Our Tech has a Climate Problem: Here's how we solve it

0:00This message comes from AT &T. Staying connected matters. That's why AT &T has connectivity you can depend on. or they'll proactively make it right. That's the AT &T Guarantee. Terms and conditions apply. Visit att.com slash guarantee for details. This is the TED Radio Hour. Each week, groundbreaking TED Talks. Our job now is to dream big. Delivered at TED conferences. To bring about the future we want to see. Around the world. To understand who we are. From those talks, we bring you speakers and ideas that will surprise you. You just don't know what you're going to find. Challenge you. We truly have to ask ourselves, like, why is it noteworthy?

0:44And even change you. I literally feel like I'm a different person. Yes. Do you feel that way? Ideas worth spreading. From TED and NPR. I'm Manoush Zomorodi. And I'd like you to think of all the things you own. So many of them need electricity these days, which used to be pretty easy to quantify. That light in the hallway, it's a 60-watt incandescent bulb, so it uses 60 watts an hour. Maybe you now have an LED bulb in there. That uses around 10. Your washing machine, that needs around 1 ,500 watts per hour. And now what about your phone? Well, a dead iPhone battery needs about three hours to fully charge, usually at around five watts an hour.

1:38Once that phone is charged, though, its energy use starts to multiply. Take an email, for example. There's the router that's connecting it to the Internet. The energy needed to write, send and store the message on your device and the cloud, both for you and the person on the receiving end. And now what if you added AI into the mix? Let's say you use ChatGPT to help you craft that email. It's really hard to quantify exactly how much energy a single query on ChatGPT uses, never mind how much was used to train that AI model to begin with. Yeah, the thing is, is that the current way we train these models is all, I call it brute force.

2:21This is Sasha Luccioni. She's an AI researcher. Like terabytes of data, essentially, from the internet, from books, from Wikipedia, what have you. And so this process, especially for large, specifically large, large language models, so billions of parameters, can take months. Sasha has become obsessed with trying to figure out just how much energy AI uses. It is a complicated endeavor, but we know it's significant. Take, for example, one early estimate by researchers. So it all started, I think, in 2019 or so. A first paper came out that estimated that training a large language model like ChatGPT emits as much carbon as five cars in their lifetime.

3:08So that's like the manufacturing of the cars and like the usage of the cars. And that's a lot. By one more recent estimate, training one of these models like ChatGPT3 took as much energy as 130 American homes in one year. And yet, climate scientists believe technology is crucial to helping find solutions to our climate crisis. But what can we do when technology itself is a problem for our planet? Today on the show, tech's climate conundrum. Ideas about how AI, electric cars, and satellite monitoring will help us reduce emissions. And how we can confront their environmental downsides, too. So, back to Sasha Luccioni.

3:58As AI gets embedded into everything we do, she says tech companies are more and more reluctant to be upfront about how much energy their products use. Something I discovered for myself. I asked ChatGPT how much energy it costs to query it. And do you mind if I read you the answer it gave me? It says, as an AI language model developed by OpenAI, I don't have direct access to information regarding my power consumption. OpenAI strives to optimize its infrastructure for efficiency and sustainability. but precise energy consumption figures would depend on various factors like server load, optimization techniques, and data center efficiency, among others.

4:42So, no, I don't have a specific answer to your question. I mean, all of that is true. But that's really frustrating. But that doesn't answer your question. No, it doesn't. And it's saying, I find it strange that the AI is telling me that its maker won't share any of the information. Is that what it's saying, basically? Yeah. It's kind of like what I've noticed a lot lately in the conversation about sustainability, environmental impact. It's like once you start gathering the information, you have less plausible deniability. And so it's like, well, we don't have the information. So they don't want to know.

5:16It's really complicated. So, you know, we haven't gone through the actions necessary to gather that information, right? And I've heard these answers coming up a lot. Well, AI is so complicated. We don't really have that information. Because AI is so distributed and complex. And so it's really even hard to, like, pin down, do you include, I don't know, Internet of Things? Do you include smartphones? Do you include, like, right, like, there's all these different, so it's like, well, I mean, no, we don't have the answer. So to try and better estimate how much some AI use, she and other researchers built their own large language model.

5:46They called it Bloom. Last year, I was part of the Big Science Initiative, which brought together a thousand researchers from all over the world to create Bloom. Sasha Luccioni continues from the TED stage. The first open large language model, like Jachibiti, but with an emphasis on ethics, transparency, and consent. And the study I led that looked at Bloom's environmental impacts found that just training it used as much energy as 30 homes in a whole year and emitted 25 tons of carbon dioxide, just so somebody can use this model to tell a knock-knock joke. And this might not seem like a lot, but other similar large language models like GPT-3 emit 20 times more carbon.

6:30But the thing is, tech companies aren't measuring this stuff. They're not disclosing it. And so this is probably only the tip of the iceberg, even if it is a melting one. And in recent years, we've seen AI models balloon in size because the current trend in AI is bigger is better. But please don't get me started on why that's the case. In any case, we've seen large language models in particular grow 2 ,000 times in size over the last five years. And of course, their environmental costs are rising as well. The most recent work I led found that switching out a smaller, more efficient model for a larger language model emits 14 times more carbon for the same task.

7:10And as we're putting in these models into cell phones and search engines and smart fridges and speakers, the environmental costs are really piling up quickly. Recently, I was giving a talk and someone came up to me and they're like, yeah, we noticed that now that we're switching to local AI solutions, our energy usage is going up. And I was like, yeah, funny how that happens. And they're like, well. And the thing is, when we talk about AI or when we hear AI being discussed, it's often this ephemeral thing that people are just like, oh, AI can read, AI can write, AI can make images. and so people don't tend to have a very physical concept linked to AI, but it actually does run on a computer.

7:51It's very physical. To help companies who train or just use AI models, Sasha built a tool that estimates how much carbon they're admitting. It's called Code Carbon. I helped create Code Carbon, a tool that runs in parallel to AI training code that estimates the amount of energy it consumes and the amount of carbon it emits. And using a tool like this can help us make informed choices, like choosing one model over the other because it's more sustainable, or deploying AI models on renewable energy, which can drastically reduce their emissions. So, for example, you know, if I'm training a large language model on a thousand GPUs that are all in Texas versus a thousand GPUs that are all in Montreal, which has hydroelectricity, it's going to be the same amount of energy, but a hugely different amount of emissions, just because the energy is generated differently.

8:43And so we want to help give people that extra information. And so Code Carbon is used by lots of developers. It's become kind of one of the de facto standards for reporting energy. We recently got a Mozilla Foundation grant for it. We're trying to essentially build it out and give it more functionalities and make it a little bit more user-friendly. So if I'm the CEO of a company, I can use your tool to figure out which AI to use more efficiently, to be more sustainable. But what if I'm just like a regular person? Let's say I am using AI too, but I want to choose the least carbon-admitting AI. Is there a tool for me?

9:20I'm really a bit of a dreamer, but I think that if people are presented with information, I mean, they'll make choices that are kind of coherent with what they care about, what their values are. And so I'm working on an Energy Star rating for AI models. So essentially, similarly to appliances or all sorts of, you know, even vehicles, they have these ratings that say, well, essentially, like, this is how it compares to other models. And so I want to do that for AI models for different tasks without going into too many details, but having like an A plus to an F or to a D rating. Essentially, so that people, when they're picking a model and they're like, okay, I want to do, for example, image generation, they kind of know, at least like at a high level, how the different models measure up.

10:05Can we talk about what drew you to AI to begin with? The good parts of AI, the problem solving it can do? Do you have a favorite example of AI being used efficiently to help the environment? Yeah, I mean, essentially AI actually is not a single technology as well. It's got a lot of underlying approaches. And so maybe people know most, you know, generative AI like ChatGPT, but there's actually all sorts of what's called unsupervised learning. So there's a really cool startup on the West Coast called Rainforest Connection, which is using a similar technique for detecting illegal logging and deforestation in the Amazon.

10:47And essentially, they have these very, very lightweight AI models that are running on old cell phones, actually, with solar panels. So they're, like, very, very green. And they listen to the jungle. And essentially, they know what constitutes rain and, you know, parrots and monkeys and whatnot. And then if they hear a chainsaw or a truck or even human voices, they can flag that as an anomaly. And then they partner it up with like rangers in different countries to send people to check if there's deforestation happening. And I find that really cool because on one hand it's very low tech. And on the other hand, it is in partnership with, you know, rangers who are already doing this work.

11:26They're already trying to stop illegal deforestation in the Amazon. And it's helping them, you know, target specific areas better. That sounds amazing. But let's say I'm at home thinking about these new tools and I, as a person, just want to do right by the planet. Is the bottom line that I should think twice before I use AI unnecessarily? I totally agree with that. The only usage I found for ChatGPT, and once again, I'm trying not to be radical here, is once I write a scientific article or a paper or something, I take the abstract, like the resume, and I put it into ChatGPT to get it to come up with funny titles because I'm typically really bad at that.

12:11And it can come up with some really good ones and then like some puns or some like word plays or whatever. And then I'll tend to like put a bunch of them together and make something that I like. And so that's like my only go-to usage of generative AI. Like tech is so pervasive in our lives. Like we shouldn't necessarily try to beat ourselves up about it. But, you know, trying to kind of stay critical. Like, you know, this thing that's being sold to me, do I really need it? It's like before, you know, I'm going to use whatever, chat GPT as a calculator. think about the options that maybe isn't AI, but that does the task that I'm trying to do.

12:45It's funny, in France, there's this concept of digital frugality that I really like, or digital sobriety. And I find that it's really interesting because they're really thinking about, like, do we need this tech? Does this thing have to be connected? Does it have to be a smart device? And I think that might help if we did it in other places too. That's AI researcher Sasha Luccioni. You can see her full talk at TED.com. Today on the show, tech's climate conundrum. It's the TED Radio Hour from NPR. I'm Manoush Zomorodi, and we'll be right back.

13:33This message comes from the Michael J. Fox Foundation. Did you know that every six minutes, someone in the U.S. is diagnosed with Parkinson's? It's the fastest growing brain disease in the world. Chances are, you or someone you know is impacted by Parkinson's. The Michael J. Fox Foundation is on a mission to find a cure. Now is a critical time for Parkinson's research, and your support will help drive progress toward a cure. Get involved today. Visit michaeljfox.org slash get involved. This message comes from NPR sponsor, Informatica. Everybody's ready for AI, except your data. Informatica offers the only end-to-end AI-powered cloud data management platform that connects, manages, and unifies your data across virtually any multi-cloud or hybrid environment.

14:21Informatica accelerates your AI readiness and scale at a pace that allows you to pay for only what you use, without compromise. Get your data ready for AI at Informatica.com slash AI. Informatica, where data and AI come to life. This message comes from NPR sponsor, Walmart. Did you know that Walmart is one of Fortune's 100 best companies to work for in 2025? It's no surprise. Walmart is investing$1 billion in training and development to help associates build lasting careers with or without a degree. In fact, 75 % of Walmart management started in hourly rolls. With benefits like health care, paid parental leave, and tuition support, Walmart isn't just a place to work.

15:09It's a place to grow. Learn more at walmart.com slash spark. Hey, before we get back to the show, I want to let you know about our next bonus episode for TED Radio Hour Plus. It's an update from psychedelics pioneer Rick Doblin. So Rick has spent years researching how MDMA can be used to treat PTSD. And the FDA is meeting to talk about whether to approve this treatment in the next few weeks. I talked about that and much more with Rick at the recent TED conference in Vancouver. He is a lovely and fascinating guy. That bonus episode is coming on Wednesday. If you're not a TED Radio Hour Plus supporter yet, please join your fellow listeners to get bonus content and all our episodes sponsor free.

15:58Just go to plus.npr.org slash TED or give it a try right in the Apple Podcasts app. It's the TED Radio Hour from NPR. I'm Manoush Zomorodi. On the show today, tech's climate conundrum. We've all gotten the message. the era of gas-guzzling cars is slowly drawing to a close. Electric vehicles are the future. My administration is investing more than$135 billion to advance America's electric vehicle future. But we've also heard that the batteries needed to power these cars require rare minerals that are hard to get. And mining for those minerals is terrible for the environment. But is that just the price we need to pay in the short term for the long-term benefit of transitioning to electric?

16:52The green energy transition, as it's often called by the politicians, it's actually maybe not as green as we once thought because it too will require a lot of mining. We're about to explore the trade-offs by tracing where exactly these minerals come from and how mining for them impacts the land and people who live there. Starting with NPR correspondent Kirk Ziegler, who's been reporting from the site of a new mine in Nevada at a place called Thacker Pass. It's hard to describe just how vast it looks, even for someone who's used to these big, wide-open western landscapes. It's ringed by mountains, huge mountains, with this massive carpet of three or four foot high sagebrush plants that unfold on the horizon for as far as you can see.

17:50Kirk says there's a push here to mine these minerals for EVs and do it fast. Minerals like cobalt, nickel, lithium, copper, and graphite. Over the next 15 years, the U.S. government estimates the demand for lithium alone will increase by 4 ,000 percent. I mean, this is a huge amount of lithium in Nevada that's sitting there in this deposit. It's often said that it's the largest lithium deposit in North America. And so now all of a sudden these lands that were largely forgotten are now very much in play and it's very much the center of controversy. That controversy pitted a mining company called Lithium Americas and the U.S.

18:35Bureau of Land Management against environmentalists and some nearby Native tribes. We're at Packer Pass and we've got protesters blocking down here. Back in 2021, Kirk talked to activists protesting at the site. One of them was Gary McKinney. Right here, I've been camped here for about a month and a half now. He's against the mine because it's on historic land and would destroy the natural environment. We can't flush out all the water from out of here and rip up everything that is out here and call it green energy. That's greenwashing. I'm prepared to stay out here and oppose this mine for as long as it takes.

19:18As long as it takes. That fight lasted three years. Eventually, the mining lease was upheld in court, and excavation began in 2023. General Motors has signed on to buy all the lithium produced here in the first decade. One of the chairmen of one of the tribes put it to me this way, that Native Americans are once again being asked to get out of the way for American progress. Yeah, that story is really repeated across the globe, whether it's South America or North America or Australia, across Asia. This is Elsa Dominish. As a researcher for the Institute for Sustainable Futures in Sydney, Australia, she has a global perspective on the impact of mining and ideas about how to do it better.

20:13But she is quick to clarify that EVs are crucial to carbon reduction. So road transport is a huge emitter. It's 10 % of global emissions. And this sector's grown more than any other end-use sector. So getting those emissions down is essential. And EVs are one of the technologies to do that. And they're a very important one. EVs are a good thing. There's no question. And we need them as part of the transition. Here she is on the TED stage. But in many ways, isn't mining and overconsumption what got us into this problem? And now we're trying to mine and consume our way out of it? So why does this matter?

20:56And surely this mining isn't as bad as mining fossil fuels? Well, yes, we need to put a halt to mining fossil fuels immediately. But it's not about what type of mining is better or worse or more or less. All of it matters. Mining for the green transition has impacts, and that matters too. We're facing a climate crisis, but we're also facing a biodiversity crisis. And many of these minerals are found in sensitive and fragile ecosystems. Half of all these minerals are found on indigenous lands. If we are doing this transition in the name of saving the planet, it should not come at the cost of sacrificing communities and ecosystems.

21:36And here, in Australia, we are the number one lithium producer. But we by no means have a clean record in relation to mining. And we are a world leader when it comes to biodiversity loss. Lithium mining was even being considered under old growth forests in Western Australia until community protests brought this to a halt. How much of the country do we want to keep digging up when so few wild places remain? Here and across the globe, mining continues to happen on lands sacred to Indigenous people who never had the option to say no. Okay, so lithium, we've talked about lithium. What are the other ones, other mines that you see being most problematic that are vital to making an electric car possible?

22:23Yeah, so cobalt is definitely the most widely known in terms of its impacts. Cobalt mining, 60, 70 % of that is happening in the Democratic Republic of Congo. Here there's both industrial mining and also small-scale informal mining. This is one of the thousands of unregulated, unmonitored mines in the DRC. Sky News film children as young as four working in appalling conditions. People are digging hand-dug tunnels, which are extremely dangerous. They extend tens of metres underground. They're obviously at risk of collapsing, of flooding. And there are many kind of deaths that have gone unreported from these mines.

23:08But the other big issue is industrial mines. I mean, I have seen videos from these mines. It's thousands of people, shoulder to shoulder, digging with their hands, simple tools, hauling out minerals with buckets. People refer to it as modern-day slavery. Exactly. So, you know, a lot of these are foreign-owned. Workers here are exploited. They're facing racism and violence and the conditions are still unsafe. They're also often paid as subcontractors, earning, you know, as little as$2.50 a day, which is below the living wage. And so they're working, but they're continuing to be trapped in poverty.

23:53Another mineral, nickel. Over half of the world's nickel is found in Indonesia and recently $14 billion of international investment has poured in for new mines and smelters. Contaminated soil from the mines has turned the ocean red and brown. Land and crops have been destroyed and sometimes the air is so polluted that it's difficult to breathe. If we continue on the pathway that we are, we could need 40 times more nickel by 2040. Is that in any way feasible or sustainable? And even the remotest parts of the ocean are under threat from plans to mine the deep sea. Communities in the Pacific are leading a protest against this, worried about the impacts it could have.

24:33We don't even understand the deep sea to know what we could potentially be destroying.

24:41Okay, so we've kind of laid out the conundrum here, which is we do need to get electric cars on the road. We have problems with how we are sourcing the materials we need to build those electric cars. So what do we do? What are the solutions? I mean, without just, you know, shrugging our shoulders and saying, well, to play the long game, we have to make some sacrifices in the short term. Yeah, so from what we've looked at, the most important thing we can do is to reduce the amount of minerals we need. And to do this, the key thing is reducing our reliance on cars. As an American, I have to chuckle at that.

25:19Exactly. Because telling people to give up their cars here, I mean, not only is it like the American way, it's just impossible in a lot of places in the United States. Yes, exactly. And I hear that a lot. I'm talking about particularly in cities. Our transport system is already going to have to change, so why don't we use this as an opportunity to redesign it to one that actually meets our needs, instead of locking ourselves into a future based on cars and all the existing problems that we know they create. We need fewer cars, but we also need smaller cars. The bigger the car, the bigger the battery, and the more mining that is required.

25:56A small EV might have a battery that weighs 300 kilograms. A big one could be two or three times that. there are some companies in the US that have even stopped manufacturing small EVs in order to make more large SUVs and trucks. It's not because people don't want to buy small EVs, they do, but it's because the larger cars are more profitable. We also need to be designing them so that they can be repaired. Whole cars are being scrapped because of a problem with the battery. How is that sustainable? Swapping one car for another? It's not going to solve our problems. Instead, we need to think differently, think bigger, and change the whole system.

26:35If we make our cities easier to get around without needing to drive, if we make public transport convenient and safe and affordable and accessible, people will use it. Okay, so public transport, smaller cars. What about recycling? Does that play any part here? So there's quite a few companies that are partnering manufacturers with battery manufacturers and recyclers in order to create closed-loop systems where batteries from end-of-life vehicles are collected, they're recycled in a really high-value process, and then those minerals can go straight back into battery manufacturing. Changing the entire transportation ecosystem is the long-term goal.

27:20But we'll still need more minerals than ever before for EVs. So right now, people are working towards making mining possible, but changing how and where it gets done. My organization has a half dozen people who all day every day they fight for electric vehicles and battery storage for the renewable energy transition. So we're all in. And that means implicitly we're all in on lithium mining. Patrick Donnelly heads the Nevada branch of the Center for Biological Diversity. He's trying to stop another proposed lithium mine from opening on the site of endangered wildflower. They chose to site it in the middle of endangered species habitat, which they knew was there.

28:02And so people say, well, we don't have time for a planning effort. No, we don't have time for all these lithium companies to be messing around in court for a decade. So we really need leadership in order to get a plan in place for how we're going to produce all this lithium without ruining the American West and driving species extinct and harming indigenous communities. Patrick points to an example in California as a model for how all sides, the government, industry, and conservationists, can negotiate and find solutions that work for everyone. This was back in 2014, and the issue was solar, but the debate was similar.