Can AI predict (and control) the weather? w/ Dion Harris and Tapio Schneider (Transcript)

[00:00:00] Bilawal Sidhu: 

Imagine fusing together massive amounts of data taken from thousands of satellites, ground and sea level sensors that have been tracking the pulse of our planet's climate and weather for decades. Now, throw in some physics and generative AI and what do you get? A digital replica, a planet earth, a real life crystal ball that shows you exactly what any street on earth looks like at any given time.

[00:00:29] Dion Harris: 

You can play around, you can drill in, you can understand what was happening in 1972 on this specific day. Well, what's really exciting is to see people who aren't climate researchers, go and, you know, literally zoom in on their neighborhood and see what was happening, or see what the foliage looked like at that time, and, and understand what's really, you know, happening.

[00:00:48] Bilawal Sidhu: 

This is Dion Harris from NVIDIA, the company that's been creating a complete digital twin of the planet, appropriately called Earth-2. So I can use Earth-2 to travel back to my current neighborhood in Austin, Texas in 1972, I can go all the way down a street level and see the oak trees lining the blocks, the green expanse of Zilker Park.

But you can also see the specifics of what the climate looked like, the air quality, the weather systems, the monsoon patterns, and how they all interact. But even more importantly, we can use these complex climate models to travel into the future and visualize multiple outcomes in extremely high resolution and do it all really, really fast.

Something that traditional climate models were nowhere close to achieving. 

[00:01:38] Dion Harris: 

Roughly about, 45,000 times faster in terms of creating an actual forecast. It's not just one forecast that makes it interesting. It's being able to do thousands of forecasts that can give you a much better representation of what the likely outcomes come out.

So just to give an example, we are working with the Essential Weather administration of Taiwan. And they're often hit by, by typhoons that are coming, you know, inland. And so what's interesting about Taiwan is it's an island and so you have to be able to, to understand how and when you can relocate. But their relocation options are somewhat limited.

And so by giving them more granular understanding of where and when typhoons will, will hit landfall, for example, they can then model and quickly build, you know, reevaluation or relocation, um, programs based on that data. So, having more resolution gives them more information quicker about specifically where and when you know areas are gonna be impacted.

[00:02:41] Bilawal Sidhu: 

So it's safe to say that AI technology is ushering in a new era of powerful tools to predict and respond to climate change, but it goes beyond that. Geoengineering is giving us the ability to intervene and control the weather. Does this mean that AI could actually solve climate change? I'm Bilawal Sidhu, and this is The TED AI Show where we figure out how to live and thrive in a world where AI is changing everything.

Now, I don't know if I buy that AI is a panacea for the climate crisis, but these models are an amazing example of how AI will be transformative in the climate space. So I'm feeling very hopeful and excited. This is where we ought to be putting our AI technology to work. But with every technology there comes great challenges and unintended consequences.

And because it's earth we're talking about, we have to tread carefully. The stakes here are massive. Our guest today is climate physicist Tapio Schneider, who's gonna talk us through the promises and the perils of this new era of climate modeling. So Tapia, you've been working in climate science for a solid 30 years now, but I wanna go all the way back to the beginning.

What attracted you to the climate space in the first place? 

[00:04:11] Tapio Schneider: 

As a physics student, I was always interested in, in the physics of everyday life. Um, you know, how does a refrigerator work and how does a transistor work? That kind of physics I found absolutely fascinating. And as I progressed as a physics student, I realized that the physics I was doing and learning was increasingly a bit further removed from daily life, and I, I decided I want to work on physics at the energy of sunlight. 

By definition, that's sort of the daily life physics, and that's how I got got to a climate as a part of physics, I must say, it did play a role in my decision making that it, it matters to people. It was clear already then, 30 years ago that global warming is happening and will impact all of our lives, and that that was a factor as well.

But the primary motivation was wanting to understand this incredibly complex system. 

[00:05:00] Bilawal Sidhu: 

What a complex system it is. Indeed. 

[00:05:08] Tapio Schneider: 

Yeah. 

[00:05:03] Bilawal Sidhu: 

I'm curious, what is the, what has been the trajectory of AI into the work that you've done? Right. Obviously, statistical analysis has been a thing. We've had good old fashioned AI as well.

[00:05:12] Tapio Schneider: 

Right. 

[00:05:13] Bilawal Sidhu: 

And now of course we've got this generative AI deep learning wave that's happening. I'm curious, what's been the trajectory of AI? 

[00:05:19] Tapio Schneider: 

I started out working in a biophysics group in the, uh, first wave of neural networks in the, uh, 1990s. So I had some early exposure to the early days of machine learning as it was then.

But what really fundamentally change for me was when I decided to work on more smaller scale processes, which are the processes that are most uncertain in climate models and wanting to use data much more extensively than they have been used before. I think that's when I really started to think about how we can use data.

I started collaborating with one of my colleagues at Caltech, Andrew Stuart, trying to formulate ways we can use data well for climate purposes, which is quite different from many other, um, applications of machine learning. 

[00:06:07] Bilawal Sidhu: 

Can you talk a little bit more about these data sources and how those have been evolving?

What's the kind of data that's most useful in your applications?  

[00:06:14] Tapio Schneider: 

So the true age of satellite observations of earth atmosphere, oceans started around 1980. And since then, data volume has, uh, has kept increasing, increasing exponentially. Right now we are receiving from NASA alone about 50 terabytes of data every day from space alone.

And in addition, we have sensors, autonomous vehicles and oceans and the like. And it's truly, it has become a very data rich field. The, the climate sciences in ways that they were not 30, 40 years ago. So the way the data that we have are most commonly used right now is for weather forecasting. So whenever you get a weather forecast, what has happened before you get it is that a weather forecasting center has assimilated all the data we have.

We use that as initial condition for forecast. This is a, that really to a big jump in the quality of weather forecast that many people don't quite appreciate. It's about 25 years ago. Something called four D variational data simulation. It led to a big jump in the quality of weather forecasts we have.

What's interesting is that in the climate space, so when you think about, now let's run a model that's a bit like a weather forecasting model, but run it for decades or centuries and say what will happen decades from now? 

[00:07:32] Bilawal Sidhu: 

Mm. 

[00:07:27] Tapio Schneider: 

Their data have been used much, much less extensively, primarily to evaluate models to say, this model is good or bad after the fact, but not directly to inform the model, and that's the piece that we want to change in in this Climate Modeling Alliance. 

The Klima project I'm leading is use the data directly to inform the model to achieve higher quality of predictions and projections for the future. 

[00:07:52] Bilawal Sidhu: 

Okay. To make this a little bit clearer, basically Tapio is advocating for using all this observational data that's just sitting around, not merely to evaluate the models, you know, based on the predictions that they produce, but to train the models themselves so they keep in mind all that historical observational data to do better climate modeling. 

So I'm curious, Tapio, has the impact of these recent AI developments altered your expectations for what's gonna be possible with these models that are specifically focusing on, you know, climate predictions on these longer time horizons?

[00:08:29] Tapio Schneider: 

Yeah, definitely. It's maybe useful to talk a little bit about how these models are being developed and what 

[00:08:33] Bilawal Sidhu: 

Sure. 

[00:08:34] Tapio Schneider: 

What a day in my life used to look like, these models are essentially solving Newton's laws and the laws of thermodynamics on a large scale, on a global grid. The problem is that these, the meshes of this grid, they have a size of somewhere between 10 and hundred kilometers, is typical today, and there's a lot of stuff that are much smaller in scale than the mesh size of a climate model.

[00:08:59] Bilawal Sidhu: 

Okay? So what Tapio is saying here is important. So let me break it down. Imagine the entire earth as a giant puzzle. Climate scientists are trying to solve this puzzle, but they only have very large pieces to work with, each piece representing an area of 10 to a hundred kilometers. They use these pieces to build a picture of how these climate systems work, fitting them together and seeing how they interact.

The problem is there are many tiny but important details like clouds that are way smaller than a single puzzle piece. It's kind of like trying to see a butterfly in a puzzle where each piece is the size of a car. In other words, the resolution of traditional climate models is way too fuzzy to be able to discern these tiny important details in clarity.

Thus creating a lot of uncertainty. And as you'll hear from Tapio, AI is able to mitigate this problem. 

[00:09:53] Tapio Schneider: 

And so what you have to do is find some empirical way of representing what clouds do, given what you know on larger scales on, on the mesh size of the climate model. And that was a pretty tedious process and it has been reasonably successful, but in this process and the lack of complete success of this process, like pretty much all uncertainties in climate predictions and what machine learning tools change is that for these small scales, now we can learn what they do from data. 

[00:10:31] Bilawal Sidhu: 

A lot of companies that are delving into machine learning are, are experiencing this issue of they're not really grounded in the physics of the real world.

Right? I mean, just to give you an orthogonal example of video generation, you know, a person's running backwards on, on a treadmill, right? Or a glass is breaking and it's behaving like plastic. 'Cause you know, this model doesn't understand sort of the cause and effect relationships and sort of the, the rules of physics that govern that environment. What's the type of work that y'all are doing to anchor these, these predictions that you do, uh, into the physics of the real world? 

[00:11:04] Tapio Schneider: 

Yeah, that, that's, it's a good analogy actually. I mean, what you don't want as a climate model that hallucinates physics, right? And for, for a video or even for a weather forecast, if there's something wrong, well it looks funny and you're correct it right, uh, for a weather forecast if it's wrong, too often we don't trust it and go to a different source.

The additional challenge of course for climate is that you do not have an easy validation case. You do not immediately know when something goes wrong, if it takes years or even decades for these changes to become manifest. So what do we do to deal with this problem? What we do is we use the laws of physics that we know and embed machine learning tools inside the laws, inside conservation laws, and that gives us, if you wish an insurance policy that what we produce is is reasonable and even more to the point. What we want to predict is something we have no data for. We don't have data for the future, and the future can be entirely different from what we are currently seeing. This is known as the out of distribution challenge, machine learning.

So by using physics as far as we can, it helps with this out of distribution challenge as well.

[00:12:18] Bilawal Sidhu: 

That's a really good point. Um, obviously the future is unknown, but this is like as close as we can get to a real crystal ball. So with advances in, you know, uh, we're obviously, uh, sensing the world in, in greater fidelity, in greater frequency.

We've got these, we've got this beautiful tool that is machine learning, and you're anchoring it in the laws of, of the real, of physics that govern the real world. I'm curious, what are your hopes for what these models will be able to do, uh, in the very near future? 

[00:12:49] Tapio Schneider: 

Yeah. So maybe let's start from what we need and then say how we get there.

I think what we need is, is assessments of risks, extreme weather, extreme climate risks for the next few decades. Climate is changing. Uh, we need to mitigate as much climate change as we can, but some climate changes unavoidable. We need to get ready for what is coming and build our infrastructure so that they're right sized and cost effective for the world we are all inhabit 10, 20 years from now. So you need two things, you need to reduce model errors and you want to quantify uncertainties errors so that we can take them into account and planning decisions. If you're an engineer building stormwater management system, you don't want to just know the mean rainfall or any kind of expected values, but you want un uncertainty ranges, you want risks. The biggest uncertainties in climate modeling come from these small scale processes. For example, clouds dominate the uncertainties in climate predictions. We don't quite know what clouds will do under global warming that dominates uncertainties and how climate will change.

And the machine learning tools we are talking about, I think have huge potential there. One concrete example that turns out to be important is how does a cloud exchange air with its surroundings through turbulence? Turns out to be it's one of the key controlling factors for how clouds behave in a global warming, that process is hard to measure.

Um, it's even hard to simulate precisely and infer from simulations, but there are machinery tools that allow you to indirectly infer what that process looks like. That you can then use in a climate model and achieve large error reductions. Our colleagues at MIT have developed an ocean model, a similar story there, a small scale turbulence in an ocean.

You can, you can similarly learn from data how to model that in the context of this large physical model and then we are still talking about perhaps order 10 kilometer scale resolution. You still need to get to this kilometer scale. And here's another really good use of AI tools. It's for downscaling or super resolution.

If you wish that you fill in the details that the climate model does not produce, using data we have for the present climate, climate projections for a future climate to produce the localized, uh, climate risk assessments that we ultimately need. 

[00:15:13] Bilawal Sidhu: 

So you, you brought up really two points. One is, rather than getting this like macro scale picture, we need to give decision makers and local authorities this micro scale picture so they can, it can be a lot more actionable.

And then as we reduce the error, the, the quality of predictions, the accuracy of the predictions, we'll obviously go up. Will there be a feedback loop there where basically machine learning will allow us to get better at predicting the future as we collect actual data, or will that window keep moving out and the future will always be unpredictable?

[00:15:45] Tapio Schneider: 

No. We will get better and better at predicting what will happen, at least say for the next 20 to 50 years or so. I think that's a good time horizon to focus on. We should be able to provide good predictions. The reason I choose this timescale is that if you think longer term. Um, uncertainties in what we as humans do will start to dominate.

How much CO2 will we emit? And that's obviously not something you can model from first principles. So that becomes a conditional for providing scenarios. Um, but for the next few decades, the uncertainties we have are dominated by model uncertainties and then by just the natural chaotic variability of the atmosphere in the oceans.

So the model uncertainties, we should be able to reduce dramatically, and then there's just no way to do this locally alone, the globe, it's all interconnected, but to get the local information, then you need these downscaling tools, risk assessment tools. So you need to build a value chain of models that are interlinked, and in the end, you want people to use all that information.

So that's one thing to say, “I have a fantastic diffusion model for downscaling.” But it's quite a different story now to give this an answer of people who need this information, a small town planner. So you need to build good user interfaces that make it very easy for stakeholders to access this information in their decision workflow.

[00:17:09] Bilawal Sidhu: 

I love your point about good user interfaces, and it actually begs the question, um, how do we translate these insights that you're getting from these models into climate action at that local level? Can you paint a picture of how those local decision makers would engage with this type of data? 

[00:17:27] Tapio Schneider: 

So climate action, I think we need to distinguish two pieces.

There is mitigation, so reduce emissions and there's adaptation, adapt to whatever is coming. I think for the mitigation part, that's largely policy problem and the technological progress problem. In some ways, we know enough about the climate system. Additional information is probably not gonna change the picture there very much, but for the adaptation part, any public private sector organization that makes any decision of, of reach of a few decades will have to adapt to climate change. 

A municipal planner will want to plant storm, storm water management infrastructure. That's one type of information they need. Uh, information on precipitation extremes, decades from now.

Um, an architect building designer will want to build a building in which it's still comfortable to be inside a few decades from now. So they want to know, um temperature, probability of temperature extremes, prolonged temperature extremes for some time at, at the downstream end of this value chain chain, what you need is an ecosystem of tools that caters to different sectors specifically and meets people in their decision making process. 

Um. I think once you have those tools available that will trigger action, right? Flood protection is of course, another good, good example. Um, rising sea level increased, uh, risk of storm surges with rising sea level.

You want to know what those risks are and then proactively, um, design your levies, your built infrastructure accordingly. I think that's, that's a kind of climate action that will be triggered with better signs and better information on the risks. Again, there's the whole mitigation side, which is hugely important of course, where these scientific information, you know, it's important, but in some ways we have what we need.

We know we need to reduce emissions. 

[00:19:18] Bilawal Sidhu: 

Exactly. I mean, it's almost like we know what action we have to take to prevent climate catastrophe, and we have to do it now, but we still have a hard time sort of visualizing the impact of that action or really even the lack of that action in a very concrete way.

And so climate change obviously has become really politicized or polarizing as a topic. Um, but maybe when you have that model in front of you, it might become more clear cuts. Do you see a world in which these models and applications built on top of them would make these risks seem more tangible and more real to people, and thus more important to address as a society?

[00:19:58] Tapio Schneider: 

I, I want these risks to be tangible and, and easily accessible to people having say apps for consumers where people can contextualize what's currently happening and put it in context of future risks. I think it would be tremendously helpful in, in raising awareness for, for what is happening. You mentioned polarization, I have to say, as far as the climate change questions are concerned, it, it is still clearly polarized, but I would say polarization is decreasing. 

Um, there's just a reality that climate is changing at this point, if you're a business and you make any decision that has a reach of a few years to decades, you would just lose money if you don't think about climate change.

And that becomes the lowest common denominator that people can agree upon, that they have to worry about it. I expect the polarization to further decrease on this issue simply because you can't deny that climate change is happening. And now it's just a question, what do we do about it? 

[00:20:57] Bilawal Sidhu: 

It it, it's amazing that economic incentives are aligned here, right?

[00:21:00] Tapio Schneider: 

Yeah. 

[00:21:00] Bilawal Sidhu: 

And like to your point, 10 and 20 years are a timescale, which impacts all of us. It's not this nebulous hundreds of years from now. 

[00:21:08] Tapio Schneider: 

Right. 

[00:21:08] Bilawal Sidhu: 

Thing that we're planning for, which makes a, a huge difference. I, I think before we get into, uh, mitigation, I do want to talk very quickly about like AI is like the panacea for everything.

There's this like, whoa, how, how do we solve this problem? Whoa, AI, of course. And then AGI will magically come along and solve everything for all of us. Uh, in your mind, what can AI not help us do from a climate perspective? 

[00:21:32] Tapio Schneider: 

I think the potential for modeling and the like is huge for anything involving software, but it's important to keep in mind that we live in a very material world.

Mitral fertilizers are produced by, primarily by the Haber Bosch Method, which relies on natural gas and uh, a large source of emissions, steel production, cement production, right? These are all enormous industries. Putting out massive amounts of material. Those things are not so easy to change quickly.

There's, of course, aviation is also a good bit harder. Aviation is 2% of global emissions. It's, it's important, but if you can deal with everything else, that's already pretty good. But everything else is, there's still a number of, uh, fairly recalcitrant problems there. So I think AI may not be the magic bullet for all of that.

But that being said, it, it can play a role in finding solutions now will the next large language model find us a, a way of producing nitrogen fertilizers that doesn't involve using natural gas and leads to greenhouse gas emissions? Well, we are definitely not there yet. 

[00:22:33] Bilawal Sidhu: 

There are folks out there that are advocating for, um, you know, more like stronger intervention.

Right. And so there's been, uh, recent popularity in cloud seating. In fact, I'm aware of, uh, um, you know, like 23-year-old kids who are doing startups that are going gung ho about this in El Segundo. And so I'm kind of curious, you know, you, you outlined the complexity of this problem of modeling this phenomena, weather phenomena at this, like it's very hard to model some, a system this complex, but does that end up having some like ripple effect in some other part of the world?

Like do we understand that? I'm kind of curious what you think about these other forms of climate inter intervention.  

[00:23:13] Tapio Schneider: 

The, the climate interventions on different levels from speculative to things that can actually be done. Um, seeding clouds to make it rain in a given place. That has been tried since the advent of weather forecasting in effect, that was the original hope when John Von Neumann and others started weather forecasting programs in, in the wake of World War II that it would lead to weather modification. 

By and large, these programs have not been successful. The other way in which you can seed clouds is, is under the heading of geoengineering, where you just change the cloud cover of Earth, especially low clouds over the oceans that will reflect more sunlight and that will, um offset some of the warming that comes from increasing greenhouse gases. 

Um, there are various ways of offsetting warming from increasing greenhouse gases, seeding clouds, it's fairly speculative. It may work what almost certainly is possible is put aerosols.

So little particles could be a sulfate or silica particles into the stratosphere with rockets. Um, they would reflect sunlight. It would lead to cooling, offsetting some warming that seems technologically feasible. It's, uh, relatively clear that that, that it would be doable to do this. Um, you would, you could offset the warming.

What you could not offset easily are precipitation changes that go along with global warming. So the real challenge with these kinds of scenarios becomes who is in charge of controlling climate globally? You know, some warming in some parts of the world may be quite desirable for agriculture and the, and the Arctic say, um, so people may not want that warming not to happen. 

Um, of course, by and large warming is not desirable. Um, it has all sorts of at attendant risks, but then if you do geoengineering, you might change. Uh, the monsoon rainfall in India and that will have severe implications that, uh, people there obviously wouldn't like. So there is a governance problem, an an ethical problem who's in charge

Um, there's the obvious moral hazard problem. So suppose we find ways of, of setting some warming. Does it give us carte blanche to keep emitting there might the, the risk if you do this for a while, any kind of geoengineering is that you are offsetting the warming. Say you do it for 30 years, you offset all the warming that would've happened in 30 years, probably something at 0.6 degrees centigrade or so. But for one reason or another, you miss the fix of injecting aerosols into the stratosphere or seeding the clouds, and then you'll get all the warming that you offset over a timescale of a few months in one bang.

That's which is what's called determination shock. 

[00:25:59] Bilawal Sidhu: 

So for those of you who haven't heard the term before or read the book, termination shock is kind of like putting the planet on a climate change painkiller if we suddenly stop the treatment, all that warming that we've been masking hits us all at once, kind of like withdrawal, but for the whole planet, it's the shock of terminating our quick fix. Hence the name. 

[00:26:24] Tapio Schneider: 

So there is another governance ethics hazard in that as well. I think it's good to have that discussion on, on a society level cloud seeding as a research program I, I I support because it, it's really one of the big uncertainties in climate protections is what will happen to clouds or how does pollution affect clouds more broadly.

And at the very least, these programs will elucidate that question. Whether we should do this on a global scale? Well, my personal take is at least to be extremely cautious, right? We are messing with the system, we don't fully understand, uh, whatever we do, we'll have global implications. We don't have effective governments mechanisms that seems very difficult. 

You could find a globally pptimal solution if you can globally agree on the loss function to minimize. 

[00:27:12] Bilawal Sidhu: 

Totally. 

[00:27:12] Tapio Schneider: 

Um, but then I think that would be possible. That, but that's a problem, right? I mean, we won't globally agree on the loss function. The, the objectives for different countries, different stakeholders will be very different.

[00:27:23] Bilawal Sidhu: 

Case in point, China and India are already saber rattling over China's weather modification program. India's worried it's messing with their monsoons and the rivers that they share across their borders. Which no surprise could have an impact on agriculture and food security. Well, I certainly hope that we come up with a globally optimal solution, and it doesn't take an eccentric billionaire to just go, you know, uh, uh, have carte blanche and just start doing this.

Um, yeah. Uh, without the consent of this like planet that we inhabit together. I'd love to change gears a little bit just on, um, the impact of AI systems, right? Like there is on, in terms of power consumption. How do you think about this race towards more data, more compute equals more intelligence, we must do this and the climate impact that this, uh, uh, this sort of race is happening, having.

[00:28:19] Tapio Schneider: 

Yeah. The climate sciences have been big users for supercomputers since they exist. So we are big electricity users by implication. And sometimes people joke, we should just take the, the, uh, attendance, CO2 emissions that come with climate simulations into account straight in the simulation. It's, it's a joke.

It's tongue in cheek. I mean, it's not that big in a global scale. Right. Um, but of course it's a serious concern. I mean, the, the silver lining here perhaps, is that we're getting ever more compute for the same amount of electricity energy use. So that's the good news. Of course, training AI models right now is incredibly expensive, given that it is so expensive, you can simply scale it with bigger computers indefinitely. 

So we probably have to find more energy efficient, more data efficient ways of achieving what today's big models achieve. And I think that's the new frontier built, built models that are more energy efficient, smaller, um, and just as good. And I think that's where we'll see a, we will see a lot of progress in the coming years.

That would be my expectation. 

[00:29:28] Bilawal Sidhu: 

Not to get too science fictiony here, but like, you know, data centers are already very resource intensive. We're gonna keep producing more data to, as a society, uh, that's certainly not going away. How do you see, sort of like this, this industrialization, the next evolution of industrialization, technol technology getting built into the type of climate models that you're building?

Like, are you accounting for this already? Like, or, um, you know, how, how does that, how does that work? 

[00:29:53] Tapio Schneider: 

When we do climate simulations, we take emission scenarios as given. So there are some economists, social scientists, scientists and the like getting together and mapping out several plausible futures for emissions, how much technological progress is expected and the like.

And what we take from that are these scenarios and then we make climate projections conditional on those scenarios. The unambiguously good news is the rapid decrease of the cost of electricity that is renewably produced, the cost of solar power has decreased by almost a factor of 10 and roughly over the last 15 years or so.

So, what we need is renewably produced electricity and that's increasingly feasible and that then can power large data centers. So I think it is a science fiction scenario where we would have to worry that, that the data centers are kind of eating our climate future. I think I'm more optimistic there, that in fact they'll be renewably powered, uh, getting more efficient and we can sustainably compute for whatever we need in the next few decades.

[00:31:06] Bilawal Sidhu: 

What is your vision for the future? Right. Um, you know, on one hand a lot of people view AI itself as a very polarizing subject. Some people are extremely optimistic about it, to the point where people are like, “We must keep accelerating, accelerate, or die.” On the other hand, people are like, “Oh no, we gotta pump the brakes.”

Are we deploying and proliferating this technology far too quickly? How do you think about this conundrum and, uh, on the spectrum? 

[00:31:29] Tapio Schneider: 

I think my own work is, it, it's certainly not representative, but I think it is indicative of how AI can be hugely beneficial, I think in, in the, in the long run for many. For us, it really takes a lot of the drudgery out of the day-to-day work.

We can learn functions from data that before we had to guess by hand, and it was tedious. It's increasingly becoming a very efficient tool, so, I think I'm extremely optimistic and I think in the long run this will increase productivity and I, I see some of that in my daily life now, and I'm very optimistic about that part.

Um, of course it does mean, you know, some jobs will become less important, others more important, they will be winners and losers. Just as they were at the beginning of the Industrial Revolution, that will happen again. I think I see huge potential. I think I see huge risk not using that potential and setting us back.

I'm less worried about, say, computers taking over those kind of scenarios, they're, they're not the large concern for me right now. 

[00:32:35] Bilawal Sidhu: 

What are you looking forward to? 

[00:32:37] Tapio Schneider: 

I mean, for me personally, of course, is I want a, an amazing climate model that reduces uncertainties, quantifies uncertainties, and then build verticals on top of it that goes to local scale information, that goes to, to apps that consumers can use to, to assess climate risks to their own properties, to, to their weekend plans and the like. 

Um, I, I want climate information to be permeating economic decisions in a rational and, and effective way, and I think it's achievable.

One easy one is anyone who purchases property, right? You'd like to know what the risk of flooding and wildfires in, in that area are. And, uh, like to know that there's granular accuracy and in a way that you can trust, and I think that's achievable. Um, and just simply, you know, when you talk with your friends about the weather today, it's an unusually hot day.

Have contextual information. How unusually hot is it in the past? What is this gonna be like 10, 20 years from now? Just inform daily discourse with, with that type of information. I would find that incredibly helpful too. 

[00:33:45] Bilawal Sidhu: 

Tapio, thank you so much for your time. It was a pleasure talking to you, and I'm really, really, really, really enjoy this conversation.

[00:33:51] Tapio Schneider: 

And me too. Thank you. Thank you, Bilawal. 

[00:33:52] Bilawal Sidhu: 

So after my conversation with Dion and Tapio, there's a couple things I just wanna stop and marvel at. Number one, our planet is covered in sensors. Let's appreciate that for a second. It's not just satellites and space and underwater drones. It's lidar measuring aerosols and clouds. It's a radar measuring ice sheet thickness.

The boatload of sensor data out there constantly monitoring this planet blows my mind. And only now are we starting to fuse it all together. And number two, we're building some predictive climate models using the same AI technology we use for sillier, whimsical things like AI, art generators, or 3D video games.

That's just cool and a big reminder that it actually matters what we do with technology. Speaking of technology, if these models are accurately predicting climate disasters, we may be tempted to use geoengineering to modify the weather, but oh boy, does something like this require global coordination in lockstep?

Otherwise, I see a new kind of geopolitical crisis brewing in the future. Literally the weaponization of weather. Instead, we should think of this crystal ball as a sandbox for scenario planning, generating a model of what the earth could look like in the future. If we make a set of decisions and put them in motion, we can use it like a canvas for global coordination.

So if we continue with the current path of excessive emissions and energy blindness. The models will map out just how dire the consequences will be. But hey, the future isn't fixed. And these models can also show what other futures are possible if we take intensive coordinated action to mitigate some of the harm caused by the climate crisis.

That is if we shift away from fossil fuel, towards clean energy sources. If we work to reduce emissions and preserve biodiversity, these next generation models allow us to travel forward in time and visualize the impact of our actions to make a more tangible vision of the world we actually want to live in.

The TED AI Show is a part of the TED Audio collective. And is produced by TED with Cosmic Standard. Our producers are Elah Feder and Sarah McCrea. Our editors are Banban Cheng and Alejandra Salazar. Our showrunner is Ivana Tucker, and our associate producer is Ben Montoya. Our engineer is Aja Pilar Simpson.

Our technical director is Jacob Winik, and our executive producer is Eliza Smith. This episode was fact-checked by Dana Calacci. And I'm your host, Bilawal Sidhu. See y'all in the next one.