Работайте офлайн с приложением Player FM !
Andrew Winston: AI and the Climate
Manage episode 431491940 series 3381317
For Sci-Fi buffs, a future infused with AI may evoke unsettling images of HAL from Stanley Kubrick’s film, “2001: A Space Odyssey.” In truth, the evolving technology of artificial intelligence may well be taking over, but not quite how the filmmakers envisioned it. This week on Sea Change Radio, we speak with Andrew Winston, a sustainability strategist and bestselling author, about what AI means for the climate. We look at how AI can help various key sectors of the global economy become more efficient, examine the dilemma of AI’s seemingly insatiable energy needs, and discuss its potential to contribute to a carbon-free future.
Narrator | 00:02 – This is Sea Change Radio covering the shift to sustainability. I’m Alex Wise.
Andrew Winston | 00:17 – It is just, we don’t really have time to work it through and double emissions and then come back down like, we’re out of time on climate. So how do we make sure we’re adding this without creating another big problem?
Narrator | 00:32 – For Sci-Fi buffs, a future infused with AI may evoke unsettling images of HAL from Stanley Kubrick’s film, “2001: A Space Odyssey.” In truth, the evolving technology of artificial intelligence may well be taking over, but not quite how the filmmakers envisioned it. This week on Sea Change Radio, we speak with Andrew Winston, a sustainability strategist and bestselling author, about what AI means for the climate. We look at how AI can help various key sectors of the global economy become more efficient, examine the dilemma of AI’s seemingly insatiable energy needs, and discuss its potential to contribute to a carbon-free future.
Alex Wise | 01:32 – I am joined now on Sea Change Radio by Andrew Winston. Andrew is a sustainability strategist and a bestselling author. Andrew, welcome back to Sea Change Radio.
Andrew Winston | 01:42 – It’s good to be here.
Alex Wise | 01:44 – So you contribute regularly to MIT Sloan’s Management Review. Yep. And you have a recent piece on artificial intelligence ai, and it’s entitled, will AI Help or Hurt Sustainability? Yes. People either talk about AI as either the end of humanity as we know it, or this wonderful life changing technology, or they just completely tune it out and don’t want to know anything about it. What does that mean? So first, why don’t you explain what AI is on the most basic level, and then we can dive into the sustainability conundrum.
Andrew Winston | 02:24 – Well, first it’s a interesting question. What is AI? So there’s a long piece in MIT technology review recently called What is AI? And it’s a long article that basically comes to the conclusion of nobody really knows, and people use the phrase in many, many different ways. The way that author described it was, AI is a set of technologies that make computers do things that are thought to require intelligence when done by people. So it just means there’s things happening. It seems to be thinking even though at this point it really isn’t and seems to do kind of magical things, right? So the, the, the everyday interaction that we have with it, which is growing by the way, if people use, have used chat GPTs, you can, you can make it do fun things like write a poem about, you know, mid medieval times. But I use it to look at my writing and say, Hey, what’s a summary of this? Or What would you title this? Or, what are five things I should think about in this topic? And it’s just a really good partner, but it’s clearly getting embedded into companies of all stripes in every sector, and the hype about it and what it’s going to do for companies and potentially for, you know, our biggest challenges and, and the planet, the hype’s really big, but it’s unclear yet what’s living up to the hype. And, and there’s some downsides which we, which we can go into. But look, I think it’s a really big deal. I think it’s, it’s one of those technologies, when I first started playing with it, um, I had that same feeling like when I first Googled, or I first bought something on Amazon of, oh, this is different, right? And this is different. This is not like a tool like word and spell check that helps you write a little better. This is something of a totally different nature, and I don’t think we’re really ready for it. And it has serious impacts for our sustainability agenda. And that’s, that’s where of course, I’m most concerned.
Alex Wise | 04:11 – Well, it seems like the marketplace is already ready for it in a lot of ways. Just driving on 1 0 1 freeway here in San Francisco, you see a lot of AI billboards that are designed for CTOs, and you look at NVIDIA’s stock price, it’s like surpassed Apple. And most people wouldn’t even, aren’t even sure how to pronounce the company’s name. So this is a, a chip manufacturer, and that’s the future success of AI is kind of built into that market price, I think in many ways.
Andrew Winston | 04:43 – Yeah, they make the chips that most of the data centers and AI constructs, I guess, are, are using. So it’s made them incredibly valuable. Look, the scale of this is, is hard to get your, your head around and, and it’s growing so fast and changing so quickly. Like most of the problems that, you know, you probably cover and that we talk about in sustainability, they’re, they’re moving quicker than you think. And they’re hard to, they’re hard to get your head around and they’re, and they’re hard to understand how fast they’re moving. You know, the scale is really big. So it’s be getting embedded everywhere, right? We’re seeing it in all sorts of ways that we hadn’t before. It’s just appearing on your computer and helping you do stuff. It’s in different apps. It’s, you know, you’re probably talking to an AI chat bot when you do customer service chats with most companies. It’s just arising everywhere. And I think we’re barreling ahead. You’re in the region where the barreling forward in tech is, is the name of the game, right? That’s what you do in Silicon Valley and San Francisco. Without probably enough pausing on what it all means and what it does to us. There’s definitely people worried about it, but that’s not, it’s not slowing, right? There’s nothing really slowing it down. Companies are investing just gobs of money in bringing it into their business without really knowing what it’s going to do for them. It’s, it’s a fascinating time.
Alex Wise | 05:59 – It’s such a large palette to work with. We probably should skip over some of the non-environmental ramifications of ai, like deep fakes and the political issues and the police state. These are all important societal topics I suggest listeners educate themselves on. But for, for the purpose of this conversation, let’s stick to sustainability. I mentioned at the outset the AI conundrum. So why don’t you spell out this conundrum first, why don’t you explain some of the pros that you outline in your MIT Sloan piece in terms of sustainability and efficiencies, and then we can look at the problems with energy usage.
Andrew Winston | 06:39 – Yep. So look, the list of ways, in theory, AI is going to help us, you know, do our jobs better and help solve the big sustainability problems. The list is as long as the list of sustainability problems, I mean, everybody is claiming there’s going to be benefits because in essence, it, it can help find, um, better ways of doing things efficiencies, better than, than in theory we humans can. So everything from the built environment, you think, think about like building controls and ai, finding more patterns in cooling and heating and helping reduce the use of energy in the building, in managing the grid, which is going to be one of the most complicated and is growing to be one of the most complicated issues really, that, that humanity has to deal with, which is a grid with millions or billions of things drawing power, tens of thousands of giant solar and wind plants, and then millions of solar panels on people’s homes and trying to manage that flow back and forth. I think AI is going to turn out to be really critical.
Alex Wise | 07:38 – People might have like a nest or something like that on their homes, which is a smart technology for thermostat. Now, if you can kind of extrapolate that to the grid that you’re talking about, how are grid operators going to be able to use smart technology or AI in a way that they weren’t able to a decade ago?
Andrew Winston | 07:59 – Well, I, I think what’s changed fundamentally and is changing fast on the grid is that the grid was primarily, you know, everybody plugs in their stuff. They’ve got multiple devices in their home and there’s power plants, right? And there was a balancing of that load that people are doing at, at power plants and have been doing for decades. And there’s a certain, you know, ability to kind of ratchet up and turn up and down the power generation. What’s changing now is as we put up more and more solar panels, um, on people’s homes, for example, the homes are now producing power too. It’s a much more complicated grid. There’s many more sites of power production when it was really just, you know, in my region there’s a big coal plant further up in our county. There’s a nuclear plant in, in the region. Um, there’s hydro coming from Canada, there’s only so many, right? But now I have solar on my roof and there’s 40 different panels and they’re all generating different amounts of power. And so it’s incredibly complicated. And for, for a person or a group to manage is difficult, right? Because the, the grid is all about kind of balancing load, so you don’t have brownouts and blackouts, and there’s going to be more and more need to control how much power is coming from things. So even cars themselves are going to be these roaming batteries. As we get more and more millions of electric vehicles, they’re not just going to draw power. They’re increasingly going to be providing it. You know, if they’re parked in the garage during the day, they’re not going anywhere. They might be able to fill in some gaps in, in the load and then hold power, as mobile batteries, millions of them. So it’s really, it, it’s just much more complicated, right?And so they claim the utility world that AI is, is really one of the big solutions. I’m, you know, I’m not a programmer, I’m not an AI guy. It’s, it’s actually really hard to get very specific examples beyond kind of the general design of what I’ve just said is companies are talking a lot, but trying to get really specific examples is hard. ’cause I think it’s very blurry. What, what AI is versus analysis that a pro, you know, that a company programmer or someone does on their own. It’s, it’s this weird hybrid of the things we do and the things that, you know, a computer would naturally do. It’s something in between. And we should say there’s, in my understanding, kind of two or three different kinds of AI people talk about generative ai and there’s regular kind of regular, I mean, the stuff going on in things like buildings and, and grids. And I was going to say also like food and agriculture, um, there’s something called precision, precision agriculture that’s been building for years, which is, you know, taking satellite data and weather patterns and figuring out wind to water or dropping just the right amount of water or fertilizer, you know, square meter by square meter, like that’s already been happening as, you know, tractors and things become basically roaming giant computers. But now AI can help manage that real time. All of these things are kind of analytical ai, the stuff that gets all the flashy, you know, attention is, is the chat GPT. And, and you mentioned before like misinformation and DeepFakes and all that, those are a little different. They’re these large language models. They’re doing something kind of different than the analytics going on in companies that I think are just better, faster, more enhanced versions of the kind of analysis that you would do in a company or hire consultants to do. And take a look at your data and figure out the best way this can be much faster, much more real time can adjust, um, and, and often finds patterns that humans didn’t see and seizes it in a, in a different way. So the look, the potential’s huge, it is, right? We have fundamentally inefficient systems. We still lose so much energy in our grid. The food system has 30, 40% waste. We throw out 30 or 40%. So if there’s better ways to reduce spoilage to to fill trucks better, so the transportation is shorter, all of that we’re going to save, right? We’re going to save embedded carbon and water. And these are huge benefits. But, you know, there’s these huge benefits. And also on the social side, there’s a lot of, um, things that can help with, uh, you know, drug discovery and health and, and finding pa health patterns, lots of things to help wellbeing. But the, the downside of the conundrum is that the energy draw of AI is growing just unbelievably fast. And that is just driving a stake through all of the carbon goals and, and projections from the big tech companies that were trying to go to net zero and have been buying more renewables than anyone in the world. I mean, to the bottom line is both Microsoft and Google announced that their emissions are way up.
(Music Break)| 12:37
Alex Wise | 13:27 – This is Alex Wise on Sea Change Radio, and I’m speaking to Andrew Winston, he’s a sustainability strategist and bestselling author. So Andrew, we were talking about the hope that AI offers in terms of sustainability and, and fighting climate change. But you also mentioned this tremendous energy suck that these data centers are creating. And I remember being pitched by a VC maybe a decade ago at Sea Change Radio about a, a chip company that was in their portfolio that they considered a green technology because it was more energy efficient. And I kind of rolled my eyes, I was like, come on, that’s not a green technology. That’s just, you know, you’re trying to make a better chip. But as this technology becomes more and more present in our lives that energy consumption is looming over everything. Why don’t you expand if you can.
Andrew Winston | 14:19 – So there’s two big draws. One is just the training of ai. So, uh, you know, a new technology, a new product like chatGPT is trained by looking at patterns of language and understanding, you know, the flow of language. What chatGPT does is actually just with enormously complex mathematical models, it knows what’s the most likely next word in a typical sentence. It’s not actually thinking, it’s just creating sentences in the most logical mode. And the way they trained it was reading pretty much everything and like a significant portion of the internet went into it. So just imagine the power of processing enormous sums. And so each new creation, the next level of chatGPT, the other AI products, there’s a training timeframe that takes enormous energy. There’s entire data centers being built just to train. And then on an ongoing basis, every time you plug into chatGPT or you say make a picture of a dog playing poker or whatever, it’s drawing more energy. It’s doing more thinking. It’s, it’s more energy than just, you know, a simple Google search. And so we’re seeing now the projections of growth and data centers are, are profound. The International Energy Agency says we’re probably going to double data center demand in the next couple years. And that’s a lot, right? These, there’s a, a big energy draw already. Now, what the tech companies and data center companies say is that, you know, we’ve seen this before. We’ve seen massive growth in technology and we get more efficient all the time. So it’s not going to increase energy as much as, as people say, you know, in the 2010s I think data center size traffic went up, like sixfold and energy was basically flat. Because they just got more and more efficient. I don’t know if that’s going to hold this time. The AI is again, like kind of a different order of magnitude than just having more websites and having more data centers to kind of grow with the digital economy, it’s something different and it draws an enormous amount of power. So you now have concerns that major utilities that they can’t keep up and that could affect just grid reliability, right? At a time where we’re already seeing weather induced grid reliability problems. So it adds a whole new level and unfortunately it’s taken the big tech companies again, off their targets. They were on their way to net zero and now their emissions are growing again. And it’s pretty much entirely because of AI.
Alex Wise | 16:40 – So we know Moore’s Law, where it’s been kind of the, the holy grail of the industry to just continually increase computing power. It doubles every year, I think is Moore’s law basically. But at what point do we stop trying to increase the capacity and the speed and really focus on that efficiency as the holy grail?
Andrew Winston | 16:59 – Yeah, I mean to me, this is the core question really is, you know, in my, in the article you saw, I kind of laid out the pros and cons, but the, the next real question is what are we using it for? One thing I’ve heard that just sounds rings really true is, so they’re saying the data centers are going to get more and more efficient like Nvidia that you mentioned, put out a chip that I think used half as much energy for the same amount of processing. The problem is if you’re building one of these big data centers to train in AI and there’s efficiencies in it and the chip moves faster for the same amount of energy, you’re not going to cut the energy, you’re just going to make it a more powerful AI.
Alex Wise | 17:34 – Right? You’re just, you’re going to, you’re going to train more things right there. It’s a never-ending thirst almost.
Andrew Winston | 17:40 – And that’s what concerns me is the growth isn’t going to just get efficient. They’re going to use that efficiency to do more. And, and we hope that we can point that in the right way. But I, I’m working on this right now, kind of really asking, trying to find stories in companies. How are you really using this? How are you thinking about how you use it? Because it is a drain on resources to do this. We, the, the idea of the cloud was one of the greater marketing kind of ideas maybe in history to call technology the cloud. Like it’s just so light when data centers and servers, they’re real products, you have to manufacture them. There’s tons of energy, you know, digitizing the world does save a lot, but it also has a footprint. So the question to me is how do we, how do we use it and how do we prioritize? And, and the example that’s in my head is that just kind of an individual use, like I said, I’ll use chat GPT to help me edit, help me think through something I’m writing for a while I was starting to use, um, Bing and Dolly, which is another chat GPT product, which does art, you, you set, you know, you give it a, a statement and, and it makes a picture. And I was doing that for, to make a picture for blogs that I wrote, you know, just to have an image. And I realized, is this a good use? Like I, there’s stock photos that, that I can use that are people that someone took, a person took a picture and I can either pay royalties or, or they’ve put it up for stock and not use all this energy to make a, you know, made up picture. And, and so I feel like we got to start asking those, those questions of our own use of it. And I think at the, at the company level, I don’t think that’s really happening. They’re just barreling forward to try to stay competitive and use it wherever they can. And it is a real question right now, are all those benefits that are promised going to outpace just how much resource, resource and energy it takes to, to run it?
Alex Wise | 19:26 – You’ve mentioned how data centers are supposed to be doubling in capacity and energy usage in the not so distant future. But maybe you can dive in a little deeper and explain, put the energy usage into context if you can. You do this in the piece very effectively.
Andrew Winston | 19:43 – I mean, I’m borrowing from the International Energy Agency and others that the way they describe the scale was that it will double the needs of, of electricity for the grid for just from data centers just in the next couple years. And in total that will equal like Japan’s total consumption. You know, we’re getting up to a scale here where it’s a large pretty energy, you know, intense country level scale for this. And so you saw recently a bunch of utilities and some are, are probably padding their numbers like, like utilities do. But they, you know, Georgia Power in the US in Georgia obviously has said that, you know, its plan was to add I think 0.4 gigawatts of, of energy that was kind of its plan by 2030. And then all of a sudden it said actually we need 6.6 gigawatts, you know, we need 15, 20 times more additional power because of this draw. Some say they’re using that so they can go build some more natural gas plants and the things they really know how to build.
Alex Wise | 20:40 – And try to skirt regulators and things like that. Right?
Andrew Winston | 20:43 – Right. And, and you know, they can raise their rates, but if they’re even partially right, it’s a lot, right? Like I, you know, the, the, the numbers of how much it wants to add just in that one kind of region would be building the world’s largest solar farm, three nuclear reactors and a few other things. And you’d still wouldn’t hit that number. So it’s big.
Alex Wise | 21:04 – And that’s just for Georgia.
Andrew Winston | 21:06 – And this is just Georgia and their region. And it’s hard to cite and build a power plant, right? You got to have community support, you got to have the, the paperwork. Like, it’s not easy. So I, I think I’ve heard concern from companies just about the grid, just, you know, what’s going to happen with the grid as this keeps drawing and drawing as it grows so fast? Like are, are there going to have to be priorities? And the tech companies, they want, they expect to have always on, right? They need power 24 7 all the time so that when you search, you can find what you want. And if they’re hosting companies, um, that doesn’t go down. So there’s going to be these tensions. Um, and it’s, what what’s really worrisome is I was, I was pretty sure and the numbers look good, that we would hit peak, um, energy use or peak emissions globally soon. And now I’m not as sure, right? We were heading down huge shift to electric vehicles around the world, you know, a huge amount of building of renewable power. All of that is still happening, but this is now kind of a monkey wrench in, in the middle of it where we were generally about to peak and maybe start heading down and, and I’m not sure that’s going to happen now. Because this is the new shiny object, the new technology. And look, it’s exciting. I get it. It does stuff I’ve never, never seen before. It’s just we don’t really have time to work it through and double emissions and then come back down like, we’re out of time on climate. So how do we make sure we’re adding this without creating another big problem?
Alex Wise | 22:36 – Sticking with the Georgia Power example you give, and I know that it’s, we should take it with a grain of salt as you mentioned, but how does a huge utility get blindsided like that in terms of, oh yeah, we need 15 to 20 times what we had just projected and how much of that, what they were pointing to was tech-based? What was the excuse for their miscalculation?
Andrew Winston | 22:59 – Well, look, I think the story, at least as I understand it, is that the difference is pretty much this demand for AI and data sets. Remember chat CPT launched 20 months ago, like the very end of 2022, and it was the fastest adopted product in history. A hundred million people were using it within a month or two faster than TikTok, faster than Facebook. So it, it literally has come from nowhere. I mean, you get that Moore’s law thing of a doubling, the doubling of processing power every 18 months, 24 months, whatever it is. You, you hit these tipping points on things, on technologies when you’re done. And that’s the thing with ex with, you know, exponential growth, it’s really hard for us to see, see, and I think Georgia Power and, and all the projections from tech companies changed because it’s really hard to, to really internalize and deal with exponential change. We’re even in the tech world, we’re just not wired for it. I think as humans we kind of see things linearly and even in the tech world where they’ve been expecting this doubling, sometimes you don’t realize the ramifications, right? Sometimes you’re blockbuster and you don’t realize that streaming movies is going to go from taking hours to taking minutes and you don’t make it, right. So I think this is very equivalent all of a sudden there’s this capability that wasn’t there both for us as individuals to use and for companies to use and we’re using it and it’s just drawing demand, you know, really, really quickly. So I, I, I guess I don’t really blame them all, like, it just, it came upon us pretty fast. I think ChatGPT even surprised a lot in the AI tech world, just how good it was and how good it is, right? There’s more coming quickly.
Alex Wise | 24:41 – And I think more useful than people surmised in the beginning.
Andrew Winston | 24:45 – Yeah. I think I am trying to use it more and more just to try to see, I just think it’s going to be hard to do your job without some element of, of AI in most jobs in the near future. Just like anybody in a kind of knowledge worker, office worker, they know how to use Word or Excel or, you know, there’s, there’s just kind of the basics. This is similar, but you know, as I said, it just has so much more capability. I think, you know, we don’t know yet what it’s going to do to our jobs and how much it will enhance and how much it will replace.
Alex Wise | 25:16 – And from a policymakers standpoint, you can see the conundrum that they have. If you’re a politician, are you going to embrace AI technology and green light tax breaks for data centers the way we see all over the country. I just read a piece on ProPublica about how Washington State is giving tax breaks to these data centers that are using enormous amounts of energy. Do you want to seem like you are trying to keep progress from happening as a policymaker? Do you want to seem out of touch with technology? No, of course not. So it seems like they’re putting this position where they almost have to politically embrace this or they look like a fuddy-duddy.
Andrew Winston | 25:59 – Yeah, I mean, look, I I risk sounding like a luddite or anti-tech and bringing up these concerns. I’m not the only one obviously, and I’m, I’m very much pro technology. I do think the potential is enormous and we should be using it to make the world a lot more efficient. And the competition you’re describing between governments, it’s at the geopolitical level, right? The US is concerned about who’s going to own ai. Is it, is it China? And then state by state attracting companies, you know, everyone wants to, the communities may not as much, but the states want to build these things because there’s a lot of construction jobs at first. They don’t create a lot of ongoing jobs like a data center doesn’t have a ton of people working at it, right? This is another element of this new digital age where there’s just not as many people per dollar or production in a digital world. And, and this is a big concern, right? What is everyone going to do for a living if, if this starts to be able to do everything for us? And you end up, even though we’re talking about the, the biophysical environmental side on the social side, where you end up on this is talking about universal basic income. And I’m amazed to see all these big tech billionaires have been talking about it for years. They’re recognizing the, the chance that these technologies eliminate so many jobs that we need some kind of safety net for people in the world. That everyone gets a little bit, everyone gets maybe enough to eat or shelter so that if there’s not as many jobs, people can still survive and thrive. I mean, these are have huge ramifications for how society works and it’s not in our nature in the US to do that kind of distribution of wealth, right? We’re the least likely of the developed country. So I don’t know where that leaves us if there’s fewer and fewer jobs. In theory, it sounds great. Maybe people have more leisure time, but only if they have enough to live on.
Alex Wise | 27:48 – Well, in this world of encroaching artificial intelligence, it’s always a pleasure to speak to someone with great human intelligence like yourself. Andrew Winston, thanks so much for being my guest on Sea Change Radio.
Andrew Winston | 28:00 – Thank you. Thanks for having me.
Narrator | 28:17 – You’ve been listening to Sea Change Radio. Our intro music is by Sanford Lewis, and our outro music is by Alex Wise, additional music by Sidewinder and David Bowie. To read a transcript of this show, go to SeaChangeRadio.com to stream or download the show or subscribe to our podcast on our site or visit our archives to hear from Doris Kearns Goodwin, Gavin Newsom, Stewart Brand, and many others. And tune in to Sea Change Radio next week as we continue making connections for sustainability. For Sea Change Radio, I’m Alex Wise.
22 эпизодов
Manage episode 431491940 series 3381317
For Sci-Fi buffs, a future infused with AI may evoke unsettling images of HAL from Stanley Kubrick’s film, “2001: A Space Odyssey.” In truth, the evolving technology of artificial intelligence may well be taking over, but not quite how the filmmakers envisioned it. This week on Sea Change Radio, we speak with Andrew Winston, a sustainability strategist and bestselling author, about what AI means for the climate. We look at how AI can help various key sectors of the global economy become more efficient, examine the dilemma of AI’s seemingly insatiable energy needs, and discuss its potential to contribute to a carbon-free future.
Narrator | 00:02 – This is Sea Change Radio covering the shift to sustainability. I’m Alex Wise.
Andrew Winston | 00:17 – It is just, we don’t really have time to work it through and double emissions and then come back down like, we’re out of time on climate. So how do we make sure we’re adding this without creating another big problem?
Narrator | 00:32 – For Sci-Fi buffs, a future infused with AI may evoke unsettling images of HAL from Stanley Kubrick’s film, “2001: A Space Odyssey.” In truth, the evolving technology of artificial intelligence may well be taking over, but not quite how the filmmakers envisioned it. This week on Sea Change Radio, we speak with Andrew Winston, a sustainability strategist and bestselling author, about what AI means for the climate. We look at how AI can help various key sectors of the global economy become more efficient, examine the dilemma of AI’s seemingly insatiable energy needs, and discuss its potential to contribute to a carbon-free future.
Alex Wise | 01:32 – I am joined now on Sea Change Radio by Andrew Winston. Andrew is a sustainability strategist and a bestselling author. Andrew, welcome back to Sea Change Radio.
Andrew Winston | 01:42 – It’s good to be here.
Alex Wise | 01:44 – So you contribute regularly to MIT Sloan’s Management Review. Yep. And you have a recent piece on artificial intelligence ai, and it’s entitled, will AI Help or Hurt Sustainability? Yes. People either talk about AI as either the end of humanity as we know it, or this wonderful life changing technology, or they just completely tune it out and don’t want to know anything about it. What does that mean? So first, why don’t you explain what AI is on the most basic level, and then we can dive into the sustainability conundrum.
Andrew Winston | 02:24 – Well, first it’s a interesting question. What is AI? So there’s a long piece in MIT technology review recently called What is AI? And it’s a long article that basically comes to the conclusion of nobody really knows, and people use the phrase in many, many different ways. The way that author described it was, AI is a set of technologies that make computers do things that are thought to require intelligence when done by people. So it just means there’s things happening. It seems to be thinking even though at this point it really isn’t and seems to do kind of magical things, right? So the, the, the everyday interaction that we have with it, which is growing by the way, if people use, have used chat GPTs, you can, you can make it do fun things like write a poem about, you know, mid medieval times. But I use it to look at my writing and say, Hey, what’s a summary of this? Or What would you title this? Or, what are five things I should think about in this topic? And it’s just a really good partner, but it’s clearly getting embedded into companies of all stripes in every sector, and the hype about it and what it’s going to do for companies and potentially for, you know, our biggest challenges and, and the planet, the hype’s really big, but it’s unclear yet what’s living up to the hype. And, and there’s some downsides which we, which we can go into. But look, I think it’s a really big deal. I think it’s, it’s one of those technologies, when I first started playing with it, um, I had that same feeling like when I first Googled, or I first bought something on Amazon of, oh, this is different, right? And this is different. This is not like a tool like word and spell check that helps you write a little better. This is something of a totally different nature, and I don’t think we’re really ready for it. And it has serious impacts for our sustainability agenda. And that’s, that’s where of course, I’m most concerned.
Alex Wise | 04:11 – Well, it seems like the marketplace is already ready for it in a lot of ways. Just driving on 1 0 1 freeway here in San Francisco, you see a lot of AI billboards that are designed for CTOs, and you look at NVIDIA’s stock price, it’s like surpassed Apple. And most people wouldn’t even, aren’t even sure how to pronounce the company’s name. So this is a, a chip manufacturer, and that’s the future success of AI is kind of built into that market price, I think in many ways.
Andrew Winston | 04:43 – Yeah, they make the chips that most of the data centers and AI constructs, I guess, are, are using. So it’s made them incredibly valuable. Look, the scale of this is, is hard to get your, your head around and, and it’s growing so fast and changing so quickly. Like most of the problems that, you know, you probably cover and that we talk about in sustainability, they’re, they’re moving quicker than you think. And they’re hard to, they’re hard to get your head around and they’re, and they’re hard to understand how fast they’re moving. You know, the scale is really big. So it’s be getting embedded everywhere, right? We’re seeing it in all sorts of ways that we hadn’t before. It’s just appearing on your computer and helping you do stuff. It’s in different apps. It’s, you know, you’re probably talking to an AI chat bot when you do customer service chats with most companies. It’s just arising everywhere. And I think we’re barreling ahead. You’re in the region where the barreling forward in tech is, is the name of the game, right? That’s what you do in Silicon Valley and San Francisco. Without probably enough pausing on what it all means and what it does to us. There’s definitely people worried about it, but that’s not, it’s not slowing, right? There’s nothing really slowing it down. Companies are investing just gobs of money in bringing it into their business without really knowing what it’s going to do for them. It’s, it’s a fascinating time.
Alex Wise | 05:59 – It’s such a large palette to work with. We probably should skip over some of the non-environmental ramifications of ai, like deep fakes and the political issues and the police state. These are all important societal topics I suggest listeners educate themselves on. But for, for the purpose of this conversation, let’s stick to sustainability. I mentioned at the outset the AI conundrum. So why don’t you spell out this conundrum first, why don’t you explain some of the pros that you outline in your MIT Sloan piece in terms of sustainability and efficiencies, and then we can look at the problems with energy usage.
Andrew Winston | 06:39 – Yep. So look, the list of ways, in theory, AI is going to help us, you know, do our jobs better and help solve the big sustainability problems. The list is as long as the list of sustainability problems, I mean, everybody is claiming there’s going to be benefits because in essence, it, it can help find, um, better ways of doing things efficiencies, better than, than in theory we humans can. So everything from the built environment, you think, think about like building controls and ai, finding more patterns in cooling and heating and helping reduce the use of energy in the building, in managing the grid, which is going to be one of the most complicated and is growing to be one of the most complicated issues really, that, that humanity has to deal with, which is a grid with millions or billions of things drawing power, tens of thousands of giant solar and wind plants, and then millions of solar panels on people’s homes and trying to manage that flow back and forth. I think AI is going to turn out to be really critical.
Alex Wise | 07:38 – People might have like a nest or something like that on their homes, which is a smart technology for thermostat. Now, if you can kind of extrapolate that to the grid that you’re talking about, how are grid operators going to be able to use smart technology or AI in a way that they weren’t able to a decade ago?
Andrew Winston | 07:59 – Well, I, I think what’s changed fundamentally and is changing fast on the grid is that the grid was primarily, you know, everybody plugs in their stuff. They’ve got multiple devices in their home and there’s power plants, right? And there was a balancing of that load that people are doing at, at power plants and have been doing for decades. And there’s a certain, you know, ability to kind of ratchet up and turn up and down the power generation. What’s changing now is as we put up more and more solar panels, um, on people’s homes, for example, the homes are now producing power too. It’s a much more complicated grid. There’s many more sites of power production when it was really just, you know, in my region there’s a big coal plant further up in our county. There’s a nuclear plant in, in the region. Um, there’s hydro coming from Canada, there’s only so many, right? But now I have solar on my roof and there’s 40 different panels and they’re all generating different amounts of power. And so it’s incredibly complicated. And for, for a person or a group to manage is difficult, right? Because the, the grid is all about kind of balancing load, so you don’t have brownouts and blackouts, and there’s going to be more and more need to control how much power is coming from things. So even cars themselves are going to be these roaming batteries. As we get more and more millions of electric vehicles, they’re not just going to draw power. They’re increasingly going to be providing it. You know, if they’re parked in the garage during the day, they’re not going anywhere. They might be able to fill in some gaps in, in the load and then hold power, as mobile batteries, millions of them. So it’s really, it, it’s just much more complicated, right?And so they claim the utility world that AI is, is really one of the big solutions. I’m, you know, I’m not a programmer, I’m not an AI guy. It’s, it’s actually really hard to get very specific examples beyond kind of the general design of what I’ve just said is companies are talking a lot, but trying to get really specific examples is hard. ’cause I think it’s very blurry. What, what AI is versus analysis that a pro, you know, that a company programmer or someone does on their own. It’s, it’s this weird hybrid of the things we do and the things that, you know, a computer would naturally do. It’s something in between. And we should say there’s, in my understanding, kind of two or three different kinds of AI people talk about generative ai and there’s regular kind of regular, I mean, the stuff going on in things like buildings and, and grids. And I was going to say also like food and agriculture, um, there’s something called precision, precision agriculture that’s been building for years, which is, you know, taking satellite data and weather patterns and figuring out wind to water or dropping just the right amount of water or fertilizer, you know, square meter by square meter, like that’s already been happening as, you know, tractors and things become basically roaming giant computers. But now AI can help manage that real time. All of these things are kind of analytical ai, the stuff that gets all the flashy, you know, attention is, is the chat GPT. And, and you mentioned before like misinformation and DeepFakes and all that, those are a little different. They’re these large language models. They’re doing something kind of different than the analytics going on in companies that I think are just better, faster, more enhanced versions of the kind of analysis that you would do in a company or hire consultants to do. And take a look at your data and figure out the best way this can be much faster, much more real time can adjust, um, and, and often finds patterns that humans didn’t see and seizes it in a, in a different way. So the look, the potential’s huge, it is, right? We have fundamentally inefficient systems. We still lose so much energy in our grid. The food system has 30, 40% waste. We throw out 30 or 40%. So if there’s better ways to reduce spoilage to to fill trucks better, so the transportation is shorter, all of that we’re going to save, right? We’re going to save embedded carbon and water. And these are huge benefits. But, you know, there’s these huge benefits. And also on the social side, there’s a lot of, um, things that can help with, uh, you know, drug discovery and health and, and finding pa health patterns, lots of things to help wellbeing. But the, the downside of the conundrum is that the energy draw of AI is growing just unbelievably fast. And that is just driving a stake through all of the carbon goals and, and projections from the big tech companies that were trying to go to net zero and have been buying more renewables than anyone in the world. I mean, to the bottom line is both Microsoft and Google announced that their emissions are way up.
(Music Break)| 12:37
Alex Wise | 13:27 – This is Alex Wise on Sea Change Radio, and I’m speaking to Andrew Winston, he’s a sustainability strategist and bestselling author. So Andrew, we were talking about the hope that AI offers in terms of sustainability and, and fighting climate change. But you also mentioned this tremendous energy suck that these data centers are creating. And I remember being pitched by a VC maybe a decade ago at Sea Change Radio about a, a chip company that was in their portfolio that they considered a green technology because it was more energy efficient. And I kind of rolled my eyes, I was like, come on, that’s not a green technology. That’s just, you know, you’re trying to make a better chip. But as this technology becomes more and more present in our lives that energy consumption is looming over everything. Why don’t you expand if you can.
Andrew Winston | 14:19 – So there’s two big draws. One is just the training of ai. So, uh, you know, a new technology, a new product like chatGPT is trained by looking at patterns of language and understanding, you know, the flow of language. What chatGPT does is actually just with enormously complex mathematical models, it knows what’s the most likely next word in a typical sentence. It’s not actually thinking, it’s just creating sentences in the most logical mode. And the way they trained it was reading pretty much everything and like a significant portion of the internet went into it. So just imagine the power of processing enormous sums. And so each new creation, the next level of chatGPT, the other AI products, there’s a training timeframe that takes enormous energy. There’s entire data centers being built just to train. And then on an ongoing basis, every time you plug into chatGPT or you say make a picture of a dog playing poker or whatever, it’s drawing more energy. It’s doing more thinking. It’s, it’s more energy than just, you know, a simple Google search. And so we’re seeing now the projections of growth and data centers are, are profound. The International Energy Agency says we’re probably going to double data center demand in the next couple years. And that’s a lot, right? These, there’s a, a big energy draw already. Now, what the tech companies and data center companies say is that, you know, we’ve seen this before. We’ve seen massive growth in technology and we get more efficient all the time. So it’s not going to increase energy as much as, as people say, you know, in the 2010s I think data center size traffic went up, like sixfold and energy was basically flat. Because they just got more and more efficient. I don’t know if that’s going to hold this time. The AI is again, like kind of a different order of magnitude than just having more websites and having more data centers to kind of grow with the digital economy, it’s something different and it draws an enormous amount of power. So you now have concerns that major utilities that they can’t keep up and that could affect just grid reliability, right? At a time where we’re already seeing weather induced grid reliability problems. So it adds a whole new level and unfortunately it’s taken the big tech companies again, off their targets. They were on their way to net zero and now their emissions are growing again. And it’s pretty much entirely because of AI.
Alex Wise | 16:40 – So we know Moore’s Law, where it’s been kind of the, the holy grail of the industry to just continually increase computing power. It doubles every year, I think is Moore’s law basically. But at what point do we stop trying to increase the capacity and the speed and really focus on that efficiency as the holy grail?
Andrew Winston | 16:59 – Yeah, I mean to me, this is the core question really is, you know, in my, in the article you saw, I kind of laid out the pros and cons, but the, the next real question is what are we using it for? One thing I’ve heard that just sounds rings really true is, so they’re saying the data centers are going to get more and more efficient like Nvidia that you mentioned, put out a chip that I think used half as much energy for the same amount of processing. The problem is if you’re building one of these big data centers to train in AI and there’s efficiencies in it and the chip moves faster for the same amount of energy, you’re not going to cut the energy, you’re just going to make it a more powerful AI.
Alex Wise | 17:34 – Right? You’re just, you’re going to, you’re going to train more things right there. It’s a never-ending thirst almost.
Andrew Winston | 17:40 – And that’s what concerns me is the growth isn’t going to just get efficient. They’re going to use that efficiency to do more. And, and we hope that we can point that in the right way. But I, I’m working on this right now, kind of really asking, trying to find stories in companies. How are you really using this? How are you thinking about how you use it? Because it is a drain on resources to do this. We, the, the idea of the cloud was one of the greater marketing kind of ideas maybe in history to call technology the cloud. Like it’s just so light when data centers and servers, they’re real products, you have to manufacture them. There’s tons of energy, you know, digitizing the world does save a lot, but it also has a footprint. So the question to me is how do we, how do we use it and how do we prioritize? And, and the example that’s in my head is that just kind of an individual use, like I said, I’ll use chat GPT to help me edit, help me think through something I’m writing for a while I was starting to use, um, Bing and Dolly, which is another chat GPT product, which does art, you, you set, you know, you give it a, a statement and, and it makes a picture. And I was doing that for, to make a picture for blogs that I wrote, you know, just to have an image. And I realized, is this a good use? Like I, there’s stock photos that, that I can use that are people that someone took, a person took a picture and I can either pay royalties or, or they’ve put it up for stock and not use all this energy to make a, you know, made up picture. And, and so I feel like we got to start asking those, those questions of our own use of it. And I think at the, at the company level, I don’t think that’s really happening. They’re just barreling forward to try to stay competitive and use it wherever they can. And it is a real question right now, are all those benefits that are promised going to outpace just how much resource, resource and energy it takes to, to run it?
Alex Wise | 19:26 – You’ve mentioned how data centers are supposed to be doubling in capacity and energy usage in the not so distant future. But maybe you can dive in a little deeper and explain, put the energy usage into context if you can. You do this in the piece very effectively.
Andrew Winston | 19:43 – I mean, I’m borrowing from the International Energy Agency and others that the way they describe the scale was that it will double the needs of, of electricity for the grid for just from data centers just in the next couple years. And in total that will equal like Japan’s total consumption. You know, we’re getting up to a scale here where it’s a large pretty energy, you know, intense country level scale for this. And so you saw recently a bunch of utilities and some are, are probably padding their numbers like, like utilities do. But they, you know, Georgia Power in the US in Georgia obviously has said that, you know, its plan was to add I think 0.4 gigawatts of, of energy that was kind of its plan by 2030. And then all of a sudden it said actually we need 6.6 gigawatts, you know, we need 15, 20 times more additional power because of this draw. Some say they’re using that so they can go build some more natural gas plants and the things they really know how to build.
Alex Wise | 20:40 – And try to skirt regulators and things like that. Right?
Andrew Winston | 20:43 – Right. And, and you know, they can raise their rates, but if they’re even partially right, it’s a lot, right? Like I, you know, the, the, the numbers of how much it wants to add just in that one kind of region would be building the world’s largest solar farm, three nuclear reactors and a few other things. And you’d still wouldn’t hit that number. So it’s big.
Alex Wise | 21:04 – And that’s just for Georgia.
Andrew Winston | 21:06 – And this is just Georgia and their region. And it’s hard to cite and build a power plant, right? You got to have community support, you got to have the, the paperwork. Like, it’s not easy. So I, I think I’ve heard concern from companies just about the grid, just, you know, what’s going to happen with the grid as this keeps drawing and drawing as it grows so fast? Like are, are there going to have to be priorities? And the tech companies, they want, they expect to have always on, right? They need power 24 7 all the time so that when you search, you can find what you want. And if they’re hosting companies, um, that doesn’t go down. So there’s going to be these tensions. Um, and it’s, what what’s really worrisome is I was, I was pretty sure and the numbers look good, that we would hit peak, um, energy use or peak emissions globally soon. And now I’m not as sure, right? We were heading down huge shift to electric vehicles around the world, you know, a huge amount of building of renewable power. All of that is still happening, but this is now kind of a monkey wrench in, in the middle of it where we were generally about to peak and maybe start heading down and, and I’m not sure that’s going to happen now. Because this is the new shiny object, the new technology. And look, it’s exciting. I get it. It does stuff I’ve never, never seen before. It’s just we don’t really have time to work it through and double emissions and then come back down like, we’re out of time on climate. So how do we make sure we’re adding this without creating another big problem?
Alex Wise | 22:36 – Sticking with the Georgia Power example you give, and I know that it’s, we should take it with a grain of salt as you mentioned, but how does a huge utility get blindsided like that in terms of, oh yeah, we need 15 to 20 times what we had just projected and how much of that, what they were pointing to was tech-based? What was the excuse for their miscalculation?
Andrew Winston | 22:59 – Well, look, I think the story, at least as I understand it, is that the difference is pretty much this demand for AI and data sets. Remember chat CPT launched 20 months ago, like the very end of 2022, and it was the fastest adopted product in history. A hundred million people were using it within a month or two faster than TikTok, faster than Facebook. So it, it literally has come from nowhere. I mean, you get that Moore’s law thing of a doubling, the doubling of processing power every 18 months, 24 months, whatever it is. You, you hit these tipping points on things, on technologies when you’re done. And that’s the thing with ex with, you know, exponential growth, it’s really hard for us to see, see, and I think Georgia Power and, and all the projections from tech companies changed because it’s really hard to, to really internalize and deal with exponential change. We’re even in the tech world, we’re just not wired for it. I think as humans we kind of see things linearly and even in the tech world where they’ve been expecting this doubling, sometimes you don’t realize the ramifications, right? Sometimes you’re blockbuster and you don’t realize that streaming movies is going to go from taking hours to taking minutes and you don’t make it, right. So I think this is very equivalent all of a sudden there’s this capability that wasn’t there both for us as individuals to use and for companies to use and we’re using it and it’s just drawing demand, you know, really, really quickly. So I, I, I guess I don’t really blame them all, like, it just, it came upon us pretty fast. I think ChatGPT even surprised a lot in the AI tech world, just how good it was and how good it is, right? There’s more coming quickly.
Alex Wise | 24:41 – And I think more useful than people surmised in the beginning.
Andrew Winston | 24:45 – Yeah. I think I am trying to use it more and more just to try to see, I just think it’s going to be hard to do your job without some element of, of AI in most jobs in the near future. Just like anybody in a kind of knowledge worker, office worker, they know how to use Word or Excel or, you know, there’s, there’s just kind of the basics. This is similar, but you know, as I said, it just has so much more capability. I think, you know, we don’t know yet what it’s going to do to our jobs and how much it will enhance and how much it will replace.
Alex Wise | 25:16 – And from a policymakers standpoint, you can see the conundrum that they have. If you’re a politician, are you going to embrace AI technology and green light tax breaks for data centers the way we see all over the country. I just read a piece on ProPublica about how Washington State is giving tax breaks to these data centers that are using enormous amounts of energy. Do you want to seem like you are trying to keep progress from happening as a policymaker? Do you want to seem out of touch with technology? No, of course not. So it seems like they’re putting this position where they almost have to politically embrace this or they look like a fuddy-duddy.
Andrew Winston | 25:59 – Yeah, I mean, look, I I risk sounding like a luddite or anti-tech and bringing up these concerns. I’m not the only one obviously, and I’m, I’m very much pro technology. I do think the potential is enormous and we should be using it to make the world a lot more efficient. And the competition you’re describing between governments, it’s at the geopolitical level, right? The US is concerned about who’s going to own ai. Is it, is it China? And then state by state attracting companies, you know, everyone wants to, the communities may not as much, but the states want to build these things because there’s a lot of construction jobs at first. They don’t create a lot of ongoing jobs like a data center doesn’t have a ton of people working at it, right? This is another element of this new digital age where there’s just not as many people per dollar or production in a digital world. And, and this is a big concern, right? What is everyone going to do for a living if, if this starts to be able to do everything for us? And you end up, even though we’re talking about the, the biophysical environmental side on the social side, where you end up on this is talking about universal basic income. And I’m amazed to see all these big tech billionaires have been talking about it for years. They’re recognizing the, the chance that these technologies eliminate so many jobs that we need some kind of safety net for people in the world. That everyone gets a little bit, everyone gets maybe enough to eat or shelter so that if there’s not as many jobs, people can still survive and thrive. I mean, these are have huge ramifications for how society works and it’s not in our nature in the US to do that kind of distribution of wealth, right? We’re the least likely of the developed country. So I don’t know where that leaves us if there’s fewer and fewer jobs. In theory, it sounds great. Maybe people have more leisure time, but only if they have enough to live on.
Alex Wise | 27:48 – Well, in this world of encroaching artificial intelligence, it’s always a pleasure to speak to someone with great human intelligence like yourself. Andrew Winston, thanks so much for being my guest on Sea Change Radio.
Andrew Winston | 28:00 – Thank you. Thanks for having me.
Narrator | 28:17 – You’ve been listening to Sea Change Radio. Our intro music is by Sanford Lewis, and our outro music is by Alex Wise, additional music by Sidewinder and David Bowie. To read a transcript of this show, go to SeaChangeRadio.com to stream or download the show or subscribe to our podcast on our site or visit our archives to hear from Doris Kearns Goodwin, Gavin Newsom, Stewart Brand, and many others. And tune in to Sea Change Radio next week as we continue making connections for sustainability. For Sea Change Radio, I’m Alex Wise.
22 эпизодов
모든 에피소드
×Добро пожаловать в Player FM!
Player FM сканирует Интернет в поисках высококачественных подкастов, чтобы вы могли наслаждаться ими прямо сейчас. Это лучшее приложение для подкастов, которое работает на Android, iPhone и веб-странице. Зарегистрируйтесь, чтобы синхронизировать подписки на разных устройствах.