welcome to ai decoded, that time of the week when we look in depth at some of the most eye—catching stories in the world of artificial intelligence. this week, we look at how big tech is searching lam your i am your naive inquisitor. i know nothing about it, so you have got good company if you don't either. this week, we look at how big tech is searching for new sources of energy to power the latest generative artificial intelligence models. it's been estimated the current ai boom could see an increase in demand for electricity soar by 40% in the next decade. to solve this issue, amazon, google and microsoft are betting billions of dollars that nuclear power can help provide their ai energy needs. they revealed plans to buy electricity from nuclear power plants. but what impact could this have on the planet? tech informed says prioritising ai over climate change could be catastrophic. recent research suggests that global ai in data centres could need as much electricity

as a small country such as the netherlands or sweden by 2027. so could the answer be nuclear power? tech giants google and amazon have both recently unveiled plans to power their ai needs using "advanced" nuclear energy, as part of their efforts to become carbon—neutral. meanwhile, bloomberg says some of the world's biggest managers of private capital are also making moves to finance and invest in nuclear power plants, latching onto a revival in the sector driven by the tech giants' electricity—intensive ai ambitions. with us is our regular ai commentator and contributor stephanie hare. stephanie, i know that bitcoin mining needs to demand of energy, what sort of levels of electricity are we talking about for al?— electricity are we talking about for al ? about for al? we're talking massive. — about for al? we're talking massive, as _ about for al? we're talking massive, as in there's - about for al? we're talking massive, as in there's notl about for al? we're talking | massive, as in there's not a single power grid anywhere on the planet that is ready to go where we need to go with artificial intelligence, so we

have to retool completely the grid and that is why you're looking and notjust renewables like solar and wind but nuclear, which is something that companies were not talking about back in the spring, when microsoft and google announced that they were not going to meet the net zero emissions commitments. let's bring in dr sasha luccioni, who's the ai and climate lead at hugging face. first of all, tell us what hugging face is. it first of all, tell us what hugging face is.- first of all, tell us what hugging face is. it is a host for models _ hugging face is. it is a host for models and _ hugging face is. it is a host for models and data - hugging face is. it is a host for models and data sets i hugging face is. it is a host. for models and data sets and who — for models and data sets and who has_ for models and data sets and who has become a hub for sharing _ who has become a hub for sharing your models and find the models you need, and we really— the models you need, and we really espouse aai responsible approach. we really espouse aai responsible auroach. ~ ., �* ., really espouse aai responsible auroach. ~ ., �* . ., approach. we don't want to break things _ approach. we don't want to break things that _ approach. we don't want to break things that we - approach. we don't want to break things that we are . break things that we are talking a nuclear power, do we come at all? ai could, some people say, help solve the climate change crisis, sasha, but it is not going to do that if it is sucking up the electricity itself. exactly. what we're _

electricity itself. exactly. what we're seeing - electricity itself. exactly. what we're seeing is - electricity itself. exactly. what we're seeing is the | what we're seeing is the climate _ what we're seeing is the climate positive applications are really efficient with a rongier laptops, sometimes on a computen — rongier laptops, sometimes on a computer. it is the largest linkage _ computer. it is the largest linkage models, the jackpots that are _ linkage models, the jackpots that are gobbling up a lot of electricity in their being deployed in summary applications and that is really adding — applications and that is really adding up. applications and that is really adding um— applications and that is really adding un— applications and that is really addin: u.~ ., , adding up. when we hear people like eric schmidt _ adding up. when we hear people like eric schmidt of _ adding up. when we hear people like eric schmidt of google - adding up. when we hear people like eric schmidt of google say i like eric schmidt of google say that al is going to solve the climate . , , , ., , climate crisis, is there any everett to _ climate crisis, is there any everett to that _ climate crisis, is there any everett to that statement | climate crisis, is there any i everett to that statement at all or is thatjust... everett to that statement at all or is that just. . .- all or is that 'ust. .. climate chance all or is that 'ust. .. climate change is _ all or is that just. .. climate change is not _ all or is that just. .. climate change is not a _ all or is that just. .. climate change is not a single - all or is thatjust. .. climate i change is not a single problem so it— change is not a single problem so it does— change is not a single problem so it does not have a single solution, _ so it does not have a single solution, and it is true there are myriads of problems, from climate — are myriads of problems, from climate prediction to actually d carbonised in our lecture secrets, _ d carbonised in our lecture secrets, that have to be solved, _ secrets, that have to be solved, but ai is always only one — solved, but ai is always only one part _ solved, but ai is always only one part of the solution we really — one part of the solution we really have to apply it mindfully, and especially with domain — mindfully, and especially with domain experts. you cannot chat buy your — domain experts. you cannot chat buy your way out of climate change _ buy your way out of climate change. is buy your way out of climate chance. , ., , ., change. is there any need to slow down — change. is there any need to slow down on _ change. is there any need to slow down on rolling - change. is there any need to slow down on rolling a - slow down on rolling a generative ai slow down on rolling a generative aland

—— into every business on the planet with these electricity needs we literally cannot sustain without going nuclear? currently we're putting the horse _ currently we're putting the horse before the cart, because we're — horse before the cart, because we're focusing on getting more energy— we're focusing on getting more energy instead of thinking about— energy instead of thinking about why we need the energy in the first— about why we need the energy in the first place, and how, whether— the first place, and how, whether we can reduce the amount— whether we can reduce the amount of energy that al needs and apply it a little bit more mindfully. i'm increasingly hearing _ mindfully. i'm increasingly hearing the term digital sobriety, and i really like that, _ sobriety, and i really like that, so— sobriety, and i really like that, so thinking about whether we really— that, so thinking about whether we really need to generative ai, we really need to generative al, every— we really need to generative ai, every thing from navigation, do we really need to talk— navigation, do we really need to talk to _ navigation, do we really need to talk to a chat bot or can we do something, a simple google search, — do something, a simple google search, or— do something, a simple google search, or do any generative summaries, right? currently we are in— summaries, right? currently we are in this— summaries, right? currently we are in this ai summer, but hopefully an ai fall will come and have us contemplate a little — and have us contemplate a little bit _ and have us contemplate a little bit on the applications of al — little bit on the applications of al in— little bit on the applications of ai in different fields. i of ai in different fields. have of ai in different fields. i have never heard that expression, digitalsobriety, i do rather like it. we did not used to pay for plastic bags when we went shopping, and suddenly we have to pay for it, you start to think, do i really

need it? iwas you start to think, do i really need it? i was reading earlier one chatgpt query needs the equivalent electricity of powering a light bulb for 20 minutes. should companies have to declare how much electricity is being required so we asked consumers can make a choice? i definitely think they should, nowadays people are considering the climate in their day—to—day actions — the climate in their day—to—day actions and their diets and their— actions and their diets and their transportation, and digital— their transportation, and digital technologies like ai are increasingly parts of our lives — are increasingly parts of our lives and _ are increasingly parts of our lives and we need to this information in order to choose. i information in order to choose. i don't — information in order to choose. i don't like _ information in order to choose. i don't like is being a bill to idon't like is being a bill to answer— i don't like is being a bill to answer the question, how much more _ answer the question, how much more electricity does the new generative ai search use compared to the good old—fashioned one from a year or so— old—fashioned one from a year or so ago? _ old—fashioned one from a year or so ago? we really need opting _ or so ago? we really need opting out mechanisms and more transparency about the energy and the — transparency about the energy and the carbon because of all of this — and the carbon because of all of this. , ., ., ., of this. sasha, one of the questions _ of this. sasha, one of the questions i'm _ of this. sasha, one of the questions i'm cures - of this. sasha, one of the questions i'm cures to - of this. sasha, one of the i questions i'm cures to your thoughts on is about the ella mai competition. it

thoughts on is about the ella mai competition.— thoughts on is about the ella mai competition. it is all very aood mai competition. it is all very good -- _ mai competition. it is all very good -- competition - mai competition. it is all very good -- competition in - mai competition. it is all very good -- competition in the i good —— competition in the market. it is all good on amazon to buy a small reactor for themselves, amazon to buy a small reactor forthemselves, but amazon to buy a small reactor for themselves, but most company cannot do that, so could we see regulators getting in on this and say, somehow we have to divvy up the resources now? ., ~ , now? yeah, i think definitely in terms of— now? yeah, i think definitely in terms of nuclear, - now? yeah, i think definitely in terms of nuclear, it's - in terms of nuclear, it's already— in terms of nuclear, it's already a _ in terms of nuclear, it's already a technology that has had its — already a technology that has had its rooms and its busts as well, — had its rooms and its busts as well, and _ had its rooms and its busts as well, and for example in the united — well, and for example in the united states you can have, you can privatise small and medium nuclear— can privatise small and medium nuclear reactors. in many countries _ nuclear reactors. in many countries that's not the case, and _ countries that's not the case, and it— countries that's not the case, and it is— countries that's not the case, and it is federal or regional, its government owned, and so there — its government owned, and so there is— its government owned, and so there is not a single solution and — there is not a single solution and we — there is not a single solution and we are going to have to see some _ and we are going to have to see some kind — and we are going to have to see some kind of legislation about, when _ some kind of legislation about, when you're making trade—offs between — when you're making trade—offs between building a new data centre — between building a new data centre or using that energy to power— centre or using that energy to power homes or to build new housing _ power homes or to build new housing or new transportation, because — housing or new transportation, because now more and more we are going — because now more and more we are going to have to electrify things. — are going to have to electrify things, have more robust energy grid. _ things, have more robust energy grid, and — things, have more robust energy grid, and if— things, have more robust energy grid, and if that energy is going _ grid, and if that energy is going towards ai grid, and if that energy is going towards aland not towards d carbonised energy,

we've — towards d carbonised energy, we've got a problem. and what burden is be — we've got a problem. and what burden is be putting _ we've got a problem. and what burden is be putting on - we've got a problem. and what burden is be putting on all of i burden is be putting on all of the businesses around the world who are wanting to use ai in their processes but also have their processes but also have their own climate and sustainability commitments? how are they going to be it was to answer to how much more electricity they are using, water they are using, if there ai providers don't give the transparency were talking about? i transparency were talking about? . , ~' . about? i actually think that companies _ about? i actually think that companies are _ about? i actually think that companies are the - about? i actually think that companies are the key - about? i actually think that companies are the key to l companies are the key to transparency, because especially if you're using a google _ especially if you're using a google or microsoft, amazon product — google or microsoft, amazon product and you have hundreds of thousands of dollars of revenue, you can ask some questions, you can ask questions, you can ask questions about, where is the energy— questions about, where is the energy coming from, how is it being — energy coming from, how is it being generated? is it low carbon. _ being generated? is it low carbon, is there it would take a low— carbon, is there it would take a low carbon? you can ask these questions. — a low carbon? you can ask these questions, because people come to be _ questions, because people come to be all— questions, because people come to be all the time asking, how do i _ to be all the time asking, how do i get — to be all the time asking, how do i get a _ to be all the time asking, how do i get a handle on this? the answer— do i get a handle on this? the answer is _ do i get a handle on this? the answer is if— do i get a handle on this? the answer is if you're using these commercial services you have to ask for— commercial services you have to ask for transparency on the other— ask for transparency on the other side, and you have to add

that line. — other side, and you have to add that line, because chatgpt is not this— that line, because chatgpt is not this magic ephemeral entity. _ not this magic ephemeral entity, it is turning on a data centre — entity, it is turning on a data centre somewhere and that data centres — centre somewhere and that data centres using energy and water for cooling and natural resources, so we need as customers, as resources, so we need as customers, as coi'isuttiei's, customers, as consumers, information about all that. i am staggered that these companies are in such a position that they can be commissioning nuclear power in this way. i was also reading amazon is already the top corporate buyer of solar and wind energy. is that scalable, that green energy source? it is scalable that _ that green energy source? it is scalable that we _ that green energy source? it is scalable that we are _ that green energy source? it is scalable that we are going to have — scalable that we are going to have to — scalable that we are going to have to see more power purchase agreements, which are essentially these ious or these promises that when a solar plant — promises that when a solar plant will be built that that energy _ plant will be built that that energy will be bought by somebody and the big tech companies, google, microsoft, amazon. — companies, google, microsoft, amazon, have been signing these agreements because it is a big risk and — agreements because it is a big risk and a _ agreements because it is a big risk and a big investment to build — risk and a big investment to build these renewable power sources— build these renewable power sources and knowing that someone is going to buy that

energy. — someone is going to buy that energy, if i have —— five to ten — energy, if i have —— five to ten to— energy, if i have —— five to ten to eight years is reassuring. also, nuclear is seen — reassuring. also, nuclear is seen as— reassuring. also, nuclear is seen as the bridge solution because _ seen as the bridge solution because it takes less time for example _ because it takes less time for example to bring three mile island — example to bring three mile island back into business, back into production, then to build out a — into production, then to build out a whole new solar power plant — out a whole new solar power lant. ., , ., , plant. one of the things i was thinkin: plant. one of the things i was thinking about _ plant. one of the things i was thinking about is, _ plant. one of the things i was thinking about is, again, - thinking about is, again, the international competition angle. the big comedies we have just listed building their own reactors or buying it in our all american. reactors or buying it in our allamerican. so reactors or buying it in our all american. so what does this mean for everyone in the rest of the world? we saw germany a couple of years ago deciding to move away from energy, does this see them needing to move back towards nuclear again? it was in 2012 the fukushima meltdown injapan, angela meltdown in japan, angela merkel meltdown injapan, angela merkel famously did the u—turn and germany backed away from nuclear. could we see ai bring countries back to nuclear who had been wanting to move away from it? ., , , ., from it? hopefully, no, hopefully _ from it? hopefully, no, hopefully countries - from it? hopefully, no, hopefully countries will| from it? hopefully, no, i hopefully countries will be going _ hopefully countries will be going more into renewables as

much — going more into renewables as much as— going more into renewables as much as possible, but i think the reason why it is all starting in the united states is because the legislations is lot more _ is because the legislations is lot more personal missive. it is not — lot more personal missive. it is not to— lot more personal missive. it is not to centralise power grid _ is not to centralise power grid the _ is not to centralise power grid. the us has dozens of independent energy providers, so it— independent energy providers, so it is— independent energy providers, so it is relatively easy. it is easier— so it is relatively easy. it is easier than in so it is relatively easy. it is easierthan in a so it is relatively easy. it is easier than in a place in germany, canada, the uk, to actually— germany, canada, the uk, to actually buy up some terawatt hours — actually buy up some terawatt hours of — actually buy up some terawatt hours of electricity, but hopefully this is not going to set the — hopefully this is not going to set the precedent forjust powering up nuclear power plants. _ powering up nuclear power plants, but on 1526 00:10:32,116 --> 00: