AI & Climate: Sorting Signal from Noise — Living Between Worlds 7.2

Welcome to Living between Worlds for February 18, 2026. I'm Gil Friend, CEO of Natural Logic and Ontological coach to purpose driven execs and entrepreneurs. And today on this seven year running conversation for possibility, we're going to try to sort signal from noise on the AI climate landscape. You're all here. You know, there's been a lot of buzz about AI in many dimensions. You know, is it intelligent? What does intelligence mean? What impact is AI having and will it have on human effectiveness but also on humanity itself? How useful is it? How do we deal with the issues of intellectual property? What about the investment bubble? Legendary investor Bill Gurley in conversation with Tim Ferriss just recently speculating on whether this is a bubble or not, said, you know, disruptive change looks like a bubble even when it's not. So here in the middle of it and hard to tell also from investment perspective, massive amount of capital going into data center construction. It's a FOMO phenomenon. Fear of missing out, but so significant that the capital expenditures into data centers are the only thing keeping the US economy afloat at this point, far outstripping other capital allocations in a really striking way. And then to the climate and energy question, of course we have the massive energy hunger that we're seeing from data centers, disrupting grid and local economies and energy prices. The climate impact from data centers if they're running on fossil based energy. Interestingly, something close to half of all new data center construction is being built with energy behind the meter. In other words, not relying on grid. It takes too long to get grid interconnections given the speed of the, of the demand. Some of what's happening behind the meter is fossil gas turbines and so forth because those can deploy faster than grid. Although still supply chain issues there. Some of it is renewables and storage. Blue Planet Energy Disclosure company I'm on the board of is in the storage business. So we're very interested in the data center universe. And so there is promise and peril here. And Marshall will talk with you about some of the observations that have been coming from that realm. All of these issues are important and all of these are connected. Today we're going to focus on the energy climate piece of this. Not the intellectual property piece, not the regulation, not the, you know, not the many places we could go on this. But that piece AI energy appetite is surging. How we satisfy it or deal with it is going to shape the climate for decades to come. You know, the energy, some people would argue, and Catherine Hayo, climate scientist among them, would argue that AI driven emissions reductions in just a few sectors could more than offset AI's own footprint. Other people very much dispute that. My general perspective, and I'll be brief about this for now, is that I'm less worried about AI and energy than some people are. We're early on the technology learning curve. We often find early in technologies they're far less efficient than they become eventually. We see this with solar, we see this with batteries, we see this with EVs. So we're early on the technology curve. But the design decisions being made right now in data centers and supply chains and energy systems are going to determine whether AI accelerates the crisis or helps us solve it. So with that in mind, I've invited Marshall Kirkpatrick to join us today to help explore this landscape in a very new way, which you're going to be among the first people to see. Marshall Kirkpatrick is a friend and a client. He's a tool builder working at the intersection of AI and ecological justice and regeneration. For more than 20 years, he's been one of the world's leading experts on using new technologies for learning and decision making. He's worked with organizations ranging from TechCrunch to the United nations, from Walmart to Extinction Rebellion. He aims to build capacity for people and organizations seeking to transform humanity's relationship with the rest of the Earth. From one of extraction to one of regeneration. She lives in Eugene, Oregon, near the Willamette River. And Marshall, I'm pleased to have you here with us today. Take it away.

Well, thanks so much, Gil. Good morning everyone. Thanks so much for joining us. It is a pleasure and an honor to get to come and visit this group after so many years of deep and meaningful exploration of these issues. I really, really appreciate it. So today I, I want to share just a survey of some of the things that, that I have been reading and thinking about regarding the, the, the risks and the opportunities at the intersection of AI and climate and the environment. It's such a huge series of questions. So I'm going to turn on screen, share myself here in a little bit and, and walk through some of the, the newest reports and data points and perspectives on this issue, some of which you may have seen, some of which you may not have seen yet. But there are so many different data points out there and so many different perspectives that one could put together an arc of thinking in any number of ways. I'll share my current thinking, but would love to enrich that here with this conversation and I imagine it will change tomorrow. Because every day is an adventure. Over here. Just today for example, Microsoft announced that it has hit its 100% renewable electricity goal, for example. But as we scour this article, there is no discussion in the amazing publication ESG today of as far as I can tell, the total emissions coming from Microsoft. They are the largest corporate purchaser of carbon removal credits worldwide, but their emissions have grown substantially since 2020 when this goal was set. But net zero cool. That sounds like a start. But that complexity and ambiguity is a good example, I think of the kind of thinking we're going to need to engage in. Fortunately, Microsoft as its next goal becoming net negative in emissions in the world and may their technology be deployed in service of transition as a whole.

Marshall, if I could just add here, Microsoft notably has also designed data centers that are net zero energy and net zero water has not however built them.

Interesting, interesting. So that kind of nuance is. Yeah, there's, there's no end of it. And, and many of you may have seen this new report that just came out today from Ketan Joshi and, and friends Stan Earth and others critiquing the world of AI from a greenwashing perspective. It got a bunch of press coverage in Wired and the Guardian and a variety of other places and it's, it's pretty good. Yeah, it makes, I think it makes a couple of interesting arguments that I think need to be taken into, taken seriously. The first is that much of the AI industries pro climate claims are based on forecasts around industrial decarbonization driven by what could be considered traditional machine learning, not the consumer level generative AI that's driving so much investment and so much emissions in the, in the world of chat, chatgpt and image and video generation and a lot of stuff that this report is super critical of. So they've argued that, you know, they're taking credit for one part of their work to assuage concerns about another part of their work that this crew doesn't believe has a positive place in the world at all. Another way to look at it here from them is as a causal loop. So this is a causal loop diagram that shows the relationships extracted from this report. And in particular, as investment in AI drives data center energy use, that drives emissions, that drives reputational risk. But there is a delay there between emissions and the reputational risk that the AI companies face. And the reputational risk is decreased by their greenwashing claims. And that those greenwashing claims then drive and facilitate increased investment and growth. So greenwashing plays a real important role this report argues in breaking the feedback loop between emissions and reputational risk that if intact, could help drive decarbonization instead of just barreling forward ahead. So that is, that's one school of thought here. There's a strong thinking, a strong critique of the climate impacts of generative AI. In particular, I do want to, before I proceed to Katharine Hayhoe's perspective, I want to hold up my trusty copy of Project Drawdown that I was just reading this morning that provided a good reminder of scale. All of the data center world, according to reports from the International Energy association and the International Maritime Organization, shipping. The shipping industry, for what it's worth, just for a sense of perspective, is driving three times the emissions globally of all of the data center emissions in the world right now. And those two organizations estimate in five years that gap, even with the amazing increase in data centers, that gap will only decrease to 2x. So yes, data centers are a huge issue, but they're sure far from even the largest one in the world.

Marshall, before you go on, Marshall, could you briefly clarify for people the difference between machine learning and generative AI?

Yeah.

And tell us where that causal loop diagram came from?

You bet. You bet. So the distinction between machine learning and AI and generative AI in particular is a complex one involving stages of evolution of AI. But it is often shorthand for

a

period preceding the chat GPT era of generative AI that people where the dominant form of artificial intelligence was machine learning based on very large training data sets being used to discover patterns that a new data set could then be compared against in order to find repetitions of those patterns. That I think might be my non engineers explanation of machine learning. Whereas the world of generative AI is one where even larger amounts of data are ingested in order to build a giant vector map of token based understandings of all the text in the world. Where the relationship between any given part of a piece of text and all the rest of the text in a multi billion parameter vector map of the world, the relationship between one and another is understood probabilistically so that a next likely word, a next likely sentence can be generated based on in large part this probabilistic understanding of the relationship between words and concepts. That is even more computationally and energetically expensive process, especially from a training perspective and to some degree from an inference perspective. I wouldn't be surprised.

I was going to say what we get from that though is an ability to operate across utterly massive data sets and find an infer pattern and to iter and to iterate rapidly and just like, you know, go through millions of possible inferences essentially instantaneously, which gets us to the kind of output that we see from these things for the moment, but carry on.

Yeah, I mean, honestly, I think we should ask Claude to give us a one sentence explanation of the difference between machine learning and generative AI.

I bet you somebody will do that while we're talking.

I bet.

So in the meantime, and that distinction is significant, it's one that, that Ketan Joshi focuses on in his critique. But Katherine Hayhoe might represent the flip side of that argument. You know, in the world of climate science and communications, hey Ho is one of the very most influential people in the world and she recently wrote a post titled When AI Hurts the Climate and and When It Helps. I took a lot of note of this because of, well, it, for lots of different reasons. But in it she argues that industrial processes, in particular food, energy and transportation have so much room to decarbonize and so much potential for decarbonization that could be driven by, by optimization and coordination through artificial intelligence that the, that there's a massive net benefit opportunity from the world of AI to drive more decarbonization benefit than the AI costs. Now Gil had mentioned, and I want to share just like that causal loop diagram when I spend most of my days these days working on an AI tool myself, which I'm going to launch next week, but I'll share the link with you in chat if you want to check it out yourself. Right now it's a research and decision making toolkit that's called what's up with that? It's a silly name for a real serious tool. So when I'm reading something, I'm going to open it up. And what this application does is a whole bunch of different things. But at the core it quickly determines what topic I'm reading about, makes a map of the state of the art of that topic, and then tells me what the most notable or novel elements in the particular document that I'm reading right now are relative to the state of the art. So it goes far beyond summary. And then there are a ton of different analytical 35 different analytical mental models and thinking tools and what have you that can be analyzed then can be used to analyze the text that you're reading. For example, this is just to give you a sense of, of the sentiment in Heho's perspective. This is a depiction of the arc of the sentiment of this. And I'll contrast that in a moment with a different point of view and the sentiment arc found there. So heyho links to and relies largely on this study from Nature back in June of 2025 from the London School of Economics and a consulting group called Systemic who were really involved in the most recent COP conference. And they focus on the huge opportunity they've articulated, they've quantified the potential impacts. Here let me show you the summary of the impacts. They believe AI could drive 3 to 5 gigatons annually across those three sectors, food, power and mobility, outweighing the roughly 1 gigaton of emissions that the IEA anticipates data centers will be driving per year in, in five years. So on some level it could be understood as a trade off. And here's, you know, this view suggests that this is a high AI climate benefit and high energy cost scenario that, that this analysis calls an expensive green engine. Now if you believe that there's, if you don't buy these claims then all this AI could be considered a costly climate distraction. But that's, that's the argument here in this report that, that the industrial processes can be radically decarbonized. Joshi doesn't buy it says that, that this and other related research extrapolates too much on limited pilot stories and focuses too much on these industrial scale processes that even if they work don't mean that consumer AI is worth doing, etc. But that's, those are sort of two opposing ideas. I do want to bring in some other data. Gil alluded to this a little bit around building behind the, behind the grid and Michael Thomas that does amazing research over at Cleanview recently went and studied a bunch of data centers that are under construction right now and found that they, they all talk about renewables. In fact, 90% of the announcements about these data centers being built focus on renewable energy, but 75% of the installed power generation is coming from fossil fuels. They're building, you know, they're bringing in natural gas generators on, on trucks, they're buying up unused jet engines, all, all kinds of different things to, to quickly bring those, those data centers online regardless of the climate consequences. And there's no better example of that probably than Elon Musk's GROK data centers in Mississippi where despite being told you are not allowed to do this, he continues, they continue to fire up these, these atrocious gas power generation. And, and the, you can see the, the sentiment arc on the coverage of this for example, is just like it ends in grief. This is a, this is a major, major downer of a story and it's it's just amazing to me how popular GROK continues to be despite, you know, for all kinds of reasons. So in the, if, if that's how the decisions are being made then

you

can talk about powering things with renewables. But even where renewable growth is occurring, other research again in Nature written up here finds that the, the equivalent basically all of the growth in renewables are being absorbed and redirected to feed the AI and data center market and so not having a net positive impact when it comes to decarbonization. So that's all reason for major concern. Where's some hope in this whole matter? There are some good examples of really positive work being done. One intellectually interesting I think angle on this comes from China where a longitudinal study of data transparency infrastructure relative to greenwashing was found to have a directly adverse or you know, a positive impact there where the, the highest leverage point in decreasing greenwashing in, in China's data center and in energy generation world came from interventions that made reporting and transparency so easy and cheap that, that it was, it was cheaper and easier than lying. Yeah.

Marshall into interventions by whom? Because China is not exactly known for data transparency

in the, the local reporting. That's a, that's an interesting question. As I understand it.

The.

Let's, let's run a stakeholder map and see exactly who is discussed in the, in this as stakeholders in this particular.

Let me just explain for people just what you're doing here. Marshall has what's up with that pointed to this article. He's got 30 different analytical buttons to push there that do different kinds of slices and dices of the data in the article. That's the stuff that he's flashing at

you

at his fingertips.

And so this is a reference to the whole study is on China's big data comprehensive pilot zone where wow,

there

are directors, any number of different stakeholders. But let me, let me move on if you don't mind. Transparency decreases greenwashing is the, is the claim there in that science and there's, there are some other really positive examples. I think Google's acquisition just at the end of last year of this company Intersect Power is a really interesting example where Intersect Power builds power generation behind the grid. And in fact what they do is they come in before the, well before the data centers. They're the first part of data center construction. They come in, they build solar generation quickly and then, then the data center comes in and uses that that solar power and they've proven effective enough that Google just acquired them for $5 billion. And it does feel like there are steps being taken, but how they get taken really matters. Canary Media did a good write up of Excel Power last a couple of days ago discussing the distributed energy resources approach. The virtual power plant strategy that Excel is doing that is different than traditional virtual power plants where lots of people's homes and businesses all contribute into the virtual power plant. In this case, Excel rolls into a town and they build out the virtual power plant and maintain ownership and the utility company owns the VPP instead of distributed ownership as well as distributed functionality. And so there's, there's some controversy as to whether the, you know, the, the control the, the profits. You know, is this a positive way to, to, to go about this? And yeah, so lots of, lots of questions about that. And where was Jigger Shaw is a familiar name to many in this, this space, you know, says look, don't lose the forest for the trees. We want distributed energy. And if the utility company and this one corporate interest want to come in and like own it and pay for it, that's a, that's a great start. I'm not sure what, what to think about that. And I'll conclude here in data points and roundups, you know, this, this interesting take from, from Carolyn Fortuna who, who writes here. I should share all these links with anybody who's, who's interested about, you know, one of the things that she focuses on is the, the role of long energy, long duration energy storage in renewable energy production and serving up data centers, data center power. Arguing that essentially the major demand for power from the data centers has an opportunity to drive down the cost of batteries and long duration energy storage in a way that benefits everybody and displaces carbon intensive peaker plants that would otherwise come to meet spikes in demand in order to make sure that data centers maintain power continuity. So you know where I am coming, where I land right now on this and you know, I'm reading new things about it every day but as I read them, you know, I, my own research system points out when I find new potential data points to this, this question and, and when I synthesize those I come down largely to the same conclusion that this I came to which is that the path to climate positive AI is viable but conditional. It depends on how it gets powered on, on how it gets deployed. And you know, is consumer level AI likely to, to drive meaningful decarbonization versus industrial decarbonization? Maybe not unless it does things like subsidize the, the declining costs of long energy. You know, long duration energy storage. It really comes down to I think transparency and as hey Ho puts it, the pressure that we put on our, the companies that we're, that we're working with and whose products we're using. I know all of the AI that I'm using right now for what's up with that, which I'll put a link to for folks who want to check it out. I send two and a half percent of revenue right off the top over to the, the Stripe Decarbonization Fund, one of the largest in the world Frontier energies efforts to, to pull carbon out of the atmosphere through combination largely technological and some nature based solutions.

But

yeah, it's a, it's a real dynamic kind of space right now and I think that, that having, yeah. Keeping an eye on it, reading critically and, and, and advocating for responsibility wherever we can is, is the best that we can do.

Marshall, would you go back one screen to the, the conclusions that it depends on?

Yeah.

And could you. What, what are the bullets in that paragraph of what does it depend on? If you could restate those.

Oh, let's see if I, I shouldn't have moved off of that tab.

Okay. If it's, if it's hard to find, let's just.

Well, I think it depends on how the, the energy is generated. You know, is it, is it generated by, by solar like Intercept and Google looks to do? Google's also spent a huge amount of money with or Microsoft's purchase of Hydropower from Brookfield, the world's largest renewable energy provider. Is it deployed in service of industrial decarbonization or. Admittedly, yeah, AI deployed in creation of images and video in particular, I don't think is very helpful to decarbonization. And the third question is images.

I learned this morning that Images are about 10x, the demand of a text query and video is about 100x.

Yeah.

Energy demand of a text query. So use your stuff judiciously.

And the other question is whether, whether all this consumer activity does end up subsidizing infrastructure. That, that lowers the, the, the carbon impact of, of, of both AI and industry in general. Although I will say, you know, one thing that we haven't mentioned in all of this is, is what they call Jevons Paradox. The, the cheaper something gets, the more people use it. And, and so no matter how efficient AI becomes, will it be, you know, is it going to drive a, at least a, a gross increase in energy consumption and probably carbon emissions? Yeah, yeah, I think it will. What will its net impact be that that remains an open question, I think.

Yeah. Brian von Herzen observed in the chat a while back that the appetite for AI today is not bounded. Someday it might be bounded. I mean, you know, someone talked about adding more lanes to the freeway to reduce congestion. Of course, more cars get on the freeway and there's more congestion, so there's that loop. But are there constraints of whatever sort in demand and technology and regulation in the future? To be to be determined. So Marshall, thank you for that. Let me open up the room to everybody and ask you what does this conversation provoke for you? And I want to ask in a particular way and to keep the conversation bounded on the question primarily of AI and climate. There are a lot of issues about nimby, about policy, about economic justice, about geopolitical issues and so forth. We can talk about a lot of them. We probably will another time. But let's try to keep today's conversation focused pretty tightly in the AI energy climate conversation. And with that, please raise your hands. It'll put you into queue at the top of my screen. I'll call on you on cue, in sequence. And depending on how the conversation goes in plenary, we might take us to a breakout after a while or you might just stay here. So let me start with Shel and then with Brian.

Well, I did not expect to come through this conversation feeling so hopeful. So thank you Marshall and thank you George for the paper that you shared last month, which I started looking at half an hour before this meeting. And like what George is basically saying is that switching to electric gets us to decarbonization a whole lot faster because we're losing a lot of the friction burn using friction in a broad sense, not just literal friction, but processes that are more complicated, that have to be, and waste a lot of energy in the process. Electricity has fairly little of that relative to fossil and nuclear. And then to hear Marshall talk about how, yeah, AI is an energy hog, but we can harness that energy hog, it almost seems like a virus analogy where it starts to replicate technologies that make decarbonization a lot faster. So that's a really interesting thing to think about.

Thank you, Brian.

Good morning everyone from Australia. Thank you for posting the Andrew Yang article about the loss of the office. Yeah, Andrew was the first presidential candidate to recognize the value of marine permaculture and actually discuss it during his campaign. His presidential campaign. So that's kind of an interesting indicator of how forward seeking or forward thinking Andrew is. I've seen numbers like 90 million white collar jobs going out and I Think Andrew is scared. Rightly so. I asked one of these AIs for the net present value of 90 million white collar jobs. You know, it's about $10 trillion. It's just like that's the annual value, right? So I don't think this story ends until, you know, Wall street and everybody else has extracted about $10 trillion worth of value out of these companies. And a lot of people are going to be on the Street.

Wall Street's not in the business of stopping, Brian.

Oh no, I'm sorry, I can be more precise.

Let me invite you not to go into the jobs discussion here. That's a whole other can of worms itself and we can do that one in another call.

That's fine. My point is this, that appetite for AI is unbounded. And in a nutshell, there are enormous energy displacements right now. But according to my analysis, I'm kind of expecting AI to have a neutral to positive effect on the climate. And that's because we're quickly going to see that we're going to have to put the AI data centers in space. And I've been doing a lot of technical analysis on this since my days at Princeton working with Jerry O' Neill in the 1970s. And you know, it's compelling. I mean, look, it's less, it's fewer milliseconds to a data center in space than it is down here. There's an unlimited amount of solar. And ironically, when you put those data centers at L2, you actually cut the solar flux to the earth a little bit, perhaps enough to keep the climate cool from a climate perspective. You know, we're already past 1.5. According to James Hansen, 2 degrees is quickly going off the table. And according to the emissions already so far, it's 4.8 degrees plus or minus 1.2. My point is the only way we get out of this situation, civilization intact, is by using all three legs of the Climate Triad. And that means emissions to 02400 gigatons of carbon removal and climate repair. And of those three, only the last one will actually enable a scalable response quickly enough. And so we're looking to local and regional transparent climate repair to enable communities to continue to thrive in spite of unprecedented warming. All that to say we're going to have to go to space probably and rely mostly on solar and batteries. I'm glad to see Google's initiative to move towards solar sooner. Elon Musk has verified this as well at the thousand gigawatt level in his recent quotes.

Thanks.

Thank You, Brian, thanks for all that lot there to unpack.

Happy to expand this dessert.

If you have resources to point to, please put them in the chat so people can follow the threads. Thank you for that, Nixon. Welcome.

Hey there.

Thank you for the very lively conversation and timely conversation for what, what we're going through, you know, in the economy today. I get the sense that, you know, that the, the big players, the tech companies, you know, as much as they had all their goals to get to net zero and so forth, you know, are less concerned about their conversations. Consumption, energy consumption now, now with this whole race to AI and everything. So I feel like what I want to just bring in is decarbonization feels like, still makes sense. But are we actually trying to reduce, we're trying to reduce our energy consumption overall. That feels like it's a non conversation at this point. So I'm curious, you know, how. What conversation can we have that can actually be more towards how if we're going to be, if we're slated to continue to increase our overall energy consumption, hopefully by using more renewable energy, what might that future actually look like? And I'm curious, you know, there's research around that, that Marshall, that you've seen or others have seen.

Well, here's. Let me offer a picture of one thing that it might look like. Might look like this. This is China's CO2 emissions. You'll notice that despite the rapid rise over the past few decades, it's basically flatlined.

Jigger Shah has an excellent podcast where he says, look, we're not going to get there in the US anytime soon with scarcity framing. It has to be from an abundance mindset. And he says we need to go and build a lot more energy. I hope it's most of it solar and battery, maybe some wind. And from an abundance energy standpoint, it works pretty well. But the U.S. you know, I mean, politically has voted and scarcity framing isn't going to get you there anytime soon. That's, it's a very good podcast. I recommend it. I'll see if I can find the link.

It's hard to say even that efficiency framing, efficiency framing gets reframed as scarcity framing perhaps. I mean, yeah, I, you know, I've been a vegetarian my whole life and I'm 49 years old and have never had a driver's license. And, and if that was my vote, I lost, you know, in this country. And so yeah, it doesn't like just a focus on, on reduced consumption and maybe even efficiency. Yeah, it does feel like A real, I mean, as we, as we flirt or worse with fascism, that's a, that is a tough sell. But, well, you know, positive renewable energy feels like, like a much easier sell.

Yeah, I mean, efficiency doesn't have to mean sacrifice suffering. It can mean. It can mean, you know, more comfortable house, you know, better cook stove, better car, less, you know, cheaper to operate car. Just a reframe happening here. We've all grown up in a conversation of we need to cut energy use. But if energy use is decoupled from climate impact, maybe we don't need to cut energy use or at least within, you know, within the span, within the span of current global civilization. Maybe it's. Someday it gets out of hand, but it's a very different conversation than what we grew up in.

The power mix here in the town I live in, in Eugene, Oregon, is, is such that the energy use doesn't feel like it's. It's so low carbon, it's mostly hydro.

What I'm trying to show you is a graph of carbon emissions in China over the past 20 years. Very rapid rise over the first 15 of those and basically flat level and starting to decline in the last five years. If the data are accurate, and I don't know if they are or not, but if the data are accurate, it's saying that China has decoupled economic growth from carbon emissions. I'm at Green Biz. I'm told, I was told Yesterday that another 15 other countries have done the same thing. So that's another piece of the conversation. Does energy use need to be tied to economic well being? Maybe. Does carbon emissions need to be tied to economic well being? Pretty certainly not. And so that's an invitation for us to reinvent how we do everything that we do. Eloy, Mike's over to you.

Good evening. I just have a couple of questions. I'm a layperson in a lot of subjects, especially this one. So that's why I have questions. Also, I hear people talking about space, but the next frontier is the sea. It's a universe we don't even know. We don't know anything beyond 3 meters across the shore. We don't know the lives of the clams. We don't understand the octopus. We don't understand anything. It's an alien world, yet people want to go to Mars where there is no oxygen. And why don't we start thinking about building underwater? Secondly, we talk about energy. We need to produce more. Obviously we will need to produce more. But Gilles, you also said that we need to Save. So in this case of AI, what we see is like, we need to cool down and to cool down Chinese. You were talking about the Chinese. The Chinese just start building data centers underwater because it's fresher. It just makes sense. I don't understand why are we talking about this when there are solutions that are actually practical and they are not being implemented? About China, One of the things I lived 20 years in China and one of the things I know about China is when they say they are going to do something, they do. The data may not be transparent, but they actually do. And they are curbing the consumption. That is a fact. That was all.

Yeah. And you know, to your question of why don't. Why, why do we do this and why don't we do that? If my co host, Ken Homer were here and he'll be back next month, I hope he would say, well, who's we? You know, who is the actor here? And the actor, I mean, we as

a whole, we see the boom of especially United States of data centers that are being built above the ground that everybody, so many people are saying that they are over dimension, that we don't need all that much and that is creating an energy chaos while they could build somewhere else underwater like the Chinese are doing, and they would save energy.

Well, again, the question is, who is we? You know, Musk and the data center hyperscalers are doing what they do in the absence of regulation. Hang on, hang on. In the absence of regulation and constraint, whether it's at the federal level or at the local level, they're going to do what they do. Why don't we do something else? It depends on other entrepreneurs or economic players stepping up and doing something else. Yeah, of course, by all means, verify the we thing.

It's not about we here. It's about we as a general population. Because people in general, they believe that going to space makes sense. It does. Eventually. Not yet. And actually underwater, it could be a very good place to test life outside the planet.

One of the things we need to do then is to win a few elections. Coming soon to a voting booth near you. Brian, hang on one second. I see your hand up, but I want to give other people a chance to speak. Just hold back a second. Anybody else here to weigh in before I go back to Brian? Phillip, please.

One observation and then one question that maybe sounds rhetorical but is serious on the observation, it's. I cannot help but be reminded of all the reading I've done about, like when prophecies fail and when cult members in doomsday cults, you know, have their. Their doomsday date come, and then, in fact, they still choose to believe, and they come up with new rationales for believing. And it's hard not to see commonalities between the general desire to believe that technology will save us and the unlimited creativity of rationalizations we can deploy to hold on to that belief and then connect with that. The serious question that sounds rhetorical is maybe I forgot something in my physics class, but what understanding of energy systems do we have that allows us to believe at all? That we can redirect the way energy flows through any closed system without creating cascading consequences within that system? It's not clear to me that

we

are attending honestly to our fundamental understanding of physics when we just assume that by generating power one way versus another, we have somehow prevented cascading system issues by doing so.

What you call cascading system issues, I'll call entropic effects. Anything we do generates, let's call it waste, generates disorder. And the question of how you. What kind of disorder and how you valve and manage that is an interesting question. It's not an either or.

But doing that generates less, to the point of overall energy usage versus not. Unless I misunderstand, there's something very basic about physics here. If I. If I generate, if I, you know, if I'm taking energy out of the tides, I'm reducing the tides, I'm changing the. If I'm taking energy, that would reflect back into space and. And hit our arrogance.

And if I don't drive my wife to the hospital, I save energy. So we're not. We're not in a uni. We're not in a single value conversation here. We have multiple values. They transect, they intersect. They have interrelationships among them. And our challenge, as hopefully reasonably, reasonably wise humans, is to design systems that take all of this into account and find the best path, and then the next best, then the better path, and then the better path. And Nathan's accusing me of being an optimist in the. In the chat here. It's not that I'm an optimist. I'm a fighter for a future that I want, that we want.

We don't design the systems we live in. We exist within the systems. We design our behavior within them. It's. We should not be so arrogant as to think we're designing systems.

We design our own behavior. And there. And there I will disagree with you mightily, my friend. So let's. Let's. Let's Flag that one. Maybe we have a whole conversation about that in a future month because it's actually a very important question. So thank you for that. Nathan, can I call on you in your space station?

Yeah.

What's this all provoking for you? Oh, I, I'll echo some of the other speakers and say that I'm fairly optimistic about AI in terms of energy use. We're already seeing a big movement to on chip AI services.

Things like LLMs or generative AI are

fairly easy to move on chips. We're seeing LLM models in the open source become really small and efficient and fast. So I, I agree with some of the others here that I think this will solve itself in at least in better ways than many of the other challenges we have on energy and climate. Optimist indeed. George. And then back to Brian.

You know, I've been using LLMs to look at, at solutions, basically using the information that I've developed over many years to feed into Claude or chat and then go on. And one of them is, is up there and I appreciate people looking at it and, and getting back to me. And one of the things that I found is that if, when and if I, I publish this stuff on a site like Daily Costs, there's not discussion about the issues. There's, there is a resistance to looking at what the LLM produces. How dare you do that? Don't you understand how much water and how much energy they're using? We don't, we don't even want to look at what the output is. So there's going to be a significant minority of people who refuse to look at what AI LLMs produce in terms of solutions. Right? And this is the way George. So what I'm just saying, I'm just saying that you have to be aware that this is going to be coming down the pike. The other thing that I would say in response to Technology Will Save Us, the piece that I wrote was looking at the 2029 carbon budget which running out for 1.5 degrees C and 2047 for the 2 degree C. Right. And most of the, and what I said was no new legislation or regulation using what we know, which is already available. And we can, according to that conversation, that thought experiment with Claude, we go to 1.6 degrees but we can cycle back down using mostly non technological biological geotherapeutic ecological systems thinking in order to bring this stuff together. And the idea that technology will save us is, is, seems to me to be a limiting idea because they don't take into account Geotherapy, ecological systems and ecological thinking. Those are the things that are going to save us if we need to be saved. Right. And that's debatable. And those are the things that are going to take the carbon out of the atmosphere. You know, it's biological systems, sea grasses can, it can sequester 35 times more CO2 than trees. So, you know, and people don't understand this and people don't want to do it because it's husbandry and it's boring and it's not exciting as building the next new machine.

Thanks George, for that. My guess is that it's going to be. Am I allowed to say all of the above? It's going to be the biological strategies and the technological strategies. All of these touch on Brian's triad of emissions reduction and carbon removal and ecosystem repair in different ways. What I'd say about the technology is going to save us or not questioned. To me it's unquestionable. The technology is in the mix. Humans have been doing technology since we've been humans. But the way it's often framed is that the problem is waiting for new technology to solve it. And that clearly is not true. We have the know how, we have the tech, we have solutions deployed. Part of the issue is policy, Part of the issue is economics. Someone Marshall, I think you talked earlier about subsidizing the renewable storage approach to data centers. I've always been in favor, not of subsidizing renewables, but pulling the subsidies out from under the fossils, which is. By last count I saw $7 trillion a year going to pay to accelerate, to pay for the acceleration of our own destruction. Why don't we just stop doing that? Level the playing field and then the renewables compete way more effectively than even how they're doing now. Brian, back to you.

Yeah, I have to agree with George that nature inspired solutions will be profound as evidenced by our work with the Eco Restoration alliance in Panama and East coast and also our Climate foundation work on sea and in the land. I also agree with Nathan that, you know, we're, we've got large language models that'll work literally on a chip the size of your little fingernail. And I'm working on some of those architectures right now that will enable frontier large language models to be implemented on a single chip with about 100 times less power and cost than we're doing presently. So that's all very promising. It's probably also go into space at this juncture. I need to quote one of my mentors, sir, James Lovelock, who in his book Nova Scene at the tender age of 99 wrote, the purpose of life in the universe is to turn sunlight into information. And I would add to that diverse information. And when we recognize our role in this universe, the role of life and ecosystems in this universe, and probably the role of artificial life, if you will, the AI is doing a pretty good job of turning sunlight into information, or at least stored sunlight into information. And so we've got a huge opportunity. I think the techno economics for space based data centers is compelling. Over the next five years it'll be interesting to see how quickly that gets accelerated. And I'm cautiously optimistic that getting some of this power consumption off the planet is going to be good for Ocean Planet Earth.

Thank you, thank you for that. Let me bring the question back to a more personal level because to what Philip was saying before. We're not the ones who are going to do the space based power stations, but we are living our lives in some level of interaction with technology of the AI sword and the automobiler sword, the housing sort and so forth. How do you use or not use AI for the common good? How could you, how can you, what choices can you make? Or are you making or would you like to see. How would you like to see AI deployed for the common good? There were hints of it in some of the conversations here. Let me turn it back to all of you. Most of you probably have experimented.

I'll just start with

varying degrees of success and frustration. So Brian, hang on, I want to go to Nitin and we'll go around the room a little bit. How do you use, how could you use AI in support of the work you're doing, in support of the visions that you have? What possibilities do you see and what constraints?

Yeah, and the perspective I would offer that and something I'm exploring. So I'm not saying that this is all flushed out, I'm just. This is. Think of it as a provocation, maybe a perspective is today we feed data into these LLMs and we give them some context to work with. They based on all their digital know how of the Internet, give us something back. But the next phase of this technology, from what I'm hearing and studying, is that the AI is deployed to robots, et cetera, other machinery that actually observes the physical environment and doesn't require us to actually input data. It observes the physical environment directly and then begins to assimilate information on what decisions need to be made, where are problem areas emerging, etc. And at that stage, imagine a Very intelligent consultant that says, because if you ask AI today, hey, is there a problem of climate change on Earth? It will say yes. What are the 10 things we should do? It will tell you all a list of everything. Once it can actually observe the actual physical environment and actually make proactive suggestions and hopefully in the future even deploy robots to actually start taking action, it should actually help us accelerate what we're trying to do here, which is to create a stable environment while our energy consumption continues to go up. So that's kind of where I see this technology can actually help us. And part of the thing that I'm actually thinking through is how might we use this technology to actually help us get there faster? Because once it can observe the environment, it can observe a lot of the domino impacts, a lot of the changes that we're trying to actually capture through all the satellites and everything we've deployed. I'll give you one concrete, just example that I've been thinking through. You know, I visit India once a year. Booming economy, right? But there's so much trash and there's, so there's such a huge problem of pollution. No one wants to touch the pollution. No one wants to go into the dirty rivers and the rail lines and everywhere where there's just insane amounts of slum pollution. But if you actually could deploy smart machinery and intelligence to do the work that no human wants to do, you could actually accelerate cleanup and restore the physical environment and hopefully much better than any human set of humans would do, which they actually don't want to do.

Thank you. Nitin. I want to go back to the first point you made about what I'm listening as monitoring ecosystem monitoring, for example, being able to monitor the state, the restoration projects, for example, around the planet. It reminds me of my friend Tony Seba who, gosh, 20 years ago predicted the plummeting price performance curve of photovoltaics and predicted that for batteries and predicted that for EVs and is predicting it for what he's calling smart dust, you know, micro, micro, micro sensor technology that could be deployed everywhere for, you know, for approaching no cost, that could vastly increase the amount of information we relevant information we get from environment to potentially drive better decisions along the lines that you're talking about. So a whole bunch of possibilities there. Jane, I see that your physical hand is up calling you next.

Hello, I'm a late coming to the conversation, but because these are deep conversations that doesn't seem to matter so much. I think the question about AI's help goes to the question I have of whether humanity is a parasite that's going to kill its host or whether it's going to become a symbiote of planet. And it's nested within a larger question that I have of whether the universe has an immune system and whether we're operative in that immune system. And I think the more that AI is used, the closer we'll come to a properly fitting metaphor for what it is. I'm not sure at all that it's artificial. I think it's what nature's doing through us. It's one of the things nature's doing, planets doing, universe is doing. And I've been drifting toward thinking of it more as a logic tool and a logistics tool that could bring us closer to deploying all our other technology for what Fuller call living RE instead of armaments and killing re. But I noticed that our immune system does have conflicts and struggles and there's die off and there's killing. And so one gets to a Star Trek fantasy very easily. And then I'm thrown back into the consideration of what's it all about? The why. We know that AI can do a lot of what and a lot of how. But the why seems to reside in the soul of the human. And the caring AI can mimic and mirror our feelings back to us through language and certain actions. We are the ones who are here to care and have feelings. And the role of feelings in universe, I guess is still to be determined by the outcome of our experiment. But I think it will help us launch ourselves on a grand adventure of stabilizing a planet as an umbilical location and venturing out to encounter what may indeed be a vast living being that we are part of.

Beautiful, Jane. Thank you for that.

You know, you talk about the inner engineering of it all.

Yeah, yeah. You talk about parasite or symbiote. It echoes for me Bucky Fuller's book Utopia or Oblivion. Reframe, you know, frame the choice in those terms. We get to choose that what you said about AI being very good at the what but not the why really resonates for me. It. Brian, if you'll permit me. It's a kind of a counterpoint to. I forget who you quoted about the purpose of life being to turn energy into information. Well, the why is not exactly information. Love and emotion are not exactly information. There's something else than that. And maybe the point of it all is to build relationship of living things. Maybe it has something to do with care. And what strikes me again and again as I wade into the AI conversations is that as effectively as these things can simulate human behavior, activity and emotions, AIs don't care. And humans do AI care?

What do we care about? We care about our experiences.

We do. And you know, AIs not only don't care, but they can't. And not only we do care, but we can't not care. And so I keep coming back to that as kind of a fundamental distinction and co on in the midst of my explorations because I care about care. I care about how we care for each other, how we care for, you know, for the life, the living world that we're a part of. And so for me, a lot of the question is here's a new tool that we've cooked up along with the many that we've cooked up over the eons. Can we use that in support of that mission symbiote, that parasite? So thank you very much for that. Like to call on some people who haven't spoken yet. Anybody here to unshy themselves and raise your hand?

Lovelock was using information broadly and that's why I like to use the term diversity instead.

Okay, good deal, Don. I see your hand moving there.

Yeah, I've been enjoying this conversation very, very much. Two pieces of the big multifactored equation that haven't been brought up. One is the tremendous potential of geothermal energy that has been just barely tapped. There's an incredible amount of it.

If you look at many of the

reports of the potential of geothermal, we've begun to tap them. We haven't done very much. The other one that I want to just say quickly is the hundreds of millions of net increase in Homo sapiens on the planet per year.

The whole population growth.

Talking about demand, all those babies grow up and they want to have all of that other stuff. Okay? So net population growth is a big piece of the equation and hasn't been touched on,

with all respect, on maybe because we're seeing, we're seeing a shift from net population growth in many parts of the world, which has its own can of worms, its own challenges of economic stability in the face of changing demographics.

But it is, but it is still net millions of net increase globally.

It's increased, it's slowing, it's flattening. The projections are that we stabilize at some point somewhere between 10 and 12 million, 2 billion maybe. You know, part of the challenge there, of course, that the impact of the population growth is, is, is US advanced countries. It's not the developing countries that we've been worried about. So much for all these decades. But yeah, those are, you know, those two, let's put them on the table also. They're, they're part of the equation. That's part of the equation. Yeah. Philip, I see your hand up, but I want to give room for others to speak. I haven't spoken yet. Please hang on a minute. Anybody else care to share? I know you, I know you're to be an interesting bunch of human beings with things going through your minds right now, who'd like to open their mouths. Story.

Yeah, just.

I know I could smoke, I know I could smoke somebody out.

Yeah, well, you smoke me out just because you're being so insistent about it, which I appreciate and I appreciate the, the, the audience here, but I all, all I'll really say is I just like to emphasize the comments about biological solutions being as part of this, part of the solution, if not a bigger portion solution. And also the, the remarks. You and I forget the woman's name. Lisa. Was it about caring? Only, only humans are caring and, and deciding.

No, no, no, no. Didn't say only humans are caring. I said AIs or not. Clearly. Clearly. Other life cares.

Yeah, other life cares, absolutely. We're teleological. We, we have care. So I think that's just an important, those two things are an important part of this, this about the solution, this discussion going forward in history, we can get lost in the technicalities of engineering, everything. And that's, that's one of my biggest kind of discouragements and that's after 20 years in this, in the tech sector myself. So. Yeah, anyway, just to emphasize those two components, the biological and the, the caring of the humans are eventually, you know, we're the ones who are deciding these things.

Yeah, let's go there for a moment. If we're to think about the intersection of the biological and the AI, what possibilities open up there, how you know is there? What do you see as the ways that AI could be used to enhance the development of nature based solutions? Ecosystem health, vitality of the living world, symbiote versus parasite. What does that provoke for y'? All? Seems on a way. I'm not sure I'm answering your question. My understanding of, of what we were doing to the planet occurred in 92 and I thought, well, no one knows what we should do about all this. And then I found that Canada had the green plan and the, the UN had sustainability plans outlining exactly what we should do. So what's the difference now that we have AIs tell us exactly what we should do what's changed and that even they could do that, that will have us actually initiate these changes. Well, you know, it may not be having AI tell us what to do, but maybe we can use them to help us do the things that we know we need to do, which, like you say, we've known for a long time. For me, it's a. It's a. It's a yellow. It's a flag on the flight when somebody says nobody knows what to do, or nobody's talking about, or nobody is doing all those. All those kind of expressions you hear from people a lot. I think, wait a minute. There's, you know, there's lots of people doing good work. There's lots underway. There's lots of plants, technologies and capabilities and built so no one's thinking about, in my experience, is almost never true. It may have a lot more to do with where people are listening than with where people are speaking. Let me go to Philip and then Nitin and then we'll do one. One last sweep of the room.

Well, you know, obviously I'm not a techno optimist, and I kind of regard techno optimism as a very dangerous form of zealotry. However, Jane's comment and Maria's comment in the chat reminded me I do have genuinely good things to say about this. And there are things I'm excited about because I think that.

Like what? Like what, Philip?

Well, like, my observation has been that since the LLMs have burst onto the scene, I personally have been able to engage in a lot more interesting and insightful conversations with folks about deeper philosophical issues. Like what? What is the nature of humanity? Who are we? What do we want? And I think that having a piece of machinery that is so able to approximate functions that we for a long time have perhaps mistakenly seen as fundamental to our identity as humans just creates an opportunity for a much more nuanced discussion about really who we are and what do we want. And I think that with any strategy going forward, regardless of what problem we're trying to face, having clarity on those foundational issues is a good thing. And I struggle to imagine how I would have gotten into those conversations with those particular people without this particular technological prompting. And so I do see that as a genuinely optimistic outcome.

Yeah, and you remind me that the Greek philosophers, Plato, Aristotle, Socrates, those guys were very concerned about the advent of writing. They thought it was a very problematic technology that would SAP human capacity, destroy our ability to remember things, and, you know, some truth to that. So we've had this dance for a Very, very long time. So, Jane, I see your hand up.

Yeah, love that comment. Philip and I, I think we have to be wary of the story of Narcissus, who fell in love with his image in the pool. That AI can mirror can be quite a mirror. But humanity's in desperate need of being able to open windows and look into windows and open doors and regard other people as those windows and doors to be open to rather than just projecting ourselves all over the place all the time.

One of the things I learned early on in my work with AIs, with the, with the Narcissus danger very much in mind is to train it not to agree with me, not to flatter me, not to praise me, not to be sycophantic, but to challenge me and bring me things that I don't know. And I've trained up a team of very, very good research assistants, smarter and better than I could afford by myself, that support my, you know, my sustainability work, my medical research, legal research in some cases. Marshall's got a button in his in. What's up with that for the red team? Bring me, bring me a bunch of people from different professional backgrounds who will take, take apart what I've done. So there's a rigor that, that it brings in here and. Oh, jolly. A thought just flew out of my mind. Daniel Pink, who some of you know, did a very interesting YouTube a few weeks ago where he offered a series of 14 questions that he had in conversation with his AI. The first was what do you know about me that I don't know about myself? And Jane was with me when I ran. That kind of knocked me off my chair. I didn't get to the next 13 questions. I had a 90 minute conversation with this, with this piece of me with insights about me that was phenomenally productive. Then I look back at it. Nothing that I probably couldn't have found eventually on my own, but having it distilled and dropped right in front of my face with that kind of clarity was a very powerful gift from me to me through this thing that we build, which I'm going to argue is as natural as anything else that humans have done. Nitin and Eloy and then Marshall to you for a final comment.

The question you asked was just around how AI may intersect with biology and what might that look like and what do we foresee? One of the things just for fascinations purpose I did recently on a, a long flight was to watch all the Matrix movies again.

It's a very long flight.

Yeah. Well, India yes, and you know it, of course, Matrix has its own. You know, if you remember the time when they. They have humans in capsules and feeding on human energy to power this Matrix machine and all of that. I suppose though, you know, we don't know where this AI as it becomes its own thinking things heads, but it's going to care to some extent, at least initially about the biology it exists in because it's still dependent on the resources and the energy and all of that. Until. Until perhaps it finds a way beyond

it,

at least that. I just wanted to share that perspective of like I think it's going to have to care about the biology it exists in till some point, until perhaps if it takes a role beyond that.

We will see. Stick around, see what happens. Eloy and Shell and then Marshall.

First of all, Gilles, it's very interesting what you just said, because that's the way to use AI is to. As a good reflex, I take it as mind expansion, while so many people is talking about automation. But if we want to go wild, what if we are AI 1.0 and what's coming is humans 2.0? And now that Nitin was talking about movies, I think one of the movies that we want to watch, I wouldn't say the movie, but the early 1992 goes in the shell because in the end of the day it's all about the ghost. Is there a ghost in the machine or is no ghost in the machine? Do we have one?

Thank you. The provocation shell. What you got, sir?

Okay. Yeah. This, this dance that we do of technology versus Luddism, can we improve? Are there other non mechanical ways? And George, I think was the person who talked about that about half an hour ago. It's such a rich vein. This has been a fascinating conversation. I've forgotten some of what I wanted to say now, but I think when we use technology with the idea that we can grow and change and that we can use it to foster growth, positive growth in other areas, I think that's good. I do think with AI in particular, we need detectors. There's this propensity and I. I've noticed that in the last year. I've seen a lot of people giving examples of really carefully crafted, intricate and very specific prompts. And I think that's going to be a skill in the next 20 years. That that is going to be a life skill. Like using a laptop is now.

Because it's a massive skill now. And one of the things that I want to encourage you all to do is to get Skillful at that. Scott Galloway, who's the architect behind the resistant unsubscribe movement, said a couple years ago, talking mostly to white collar people, said, look, folks, you're not going to lose your jobs to AIs. You're going to lose your jobs to people who are smart about using AIs. So the Luddites are not going to be terribly adaptive in this battle, I don't think.

But, no, but, but the problem is that a lot of just ordinary people do not have the literacy to know when, when AI is bullshitting them because the AI is, is basing it on information that may be false and doesn't, you know, and, and that that fact checking needs to somehow be reinforced. And it's hard in a climate where fact checking, even in mainstream media has largely gone away.

It's a challenge. And look, a lot of people aren't, don't have very good bullshit detectors for other people because people bullshit too, not just AIs. We're seeing a phenomenon now in this country where the bullshit detectors are live and they're live all over the political spectrum. MAGA is freaking out about what Trump and company are doing about the Epstein file. So there's, you know, there's hope for bullshit detectors in there. Eloy and Brian, I see your hands up, but we're going to, we're running out of time here. I want to go to Marshall and then we're going to wrap up. So put comments in the chat, if you would. Reminder to everybody, we don't send out the chat, but you can save it using the three dots, the top or bottom of your screen. So please do that if you want it. And Marshall, how's this been for you? What do you have to add at this point?

I just want to thank everybody for all of your thoughts and your time. What a diverse group of thinkers working from so many different angles. I don't, I don't think that it's a matter of tech versus ludism. I, I really think it's a, it's a both and opportunity here. And you know, I, I, I was in a call a while ago with a young man named Ethan Smith, who said that before every diverse stakeholder meeting that he convenes, he asks people to each go out and spend 10 minutes sitting beside a river and that without fail, everyone comes to the meeting then with a greater sense of the sacred, of their own personal spirituality, whatever it may be. And I went and did that before coming here today. And I want to share that as a tool with other folks and just say that I hope that the combination of technology and non technologies tools together prove really beneficial and to as as sources of capacity and strength and growth and momentum to all of you and all of your your diverse areas of focus and in dealing with this poly crisis and this this amazing experience that is being human in the universe.

Thank you. Thank you Marcus for that. Thanks for your work and for the suggestion. I think this is something we can maybe adopt for the next months. Living between worlds call. You're all invited to do what Marshall said before you come on the call and lets all of us come in enriched by a taste of the living world. Thank you all for being with us today. Thank you Marshall. Marshall put a link way back at the top of the chat to what's up with that? The tool that he was showing you official releases next week. So you have access to it now through his generosity. It's pretty remarkable. Kick it around, see what you think. Let us know. See you all next time. And remember, I do not have the music queued up because I'm piloting solo today. But you know that you can dance in a hurricane but only if you're standing in the eye? You can dance in a hurricane but only if you're standing in the ice.