4 ms·

are you sure we aren't on the >4°C path? AI could very well put us back on it.

brynnbee 8mo ago
Considering that current energy usage of AI is very likely to plummet in the coming years due to efficiency gains (which we've seem massive improvements on in the last couple years) and most of these large companies building datacenters are looking for and investing in specifically clean energy sources (nuclear) I don't think AI is a meaningful contributor compared to all the other high causes of global emissions.
- judahmeek 8mo ago
  Jevon's Paradox would like a word
  - brynnbee 8mo ago
    Sure, but I meant consumption on a per-task basis. I would also add that at some point building new software is going to become rather valueless as society trends towards building effectively any variant of software anyone could want relatively instantly and cheaply. There is also a finite number of consumers of AI content on the planet - sure maybe everyone want 4k-per eye live feed video of anything they want 24/7, but once you accomplish that for all 8 billion humans, there is no further demand for it. Generation costs are very likely to continue to trends rapidly downward and there is a likely final wall of saturation of demand.
baq 8mo ago
AI which helps us design a commerciallizable fusion reactor might buy us a century.
- ben_w 8mo ago
  I don't expect fusion to make a difference in either direction. For big reactors: At current aluminium prices, the bill of materials for a global power grid with 1 Ω resistance the long way round is only about 10x the cost ITER (the organisation) expects ITER (the reactor) to be. Plug in your own numbers for target resistance as desired, halve resistance needs double the material. With such a grid, you can put the PV in Angola and still get useful output in mid-winter nighttime in Anchorage. For potential small reactors like Helion's "shipping container" target size, I won't say it can't work (I don't know enough to be confident), but I will say that we immediately find we have bigger problems because any hostile actor can simply choose to run them in neutron-source mode and turn everyday cheap depleted uranium* into weaponizable plutonium. * I note that eBay still clearly has a dictionary merge on all nouns, given my search results came up with this: Get the best deals for Depleted Uranium Metal at eBay.com. We have a great online selection at the lowest prices with Fast & Free shipping on many items! Also be aware that US restrictions on sales aren't particularly relevant to this, as the moment this Columbus' Egg** gets solved it rapidly becomes a global problem. ** https://en.wikipedia.org/wiki/Egg_of_Columbus https://en.wikipedia.org/wiki/Egg_of_Columbus
  - myrmidon 8mo ago
    Completely agree with the outlook ("fusion power irrelevant for climate change in every realistic scenario"). But where did you take those grid cost numbers from? Iter costs are <100bn AFAIC; and Germany alone (!!) projects more than that (top end) for grid expansion/operation within 2040 (mainly north/south and offshore connectivity).
    - ben_w 8mo ago
      Putting this note first, because it's probably the main point of confusion/surprise: I did say "bill of materials"; the estimated full cost for European and US grid upgrades that we need anyway for other reasons, with far less material, is order-of a trillion or so for US, half a trillion for EU. For the material cost, just applied maths. By sheer coincidence, 1Ω of aluminium around the world is very close to 1m^2 cross section: https://www.wolframalpha.com/input?i=40000km+*+resistance+aluminium+%2F+1m%5E2 https://www.wolframalpha.com/input?i=40000km+*+resistance+al... This is almost exactly 1e8 (100 million) tons: https://www.wolframalpha.com/input?i=40000km+*+1m%5E2+*+density+aluminium https://www.wolframalpha.com/input?i=40000km+*+1m%5E2+*+dens... This is $223bn at current prices: https://www.wolframalpha.com/input?i=40000km+*+1m%5E2+*+density+aluminium+*+price+aluminium https://www.wolframalpha.com/input?i=40000km+*+1m%5E2+*+dens... For scale, this is about what China makes in 2 years; if this is rolled out over 30, which would be optimistic but plausible, it's within the realm of just how much China increased production between 2023 and 2025, being spent every year. To get to 10x ITER's own estimate for ITER, the wikipedia page says the organisation estimates the reactor will cost about €18-22bn: https://en.wikipedia.org/wiki/ITER https://en.wikipedia.org/wiki/ITER There are a lot of reasons not to do this as a single big 1m^2 "wire", amongst them being that the surface magnetic field is strong enough to be dangerous to approach with ferrous materials.
      - myrmidon 8mo ago
        The "square meter of aluminium" is an interesting take. Not sure how much power you'd get over that thing; extrapolating from existing HVDC systems (Inga-Shaba is 1GW over 2x520mm²), I'd expect around 1TW, so twice the US demand? But because Nimbys have no appreciation for beautiful pylons, projects in that direction are doomed for now anyway and everything needs to be buried underground at extra cost :(

Considering that current energy usage of AI is very likely to plummet in the coming years due to efficiency gains (which we've seem massive improvements on in the last couple years) and most of these large companies building datacenters are looking for and investing in specifically clean energy sources (nuclear) I don't think AI is a meaningful contributor compared to all the other high causes of global emissions.

judahmeek 8mo ago
Jevon's Paradox would like a word
- brynnbee 8mo ago
  Sure, but I meant consumption on a per-task basis. I would also add that at some point building new software is going to become rather valueless as society trends towards building effectively any variant of software anyone could want relatively instantly and cheaply. There is also a finite number of consumers of AI content on the planet - sure maybe everyone want 4k-per eye live feed video of anything they want 24/7, but once you accomplish that for all 8 billion humans, there is no further demand for it. Generation costs are very likely to continue to trends rapidly downward and there is a likely final wall of saturation of demand.

baq 8mo ago

AI which helps us design a commerciallizable fusion reactor might buy us a century.

ben_w 8mo ago
I don't expect fusion to make a difference in either direction. For big reactors: At current aluminium prices, the bill of materials for a global power grid with 1 Ω resistance the long way round is only about 10x the cost ITER (the organisation) expects ITER (the reactor) to be. Plug in your own numbers for target resistance as desired, halve resistance needs double the material. With such a grid, you can put the PV in Angola and still get useful output in mid-winter nighttime in Anchorage. For potential small reactors like Helion's "shipping container" target size, I won't say it can't work (I don't know enough to be confident), but I will say that we immediately find we have bigger problems because any hostile actor can simply choose to run them in neutron-source mode and turn everyday cheap depleted uranium* into weaponizable plutonium. * I note that eBay still clearly has a dictionary merge on all nouns, given my search results came up with this: Get the best deals for Depleted Uranium Metal at eBay.com. We have a great online selection at the lowest prices with Fast & Free shipping on many items! Also be aware that US restrictions on sales aren't particularly relevant to this, as the moment this Columbus' Egg** gets solved it rapidly becomes a global problem. ** https://en.wikipedia.org/wiki/Egg_of_Columbus https://en.wikipedia.org/wiki/Egg_of_Columbus
- myrmidon 8mo ago
  Completely agree with the outlook ("fusion power irrelevant for climate change in every realistic scenario"). But where did you take those grid cost numbers from? Iter costs are <100bn AFAIC; and Germany alone (!!) projects more than that (top end) for grid expansion/operation within 2040 (mainly north/south and offshore connectivity).
  - ben_w 8mo ago
    Putting this note first, because it's probably the main point of confusion/surprise: I did say "bill of materials"; the estimated full cost for European and US grid upgrades that we need anyway for other reasons, with far less material, is order-of a trillion or so for US, half a trillion for EU. For the material cost, just applied maths. By sheer coincidence, 1Ω of aluminium around the world is very close to 1m^2 cross section: https://www.wolframalpha.com/input?i=40000km+*+resistance+aluminium+%2F+1m%5E2 https://www.wolframalpha.com/input?i=40000km+*+resistance+al... This is almost exactly 1e8 (100 million) tons: https://www.wolframalpha.com/input?i=40000km+*+1m%5E2+*+density+aluminium https://www.wolframalpha.com/input?i=40000km+*+1m%5E2+*+dens... This is $223bn at current prices: https://www.wolframalpha.com/input?i=40000km+*+1m%5E2+*+density+aluminium+*+price+aluminium https://www.wolframalpha.com/input?i=40000km+*+1m%5E2+*+dens... For scale, this is about what China makes in 2 years; if this is rolled out over 30, which would be optimistic but plausible, it's within the realm of just how much China increased production between 2023 and 2025, being spent every year. To get to 10x ITER's own estimate for ITER, the wikipedia page says the organisation estimates the reactor will cost about €18-22bn: https://en.wikipedia.org/wiki/ITER https://en.wikipedia.org/wiki/ITER There are a lot of reasons not to do this as a single big 1m^2 "wire", amongst them being that the surface magnetic field is strong enough to be dangerous to approach with ferrous materials.
    - myrmidon 8mo ago
      The "square meter of aluminium" is an interesting take. Not sure how much power you'd get over that thing; extrapolating from existing HVDC systems (Inga-Shaba is 1GW over 2x520mm²), I'd expect around 1TW, so twice the US demand? But because Nimbys have no appreciation for beautiful pylons, projects in that direction are doomed for now anyway and everything needs to be buried underground at extra cost :(