61–70 of 168 posts
Ugh, this is horrible writing. For instance, "Data centres cause pollution through high eletricity use" - high relative to what, residential housing? Aluminum smelting? It's fine to be skeptical of the AI bubble, but this article (ironic for FT) reads like veiled de-development. We also don't need any fizzy-drinks - shall we calculate the deaths due to pollution caused by that, let alone the primary effect on health…
It’s deflection from the real polluters
This reads as less than useful research to me. E.g.: "The analysis does not account for the purchase of market-based instruments that are meant to represent investments in new renewable energy in the US and that tech companies buy to offset the pollution from their electricity consumption. ... 'Unlike carbon emissions, the health impacts caused by a data centre in one region cannot be offset by cleaner air elsewhere,…
> Why not? If your house had a carbon monoxide leak but your neighbor’s house did not, would you consider your carbon monoxide leak to be “offset” by that fact?
Not sure why we need an analogy, though. How about sticking to standard air pollution, which has direct impact of its own? CO poisoning seems a bit extreme as a comparison.
This reads as less than useful research to me. E.g.: "The analysis does not account for the purchase of market-based instruments that are meant to represent investments in new renewable energy in the US and that tech companies buy to offset the pollution from their electricity consumption. ... 'Unlike carbon emissions, the health impacts caused by a data centre in one region cannot be offset by cleaner air elsewhere,…
Likewise if you pay someone to plant a tree that'll reabsorb the carbon you've emitted over the course of 40 years, and they cut it down and burn it after 15 years - the carbon's still in the atmosphere.
So putting your data centre right next to a hydroelectric or nuclear power plant is the gold standard.
[1] https://www.theguardian.com/environment/2023/jan/18/revealed...
Earlier quoted context omitted.
> Why not? If your house had a carbon monoxide leak but your neighbor’s house did not, would you consider your carbon monoxide leak to be “offset” by that fact?
If a mitigating action had decreased CO in my neighbor's house by at least as much as it increased in mine, then potentially yes, a neutral researcher should consider that as an offset at the population health level. Not sure why we need an analogy, though. How about sticking to standard air pollution, which has direct impact of its own? CO poisoning seems a bit extreme as a comparison.
Earlier quoted context omitted.
Even if you account for chernobyl and fukushima the death count from nuclear is orders of magnitudes less than the death count from coal, oil, solar, etc. We overregulated nuclear because the risk of nuclear disasters feels scarier than the real risk of deaths downstream of airborne particulates. There are tradeoffs to every safety regulation on nuclear and those tradeoffs include more climate induced disasters, a si…
>nuclear disasters feels scarier I am in favor of more nuclear power and I know very little about the field, but this argument never resonated with me. If the Chernobyl series paints a reasonably realistic image of what could have happened had they not contained the melt down, we are talking "better part of a continent being uninhabitable for thousands of years". When that is the potential risk, comparing death count…
It's not a realistic risk.
Chernobyl was absolute worse case. A massive reactor, with no containment structure (it was essentially in a shed), with a flammable solid state moderator, and no passive cooling. Yet, Europe is not uninhabitable today, infact the above-natural-background radiation in Europe is almost entirely coming from burning fossil fuels, not from Chernobyl.
Earlier quoted context omitted.
> Why not? If your house had a carbon monoxide leak but your neighbor’s house did not, would you consider your carbon monoxide leak to be “offset” by that fact?
If a mitigating action had decreased CO in my neighbor's house by at least as much as it increased in mine, then potentially yes, a neutral researcher should consider that as an offset at the population health level. Not sure why we need an analogy, though. How about sticking to standard air pollution, which has direct impact of its own? CO poisoning seems a bit extreme as a comparison.
Earlier quoted context omitted.
Nuclear is not economically viable compared to renewables (wind & solar), and it's rightfully on its way out. What is really needed is for humanity to say "enough" and stop using fossil fuels. Would it be a problem if training AI models (or running any other analytics) be lot more expensive in winter than summer? I doubt it. There is no reason not to use more renewables for this use case, then.
Nuclear has a far lower LCOE than renewables, when you mass produce them rather than do ad-hoc bespoke builds. For example, the Gen II+ plants built in the 90's had an (inflation adjusted) price of about $500 million per GWe. Whereas, the cost of the recent Vogtl build was $15 billion per GWe! Renewables proponents point to the cost of Vogtl (or similar) when they say that Nuclear has a higher cost than renewables, w…
Why not? Just have the excess of computational capacity in many different regions and compute where the sun is shining bright now.
Since every activity has some externalities and also unexpected positive effects, isn’t focusing on just the negative effects of this one thing very lopsided?
Earlier quoted context omitted.
Nuclear has a far lower LCOE than renewables, when you mass produce them rather than do ad-hoc bespoke builds. For example, the Gen II+ plants built in the 90's had an (inflation adjusted) price of about $500 million per GWe. Whereas, the cost of the recent Vogtl build was $15 billion per GWe! Renewables proponents point to the cost of Vogtl (or similar) when they say that Nuclear has a higher cost than renewables, w…
>The other part of this is that expensive silicon is not going to sit idle. It needs baseload power Why not? Just have the excess of computational capacity in many different regions and compute where the sun is shining bright now.
It's cheaper to build a 1 GW baseload power plant to sit beside your existing GPUs, than to buy another 1 GW of GPUs to place on the other side of the planet. Also, latency.
Also, a large scale roll out of nuclear would be cheaper than solar regardless. Most of the cost of recent nuclear builds is learning cost. We build them so infrequently that people retire or move on before the next one is built, so we have to learn how to build them from scratch, and develop the supply chains from scratch each time.