different decisions. Dramatically think... And when everybody says the economy is going to grow, they're preparing for a global financial crisis. And to give them the courage to act in new ways that are almost unimaginable, prepare for events that are almost unthinkable. So that's my kind of first hat. Yeah. Okay.

You're talking to people about assumptions in areas that they don't even think about this, and you're collecting those things and you're working together through... With a small decision making group through thousands and thousands of pages of these things. Increasingly, you could do some of that, using technology. So we developed a piece of technology that could map and build graphs of how all of these things connect, and you can start to see that. But we were not alone in this. There's many, many attempts at kind of doing that that that tooling. And often you're seeing things at the edge. There's something at the edge which is like, why is this there? And I and I give an example from, COVID. So you're doing these discussions. And why why you're doing that is there's a whole set of assumptions about the world which are not yet written. So even in AI, you're building on things which are written, but there's things which are not yet written. So if you look at the discussions on global financial collapse in 2006, they're almost zero. So it's not something that you track momentum of or any of those sort of naive assumptions. You're you're you're looking for things which are not yet written and you're surfacing that from it. But, you know, there'd be some notions of that at the edge. And so in COVID, the question was we did scenarios on what does the post COVID world look like. And, this was in the middle of the COVID days. So what does that look like? What is the structure of how it will cascade through the world? And one of them was at the edge, so using this technology, is food security. You go, what is food security gotta do with COVID? And, the Nobel the Nobel Prize for economics was one on this. So, normally, when we think about food security, we would say, that has to do, with supply. We're just not growing enough tomatoes or dry goods or these type of things. We're not doing that. And so, but what what was shown there is that it's really because of market failure. So somehow there's almost always enough supply. There's kind of isolated cases, but somehow that doesn't connect with demand. And so here, because everybody was was staying and in lockdown, would the markets go in lockdown for food? And what does that mean in a developing world where people don't really have the resources to find alternative means to guinea pigs? And as a farmer, if you're sitting on the farm, your biggest capital costs really is... If I've grown the tomatoes, is do I go and harvest the tomatoes, which requires labor, and do I pay the money to have those tomatoes go into town to be sold? That's that's my biggest capital cost. And so farmers were expected to make very rational decisions and say, maybe this... Not this crop of tomatoes. I'll gather some seeds. I'll let it go to waste the tomatoes themselves, and we'll kinda sit this one out, this season out. And that would have, and I remember, it would lead to a hunger of biblical proportions was the statement, which was said. Not... It's sort of slightly different, but but with the underlying same way of thinking. And so in the room, you could sit there. This is a virtual room. It says, well, this is coming, this food security. Can we intervene and shift this? So the mental model is food security is not an issue, and then the... Our models are wrong. Right? The mental model and say, actually, it is an issue. You discover at the edge some of it technological means, and then you bring it back in the decision making process and say, is there an intervention strategy? In this case, it was a bank again. And the bank then said, well, you know, the the the public safety of our area is a primary concern for us because the bank doesn't necessarily normally get involved in food security. So they had a small product... Project around animal feed, and the team came away and says, no. We're gonna rechange that and change the entire process using... I I didn't go into the detail of that if we want, and we can apply it to to human food and make sure that that enables the market to exist. And so within a month, this thing was built, and within three months, we had supplied for free, a million meals. And that process has extended on. And to date, I think it's done about a 120,000,000 meals, which is given away through intervening into that system. And I can go into detail if I want, but it's really about how one does that model, in this case, technology help to challenge how we think the world fits together. So you can get with tiny little changes, dramatic shifts in actions. Yeah.

Yeah. Yeah. Yeah. Or at least you're playing a game of probability. So you're surfacing things which wouldn't be seen otherwise. Does it mean that you're gonna surface everything that's important? You don't know. Because the future is fundamentally uncertain. And anybody that tells you they know the future is a charlatan.

There'll be... I mean, it's also... There's a beautiful painting by Caravaggio. And you can go in the Louvre. I took my kids to go watch this painting. It's a... It's not a very big, but it's a beautiful painting. And, so in this painting, a person shows their hand, and a palm reader is reading it. And if you look at it really closely, you can see that the palm reader is stealing the ring of the person's finger as they're reading it. Right. And so he captured that moment, and he was commenting on fortune tellers there. He was like stealing that ring. And so this is 500 year old comment on this. There's a lovely book by Sherman, the Fortune Sellers. And it's an industry sort of around $200,000,000,000, today in scope of analyst reports telling us where the world is going or models or all of these various things, everything from economic models. If you measure all of them, which which Sherman did, but many parties have done this and say, who's been right? Out of all of these things, who's been right? Whichever. Whether whether that's crystal balls or models or analyst reports or these things. Who's been right here? Two. Demographics, which is kinda predictable. Short term weather. The end. So we spend 2 to $250,000,000,000 a year on analyzing and anticipating the future, and we know that almost all of it's wrong.

Excuse my sins. Sorry. I'm a I'm a recovering strategy consultant. Yeah.

someone who had already made a decision. Yeah. Yeah. Amazing. Yeah. So in scenarios, which is slightly different because this advisory... And strategy would follow from scenarios. Yeah. What you... What you're trying to do is we go into the process, a deep process of deep trust, and, you know, these are, you know, rather expansive affairs and also costly affairs. It takes real time to do this. But, you go into a process and say, it's better to be surprised in the meeting than in the market. So let's find out what are those unknown things that could happen to us that'll surprise us. And if you want to have the usual type of things, there are plenty of advisory practices that are very good at cutting costs or firing people or just... God bless them. That's it. But this is not what we're doing in scenarios. Of course. In scenarios, we're trying to find where are the surprises and how can we be surprised and prepare for those things a little bit earlier than most to dramatically different outcomes. Mhmm.

You know, we've done as, you know, a good two two and a half... A quarter of a century of the scenario work, ranging on on on various. And then I I stood and I asked myself, well, what are the real global challenges here for my children and their children's children? So what what kind of work does one want to do? And, you know, like I mentioned, the legacy of that work is, you know, hundreds of millions of meals, billions of dollars for clients, radical shifts and things. And we can go through all kinds of more battle stories if you want on that. But he said humanity's greatest challenge. I was asked to give a speech, at, in San Francisco on the promise and perils of AI, you know, overlooking the Bay Bridge. It's a a place called Shack 15, which ain't a shack at all. It's a spectacular view there on the Bay Bridge. And so, and you're looking and there's a heart felt set of technologists in the room, you know, everything... Everybody from, you know, Rusty all the way throughout. So very heartfelt California technologists in the room, with a great deal of optimism on AI and these type of things. And I said, you know, it's a framing I recoil from. Because the promise and perils of AI is the wrong framing here. The the real peril we face as humanity, the real peril is climate change. It's not a hypothetical thing. It's not something that might happen, that AI overtakes, you know, and and plays chess better than people and whatever that ends up. It's nothing like that. The real challenge here is climate change. And best as at the scale of the planet, we're gonna have to get really good at preparing for different worlds. Certainly for our children, absolutely for our children's children, and perhaps even for ourselves. And so we're at the cusp of these massive changes. So what can I do to contribute to our capability to change and to to act in this? On these things, change our own actions to act on this? And so,

It it was it was bit with some confusion and trepidation. No. They took the message well. But you're you're kind of the person in the room that that questions the base assumptions on that. And, you know, after spending a quarter of a century doing that with some of the most powerful rooms on Earth, that's part of the course. You know? You're gonna question things, and that's the job. So, you know, in 1972, so, in the oil crisis, an interesting bit of work was done, by a good friend, Jorgen Randers, and and Donis and Donatello Meadows of the Club of Rome. So I'm I'm I'm very honored to be a member of the Club of Rome now. But that work said, you know, we've had all of these years of economic expansion post war. So since after the war, the world's economy has expanded. What are the limits here? Are there any limits to the economic expansion? And let's build a model which will model the Earth. And Jorgen put punch cards in in MIT using system dynamics to model the earth and came to the surprising conclusion is there's very real limits to the world, and that limit is the carrying capacity of the biosphere. And there's really just a couple of scenarios. Either we recognize that we're in the carrying capacity, we're approaching that limit, and we say, okay. Hang on. We gotta change how we run our economy. Or we go overshoot, and suddenly the the the the the natural systems can't carry our economic activity. And we figure that out, and we return back to within the current... Or we overshoot, and it's too late. Those are more or less the scenarios on... And what the limits to growth was about. So the club of Rome then said, and, let's think about this again fifty years after it happened. So... And they bid... Did a study, the earth for all. So not the earth for some, earth number for all. So the earth for all. And said, what does that look like, the limits to growth study, but now fifty years later? But it came with five intervention strategies. So there's five areas which we can intervene to radically shift our outcomes. So not a report about just the limits, but a report about what can we do about those limits to ensure happier, healthier, whole lives for ourselves, our children's children, etcetera, etcetera. And so the question that we were faced with was it's been 36 doublings of of, you know, capabilities of computers, since since that. So, '38. I I could be wrong about that, but it's it's, are there other technologies to apply here than system dynamics to build this model, earth models with. And so we decided, let's build a model which is an AI model on climate change with the input of the Earth for all work and all of that and the SDGs and this and say, we're gonna build Climate GPT, but on a slightly different problem than climate is the weather. Because we've built all of these inscription devices on, tagging carbon stocks, looking at the the future metallologically, using supercomputers to model the Earth, the international instrumentarium around the IPCC, which works at this. So every few years by the thousand scientists work on what is the best knowledge of where the Earth system is going. It says, we model the weather extremely well, and we're getting better and better at this heat integration. But what about the social implications here? What does that mean for us? What do the implications mean that if the earth is changing, what does that mean for us? So we get these numbers on maps. It'll be 50 degrees in Morocco in the next ten years or so. So just a bit over fifty years. I don't know exact time, but sometime in the future, it'll be there. But these end up being numbers on maps. And what we really wanna know is what does that mean to our social decision making? Water security, food security, institutional failure as political institutions fail, as everybody is under that stress, mass migration, likelihood of conflict, all of those things. Those are models we don't build. So we have deep expertise in modeling what's happening, but not in terms of our own. So let's increase our capacity to act on it. So we build ClimateGPT around that idea to change our capacity to act on it and build the base tooling, for acting. And, you know, whether it's adaptation or mitigation, those are different approaches and different ways of thinking about it. There are many, many different ways. Let's incorporate all of that and put the best knowledge we have on that, including the best scientific papers that we have on this into that. And that model is now in its third iteration. We're just about to launch model 3.5, and by September, it'll be model four. Yeah.

What does this give us? Do we do we know the changes that we need to make? If we were to stop all oil, today, it's... The the bulk of the changes in the climate is... The climate system is characterized by amplified, feedback and delayed response. So what we see now has already happened, and it's slowly, getting larger and larger and larger. So if you were to stop all oil, all of the carbon in the air, we're above 400, parts per million already. So it... That's not going to solve things. It really isn't. Those things are are already happening and already baked in into our future. So it's much more open for debate what needs to be done here, than what typically thinks. And in order to make these kind of decisions, so do we spend a dollar on stopping the coal fired power plant or building a dike which makes it safer and and better, for people living in low lying coastal areas? Those is an allocation of resources question, and that depends on is, can the... That country get access to those resources or not. So there's very deep and subtle ways in which we responded. So even though we know the climate is changing, our social responses to that is very nebulous and up for debate and up for discussion. You know, that... So what we said is, a, is let's build it on green power. Can you build an at scale model, a foundational model, and fine tune models on green power? And since what we're going to do is we're gonna build models on green power. That's one. Two is let's make the model as small as we possibly can, not as large as we possibly can, but as small as we possibly can, preserving results and all of these type of things. So that is democratically more people can run it on more humble hardware, than a very, very expensive chips, that we're in. And when it runs, it uses less power. And so... And and when it runs, you could do more cycles of that agentically, by using a fraction of power. So let's make it as small as we can, this model. Let's make it open so it's open source. Let's make it open access. So you don't have to pay to use it, but, you know, it's there. It's there. It's available. You just apply, and then you can use it. And let's make that model in a process that learns and amplifies and deals with particular aspects of the climate problem through each generation of those models. And let's incorporate the best of our knowledge into that model. So... And it's a standard open source, open knowledge play. Let's make the barriers to decision making and knowledge about this a little bit easier, a little bit lower, and so that when I'm a city and I need to make a climate resilient strategy, it's just that much easier to do that. So training it on green power and still running it on green power, and every round, we still do it on green power. But it ended up that the model, when we first published it, it surprised us. So by doing something which is domain specific, now it's a massive domain. It's climate and our human actions on that, so it's not, you know, playing chess model. It's it's a massive complex domain. We were able to use, even if you run it on dirty power, a comparatively performing response, if you look at it across a bench or benchmarks, is, equal to models 10 times their size. So models are measured by parameters. So 7,000,000,000 parameter model would perform equivalent to a 70,000,000,000 parameter model, and a 70,000,000,000 would perform to a model 10 times their size. So it surprises, but you can shrink models 10 times smaller. And that means that you can do more regular updates. You can do more regular things. It's lower cost of training, and it's easier access. It's a lower cost of GPUs that one needs because that doesn't scale area. So there's... And to quote Feynman, there's lots of space at the bottom for this. Now if you look at the sort of general data center discussion, which is a valid discussion we have, data centers today command, and produce, around 3% of greenhouse gases. Airline travel is more or less in the same thing. Shipping is more or less in the same amount. Fashion is 14%. You know, the built environment is 30 to 40%. So there's whole set of other things we can do. Insulating our buildings would have massive more effects. So even if we were to buy everything kilometer zero, all our food and those things, those are single percentage points of a problem, which there's a lot of other things that one can do. Doesn't mean we shouldn't solve them. But as an engineer, which is my basic training, I think there are bigger parts of this system that we can start to think of and address. And it doesn't mean that this is not serious. For our own practice, we could make it, and run it on green power. And remember, the world is radically shifting. We've done energy scenarios back. And when we did that, 2011, I think, renewable solar power was a 100 times more, than coal. Today, in the best of times, in a 190 out of a 193 countries, solar power, photovoltaic solar, is the cheapest source. The sun, which is hitting us now, is the cheapest source of new energy in all but three countries. The other one is wind for the remaining three. This year, so last year, 2025, we shipped 600 gigawatts of photovoltaic, capacity is shipped. That's 600 nuclear stations built in one year, when the sun is shining. Not all the time, but when the sun is shining. 600 nuclear stations. This year, it'll be more than that. Around the globe? Around the globe. Yeah. I mean, I imagine China is probably huge portion of that. China is is a significant portion of that, but it's not, you know, there's a distance between rhetoric and reality often. So out of the 600 gigawatts, 90 of them are made in The US. So one out of six is actually made and deployed in The US of photovoltaic panels. Very different from the rhetoric we would hear. 45 gigawatts of that is, made from sand in The US all the way through crystals, the way... For right all the way from sand into panels. The entire body of of GPUs that is running in The US is, 15 gigawatts. You know, all of Europe is one gigawatt, which is a dramatic difference. That's all of the GPUs running. Now there's other things and there's questions on that of what it means because it disrupts systems as they are, as they're transitioning. I'm not saying everything is perfect about that at at all. And I'm saying many of the sources that are being used needs much, much different practices and better practice, and we're working hard at that. But, there are often in this space bigger fish to fry, for a global transition that we make, and we would be better held as as a world of looking at what are the complex things that we want to change rather than the things that we want to signal we're changing.

It's not gonna help you do your homework a little bit better. That's what they tend to... Lawyer's letters or any of those things. Yeah. Yeah. No. I mean, so the... I think this is, like, one of the things that is often missed in these conversations,

technologies tend to diverge, not converge. So the internal combustion engine... And it doesn't mean perfectly for all technologies that are kind of isolated cases. But in general, the internal combustion engine became the motorbike, the speed car, the sport utility vehicle, the the the the station wagon, the family car, the the the the fast motorbike, the the the the boat, all of these things. So technologies tend to diverge to purposes. And for us, what we want to do is the most important challenge for humanity is climate. Let's build that thing and enable that so that a city can make slightly easier plans for their carbon transitions so that companies can make easier plans so that countries can make easier carbon policies. I can discuss about all of those type of things so researchers can work a little bit better and have that as small footprint as we can for outsized quality. And by... When we started it, the idea of building a domain specific model, which was done, for COP twenty eight, at the time, so in Dubai, was a radically unusual idea. We're starting to see more and more parties doing domain specific models. And as agentic AI starts to become a normalized process, what one can see is that those kind of questions are asked of specialized models doing specialized things as part of a broader agentic flow. And what we wanna do is make sure that the Earth and climate and our change is part of that discussion. Hence, we make it open, and freely available there, as a just normal open play. Because I think, any barriers to entry that we can remove, in this is is critical because this is humanity's greatest challenge. Yeah.

So the Internet Archive, none of this would have, you know, been possible without a set of very gracious grants, that we've received on this. And it's it's largely... We're funded through grants and to enable it to be open, on this. And so the Internet Archive has supported us along the way. And and in Europe, there's a special provision. So under, the European Copyright Act, there's the text and data mining exemption. And so as it's a foundation, so the Climate GPT Foundation, what one could do is you can also do data analytics on copyrighted materials. So we try in our, to pull in the best of science that we know in there. So it has policy clarity here, which opens up bringing our best of science because we don't wanna build it on last year's science or the fifteen years ago, etcetera. We wanna build it on the best knowledge that we have, in terms of this. And so, we've limited the RAI corporate to open access, open access journals, and open access science. But increasingly under the TDM, one can increase that and start to add, more copyrighted materials so that we get the best, material going into our decisions. So... And and, you know, the Internet archive and particularly, Bruce Terkel has been a mentor and a supporter, of us, which enabled saying there are this incredible library of knowledge there. What... How can we use that library to solve some of the deepest challenges that we have? You know, climate is one of them. Educating the next generation one can go through a whole list of those things, and I think it's one of the treasures, of... On Earth. And and, personally, I think it should be a world heritage site, to be honest here.

Yeah. I I I think the archive, you know, in its practices and and its collections, it's it's really, I mean, in many ways, doing God's work. You know? And and in many... You know, how do we get access to those microfiche materials? How do we build this? And, we're at a moment when, you know, next week, I'm I'm at the Vatican where we're discussing sort of AI policy, and response. To the pope. Well, it's it's part of a larger process there, but it's a it's a very complimentary. So we'll be large group of Nobel Prize winners, and the Catholic church has taken a lead on this and say, you know, we're at the point of, significant changes in how we think, how we go about thinking, how we learn, what we should be learning, and also the safety of these technologies as they're, become part of the human fold, here. My view is as an AI builder, as a model builder, is very much... I think it... It's useful to think of these things much more as big statistical machines. You know, there's a humility in building these things, but it's also we're not really artificial intelligence or artificial life. We're engaging as big statistical machines as part of sometimes automated processes where they're making decisions on our behalf. Sometimes, those things will go away. We we know that's happening. And what we want to be is have the Earth as part of that conversation. Have the limits of the biosphere as part of that conversations because agents can be incredibly exploitative on this. They can, you know, take all the copper, in Zaire independent of the consequence. And unless we put the earth and those things as part of that conversation, human and, and artificial, we stand at tremendous risks. And so having a history of our best knowledge, which is digitized, which is what the Internet archive... You want to have the system builds on the best we have and perhaps not Reddit posts only.

is something that we need to be doing. Look. It's it's... There are different ways of thinking about climate change. So if you look at the inscription devices we have now, so, we have, scientific processes modeling Earth physical system. We have scientific processes modeling, carbon stocks, you know, tracking carbon stocks around the world. Do we have all of those processes are highly mature of how carbon... How we put carbon in the atmosphere, how we're doing that? And we we typically frame that in economic terms, so as a tragedy of the commons type of thing. It's in the commons. We have these externalities. We're not doing that, etcetera, etcetera. That's one framing. Alongside of that, there are other ways of framing this. And one of the interesting principles coming out of COP was that polluter pays. So the bulk of the problem with the carbon in the air is really produced by Europe, United States, and China. So if you're in Bangladesh, your problem isn't externalities. Your problem is that there's been pollution. So if we're sitting in San Francisco Bay and an oil tanker breaks apart there and start leaking oil, everybody points and say, the oil tanker needs to clean this up. That's the polluter they need to pay. So these debates have multiple ways of thinking about that. And so framing it from a polluter pay, which is the principle of COP twenty eight, suddenly you start to say it's a liabilities problem. It's not an externalities problem. It's a liabilities problem. Who pays, for the damages that pollution is creating around the world, particularly in vulnerable communities throughout by the hundreds of millions of people. How do we create those processes, and how do we get that carbon out of the air? What are the... And the surprising thing about the club of roam work, the Earth for All is it's doable. We can get the carbon out of the air. It's very very viable, to do that. Not through capture or anything like that, but through the very thing we're sitting in, the forest. This very thing or the breathing, every breath you take, you know, inside that breath, there's a little, you know, just a couple of molecules, but of the last breath of, Socrates. Like, it's right there. You've just taken that breath. Anybody listen to this? I can feel him inside me. That'd be a great place. But every breath that you take when you're listening to this, you know, when you're listening to this, you're breathing some of the last breath of Socrates. We all share this planet together, and these trees are recycling nutrients, and in this case, carbon. They're cycling and fixing that there. We've got the sun, you know, nuclear reactor right above us, That's doing all of the fixing, providing the energy for all of the fixing right here. And by creating natural system restoration where those natural systems function, and we can go into that in more detail, are imminently viable pathways of fixing carbon. Doesn't mean that it's on time, doesn't mean it's gonna be easy, doesn't mean it's there, but these are very, very viable pathways, that we should be undertaking now.

Wonderful. Thank you for having here, and it's a great series.