Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
AI Google Earth

Former Google Chief Urges AI Investment Over Climate Targets (windowscentral.com) 81

Posted by msmash from the how-about-that dept.
Former Google CEO Eric Schmidt urged prioritizing AI infrastructure over climate goals at a Washington AI summit this week. Schmidt, who led Google until 2011, argued that AI's rapid growth will outpace environmental mitigation efforts. "We're not going to hit the climate goals anyway because we're not organized to do it," Schmidt told attendees, addressing concerns about AI's surging energy demands.

Data centers powering AI are projected to consume 35 gigawatts annually by 2030, up from 17 gigawatts in 2023, according to McKinsey. Schmidt, now heading AI drone company White Stork, suggested AI could ultimately solve climate issues, stating, "I'd rather bet on AI solving the problem than constraining it."
This discussion has been archived. No new comments can be posted.

Former Google Chief Urges AI Investment Over Climate Targets More

Former Google Chief Urges AI Investment Over Climate Targets

Comments Filter:

  • Make 'em all nuclear (Score:5, Insightful)

    by bettodavis ( 1782302 ) on Friday October 11, 2024 @11:46AM (#64857051)
    I'm serious. All big AI cluster investment ought to have a nuclear plant in it, ensuring it gets properly juiced.

    And once they're there, lobby to grow nuclear everywhere, so we stop the trend of fake de-carbonization through repentance and self-impoverishment.

    • Re: (Score:1, Redundant)

      by ShanghaiBill ( 739463 )

      Building a bespoke generator for each consumer is dumb. We have a grid for a reason.

      Nuclear has a twenty-year lead time.

      The last nukes built had massive cost overruns, and there is no reason to believe next time will be different.

      • Re:Make 'em all nuclear (Score:5, Insightful)

        by cayenne8 ( 626475 ) on Friday October 11, 2024 @12:23PM (#64857135) Homepage Journal

        Nuclear has a twenty-year lead time.

        Hey, you gotta start somewhere/sometime, eh?

        The last nukes built had massive cost overruns, and there is no reason to believe next time will be different.

        I am by NO measure any sort of expert on nuke tech, but, I do seem to be hearing that in all these years, there have been major strides in the tech...and how to build multiple smaller nuke generators that are safer and faster to stand up?

        I believe we also need to review all the red tape for nukes....not to cut regs just to cut regs, these things need to be safe, but let's look into what's on the books that piled up over the years and make sure whatever doesn't make sense anymore and hinders progress, is cleared off the books.

        • Re: (Score:2, Insightful)

          by sfcat ( 872532 )
          First nuclear plant from design to electrons on the grid was completed in 18 months. Both the Chinese and Koreans deployed AP1400 from order to electrons on the grid in about 3 years. 20 years is only in the west and not due to technical reasons.

          • Re: (Score:3)

            by AmiMoJo ( 196126 )

            The first nuclear power plant was the APS-1 Obninsk, which was started in 1951 and connected to the grid in 1954. It produced 6MW, so was more of a prototype, and didn't have a containment building or really any other safety features.

            The AP1400 takes about a decade to build, but only at sites that already have reactors and all the other infrastructure. For example, the latest units at the Kori and Hanul plants in South Korea took a decade each. It takes longer if you are building a whole plant from scratch.

            • Re: (Score:2)

              by q_e_t ( 5104099 )
              EBR-1 took 18 months to build (1949-51), but wasn't grid connected and its output was kWs.

          • 20 years is only in the west and not due to technical reasons.

            Question: Is that going to change?

            Answer: No.

            Question: Is it possible to take 20 years to finish a nuke, paying interest the whole time, and do so profitably?

            Answer: No.

            Question: So, will the next nuke be yet another financial debacle, just like Vogtle and Summer?

            Answer: Yup.

          • Re: (Score:2)

            by q_e_t ( 5104099 )
            The first plant from design to grid didn't take eighteen months, it took just over three years from breaking ground, longer from design. This was https://en.m.wikipedia.org/wik... [wikipedia.org]. EBR-1 was not grid connected. EBR-1 took 18 months to construct, not design to operation. Since it only powered four light bulbs it isn't really comparable to any useful, current reactor. The typical time to build a reactor in South Korea is currently fifteen years, not three. In China it takes five. I'm actually in favour of som
        • It's all bullshit. Nuclear has to be safe for 20,000 years give or take. Anybody that says they have it handled is a liar, full stop.

          • Nuclear has to be safe for 20,000 years give or take.

            Define "safe".

            After 500 years, nuclear waste emits less radiation than the original ore. If that isn't good enough, then why not?

            Anybody that says they have it handled is a liar, full stop.

            We have it handled. Yucca Mountain is a reasonable solution that has been held up only because of politics.

            The problem with nukes isn't waste. It isn't safety. It's economics.

            • Re: (Score:2)

              by q_e_t ( 5104099 )

              Nuclear has to be safe for 20,000 years give or take.

              Define "safe".

              After 500 years, nuclear waste emits less radiation than the original ore. If that isn't good enough, then why not?

              Biological accessibility of heavy metals is also an issue.

        • Re: (Score:2)

          by AmiMoJo ( 196126 )

          The AI guys are not going to wait 20 years. The only hope of them not buggering up the climate even more is to force them to power their crap with renewable energy or by extending existing nuclear plant lifespans (if you dare).

          You have been misinformed about "major strides" in technology, it's just the same old stuff repackaged. The smaller reactors you mention are not commercially viable, as they consume more fuel, produce more waste, and still need most of the expense of large ones, like containment build

        • multiple smaller nuke generators that are safer and faster to stand up?

          Number of SMRs that have been successfully built: 0.

        • It's not the red tape that makes it so expensive, it's the construction delays caused by the constant injunctions filed against them that really drives up the costs. https://en.wikipedia.org/wiki/... [wikipedia.org] I used to live about 2 miles from this site about 20 years ago.

          • Re: (Score:2)

            by q_e_t ( 5104099 )
            Costs seem to spiral even without injunctions. Hinckley C had some legal opposition but not very much and that ended a long time ago, but it over budget. It's an endemic issue with large projects of any sort, and is only partly due to engineering as a large part of the issue is raising finance. Using SMRs would get around that particular issue, but they are less efficient, so are not likely to be a solution.

      • Regular* Data centers already consume enough that they need their own substations and grid infrastructure. It's not as easy as just plug it in the wall. A new data center requires months of planning with utility companies to make happen. AI and crypto data centers often use as much energy as whole-ass powerplants produce. It's an order of magnitude different.

        *you could now effectively call them "low-power" data centers after AI and Crypto started filling racks with GPUs instead of cpus, switches, and r

      • Some costs are from 'obstructionists' (Score:5, Informative)

        by david.emery ( 127135 ) on Friday October 11, 2024 @12:51PM (#64857217)

        From my observation of a couple nuke projects where our local power company was part of the investment back in the 80s and 90s: A lot of the source of overruns was the impact of delaying tactics and opposition by governments, "public interest" legal lobbies, and local opposition. That's not to say that nuke plants haven't had technical and construction problems. But it's important to recognize the cost impact created by groups seeking to block the construction. And of course, there are no penalties to governments or lobbyists when they cause these massive delays/overruns.

        • Re: (Score:3)

          by AmiMoJo ( 196126 )

          The UK's currently under construction Hinkley Point C reactors are at a site that already has other reactors, so most of the environmental and legal issues were sorted long ago and any lingering ones were quickly resolved.

          It is still taking 20 years to build, still massively expensive and over the original budget, and will still produce the most expensive electricity in the UK.

          It's not particular to the UK either. The builder and operator, EDF, is having the same problems in other European countries, and gi

        • Re:Some costs are from 'obstructionists' (Score:4, Insightful)

          by q_e_t ( 5104099 ) on Saturday October 12, 2024 @01:04AM (#64858577)
          In the UK, the government is very supportive of nuclear power, but there are still cost overruns. Ditto in other countries. The evidence that opposition by governments or environmental groups cause cost overruns is quite weak. Cost overruns in large, capital and engineering projects are par for the course - it isn't even specific to nuclear power. For example, in the UK, major rail projects also seem to have massive cost overruns and in general, environmental groups are quite supportive of rail, and successive governments supportive of the projects.

          • Every delay in licensing is a day of cost overrun, for the interest payments on the construction loans without any revenue coming in. That seems pretty obvious to me. And in Seabrook, the delays caused by the state of Massachusetts not submitting the required plans, etc, -as directed by the governor of Mass- (he was honest about that) stretched for several years.

            • Re: (Score:3)

              by q_e_t ( 5104099 )
              Whilst it is true that any licensing delay costs money, even where there are no such delays, cost overruns abound. In the UK, the licensing went pretty quickly, but things are still over budget. Delays aren't an obvious smoking gun for cost overruns. And as I noted, pretty much all large projects have cost overruns, often large, so nuclear isn't in any way unique.

      • Re: (Score:2)

        by PPH ( 736903 )

        We have a grid for a reason.

        To heat my Aunt Millie's apartment.

        In Washington State, it was common practice to site aluminum smelters just down the road from hydroelectric plants. The same logic should apply to data centers. Except that for many power sources, both their sites as well as the DCs are flexible.

        The last nukes built had massive cost overruns

        The revenue potential of AI should cover that easily.

    • Re:Make 'em all nuclear (Score:5, Funny)

      by MobyDisk ( 75490 ) on Friday October 11, 2024 @01:48PM (#64857377) Homepage

      Don't do that. Then the AIs will still have power even after we scorch the sky.

    • Have you ever seen "Colossus: The Forbin Project" (1970)???

      That shows what happens when you hook an AGI up to a nuclear reactor:

      Hint: It becomes an ASI in less than 24 hours and takes over the entire world, turning some of us into pets, and killing the rest...

      Fantastic movie!

  • The world uses 3,200 GW of electricity, so the projected consumption by AI is about one percent.

    Most AI data centers are in 1st world countries which are transitioning to renewables.

    • They burn 'renewables' here at our local power plant, wood chips. Isn't that worse pollution wise than coal? They pitch it as green though.

      • Is wood greener than coal? Re:35 GW is one percent (Score:5, Informative)

        by XXongo ( 3986865 ) on Friday October 11, 2024 @01:09PM (#64857271) Homepage

        They burn 'renewables' here at our local power plant, wood chips. Isn't that worse pollution wise than coal?

        Hard to say. If the only criterion were carbon dioxide emissions, growing wood for fuel and then burning it is net zero: the carbon dioxide emitted in burning equals the carbon dioxide captured in growing the wood. So, yes, it's a lot greener than coal in that respect.

        Burning creates other pollution, though. Wood doesn't contain any significant sulfur, so it's lower pollution that high- or medium-sulfur coal. Word burning does have other pollutants, though. And emissions will depend a lot on how it's burned-- how hot, how good the airflow is, whether the stack has scrubbers, etc.

        As a quick estimate, including carbon dioxide as "pollution," I'd say that if the combustion is up to modern standards, it's probably greener than coal, but it's not pollution free.

        • growing wood for fuel and then burning it is net zero

          You ignored the fuel consumed in growing, harvesting, chipping, drying, and transporting the wood.

          Most chips are grown in North America but burned in Europe. That's a long journey, all powered by fossil fuels.

          it's a lot greener than coal in that respect.

          Comparing wood chips to coal isn't reasonable since coal isn't the alternative.

          • Re: (Score:2)

            by XXongo ( 3986865 )

            growing wood for fuel and then burning it is net zero

            You ignored the fuel consumed in growing, harvesting, chipping, drying, and transporting the wood.

            Not in any way comparable to removing a mountaintop to dig out coal.

      • "They burn 'renewables' here at our local power plant, wood chips. Isn't that worse pollution wise than coal?"

        No.

        Coal contains fissile nuclear material and it is sequestered carbon.

      • They burn 'renewables' here at our local power plant, wood chips.

        I presume you're in the UK.

        The UK wood chip subsidy scheme is widely recognized as even stupider than America's corn ethanol subsidies.

        At least the corn subsidies go from Americans to Americans. The wood chips are mostly grown in North America and shipped across the Atlantic in bunker-fuel-burning cargo ships. British taxpayers are subsidizing North Americans.

  • Eric Schmidt (Score:3, Insightful)

    by Savage-Rabbit ( 308260 ) on Friday October 11, 2024 @11:53AM (#64857065)

    Eric Schmidt is a toxic sociopathic narcissist who is now fishing for taxpayer money for AI ventures. No amount of AI will create a world where we ‘solve’ climate change while still de-sequestering exponentially increasing quantities of fossil carbon. The only way to deal with climate change is to move on from this antiquated fossil carbon economy.

    • Eric Schmidt is a toxic sociopathic narcissist

      Sure. But that doesn't mean he's wrong.

      while still de-sequestering exponentially increasing quantities of fossil carbon

      Carbon emissions are growing, but the rate of growth is slowing, and fossil fuels are declining as a percentage of energy consumed. They certainly are not "exponentially increasing".

      • Re:Eric Schmidt (Score:5, Informative)

        by Savage-Rabbit ( 308260 ) on Friday October 11, 2024 @12:59PM (#64857241)

        Eric Schmidt is a toxic sociopathic narcissist

        Sure. But that doesn't mean he's wrong.

        Yes it does.

        while still de-sequestering exponentially increasing quantities of fossil carbon

        Carbon emissions are growing, but the rate of growth is slowing, and fossil fuels are declining as a percentage of energy consumed. They certainly are not "exponentially increasing".

        Really? Even if carbon emissions are 'slowing' they aren't doing it anywhere near fast enough. If we want to avoid a very bad situation we are going to have to more or less end fossil fuel use and the sooner the better, i.e. within two or three decades. Looking at this graph there's a slight dip because of Covid but other than that we seem to be on a path back to more or less exponential CO2 emissions growth: https://ourworldindata.org/co2... [ourworldindata.org] What Schmidt is essentially saying is that we should abandon all efforts at reducing emissions, continue to 'Drill baby drill!!' and bet the planet on the fact that the AI Bros will in some future utopia, somehow swoop in and miraculously 'solve climate change' like the the Avengers in a Marvel Comics movie.

        • Really? Even if carbon emissions are 'slowing' they aren't doing it anywhere near fast enough. If we want to avoid a very bad situation we are going to have to more or less end fossil fuel use and the sooner the better, i.e. within two or three decades. Looking at this graph there's a slight dip because of Covid but other than that we seem to be on a path back to more or less exponential CO2 emissions growth

          And that is why we need to try for nukes NOW....and also a "Plan B" for when it warms past the magic

          • Really? Even if carbon emissions are 'slowing' they aren't doing it anywhere near fast enough. If we want to avoid a very bad situation we are going to have to more or less end fossil fuel use and the sooner the better, i.e. within two or three decades. Looking at this graph there's a slight dip because of Covid but other than that we seem to be on a path back to more or less exponential CO2 emissions growth

            And that is why we need to try for nukes NOW....and also a "Plan B" for when it warms past the magical 1.5C mark, because it IS going to happen.

            There is just no stopping this....people around the world, in different countries with different needs, lifestyles and aspirations are not going to willingly shut down civilization at the same or worse levels that some of those places did when they locked down for Covid.

            It simply is not going to happen.

            So, taking that as fact...what else do you propose to do?

            People are NOT going to stop their day to day lives, stop using fuel, stop going to work, stop living a modern day live, not going to eat 'bugs' or whatever alternative to meat you think will halt the global temperature rise....

            The only way that's going to happen is if the whole world undergoes some dystopian transformation where there is no freedom, one govt to rule the world and dictate what formerly free people can do or live like.

            Frankly, I'd rather it get warmer....

            Really? If you are not a troll you are the only illegitimate love child of a snowflake and a drama queen that I have ever come across.

        • bet the planet on the fact that the AI Bros will in some future utopia, somehow swoop in and miraculously 'solve climate change' like the the Avengers in a Marvel Comics movie.

          There's no need to wait. AI can solve the climate crisis right now.

          Human: Dear AI, how do we solve climate change?
          AI: stop using fossil fuels lol
          Human: HOW DARE YOU

      • Re: (Score:2)

        by XXongo ( 3986865 )

        Eric Schmidt is a toxic sociopathic narcissist

        Sure. But that doesn't mean he's wrong.

        ROFL.

        But, actually, that's pretty much what his supporters say about Donald Trump.

      • Yes, he is wrong. "We won't make any goals, so fuck it, burn it down and let us oligarchs get rich while we can" is a bullshit response.
        Remember, he is the the asshat who said "Google isn't free, the cost is your private information".

        • Re: (Score:2)

          by sfcat ( 872532 )
          So if you don't like them then they are automatically wrong? That's not reasoning, that's a personality flaw you can see from orbit. Also, you are just plain wrong about nuclear. But then again, your opinions seem to be motivated by if you like the speaker, not the validity of their arguments. Are you sure you belong here?

    • Eric Schmidt peaked with lex.

    • "AI will solve all our problems with immigrants! Whether you are for it or against it, AI will do it your way!" --Eric Schmidt

  • Guy worth hundreds of millions says... (Score:1)

    by Anonymous Coward

    The man has earned enough to never have to work another day in his life and retire gracefully.

    At least Bill Gates had something of an afterlife as a philanthropist.

  • Priorities (Score:5, Funny)

    by medusa-v2 ( 3669719 ) on Friday October 11, 2024 @12:01PM (#64857083)
    Stop worrying about life for future generations. There are billionaires who need more money now!

  • We're not going to hit the climate goals, so... (Score:4, Funny)

    by doragasu ( 2717547 ) on Friday October 11, 2024 @12:07PM (#64857099)

    ... (checks notes) let's make things even worse.

  • Let it die, let it die, let it shrivel up and die.

    WTF is wrong with this guy?

    https://www.youtube.com/watch?... [youtube.com]

  • I think he is biased.
  • He just doesn't understand anyone who isn't a fellow plutocrat, or acknowledge anyone who he can't manipulate with his vast wealth. Why does every tech bro billionaire become such an uncaring sociopath? Is there a course? Sociopathy 101 maybe?

    • Why does every tech bro billionaire become such an uncaring sociopath?

      You have the cause and effect reversed.

  • But for a beautiful moment in time we created a lot of value for shareholders."

  • new data centers should run on new renewable power (Score:5, Insightful)

    by votsalo ( 5723036 ) on Friday October 11, 2024 @02:33PM (#64857499)
    Is it too much to ask that every new data center should be powered by new renewable power installation paid by the company that owns the data center?

    According to this 2019 Department of Energy article [energy.gov], a 240 Mwh lithium battery storage site costs $91 million (in 2019). Considering that data centers run around the clock, the cost per W is $91 /(240/24) = $9/W. Add another $1/W for the PV, and the cost is roughly $10/W.

    35GW of data centers would require investing $350 million for renewable power. They could then use this power for the next 2 decades for free. That seems fair and probably also profitable for the data centers.

    This could apply to every new electric installation, not just data centers. Are you building a new 1KW house? Pay $10K (once) for new renewable power generation. Or put PV panels on your roof. California already requires that all new homes have rooftop solar PV panels. Do the same for data centers (except not on their rooftop).

  • Without pressure to search for more efficient ways to do the same thing, often we won't try as hard. And we won't aim resources at the problem.

    Otherwise this is like wishing, or praying, or whatever your favorite unproven/unreliable solution may be. "If this works it'll be great!" then lets wait for it to work before we rely on it.

  • Of course they're power hungry in more ways than one. It's not a surprise. They'll just spend the rest of their lives in their billionaire prepper bunkers should the SHTF from them failing to lead and take action to turn around the climate catastrophe. In the meantime, they DGAF and are fine with ignoring the needs of humanity's long-term survival.
  • Why does everybody think that the future of AI lies in ever more gigantic datacenters? That really only makes sense as long as edge/local solutions aren't powerful enough. Oh btw., "projected to consume 35 gigawatts annually by 2030" is scientific nonsense. Someone didn't know the difference between energy and power.

Slashdot Top Deals

Two can Live as Cheaply as One for Half as Long. -- Howard Kandel

Close