Eventually we have to either give up on the hopes of what could come from LLMs with enough investment, or give up on very loud but apparently hollow arguments related to the damage we are causing to the planet.
Eventually we have to either give up on the hopes of what could come from LLMs with enough investment, or give up on very loud but apparently hollow arguments related to the damage we are causing to the planet.
The people driving AI investment will simply not be significantly affected by climate change. They don't care that hundreds of millions in the tropics will die, and that much of organised human activity will collapse, because up until the last possible moment they'll be insulated from the consequences.
Which 'hollow arguments' are you referring to?
I call them apparently hollow in this context because we can't both chase the resource behemoth that is LLM tech and make any meaningful change to reduce our impact.
Its a reasonable hope that we could discover a new energy source that can produce orders of magnitude more energy with even less impact than today's sources, but that is just a hope. In the meantime we would be committing ourselves to a new, much higher baseline of energy needs whether we make that discovery or not.
Big tech companies have a long list of promises made over the last 5-10 years promising huge cuts in their environmental impact. Those same companies and their leaders largely abandoned those goals.
I didn't really have political leadership in mind when writing that comment, though they could be part of that "we" as well.
> The people driving AI investment will simply not be significantly affected by climate change. They don't care that hundreds of millions in the tropics will die, and that much of organised human activity will collapse, because up until the last possible moment they'll be insulated from the consequences.
We've spent 80 years globalizing economies in an effort to avoid another world war. We'll all be impacted by it if some of the climate predictions are accurate.
Edit: to add that many of the same leaders developing LLMs make claims that LLMs and AI (if we get there) may be our only hope for finding ways of reversing our environmental impact. Either they are making that up as a sales pitch or they do in fact fall into the "we" here of people that care deeply about our impact while simultaneously burning massive amounts of resources on the hope that LLMs may fix it for us.
Reactors themselves take a large amount of resources, some rare, to build. Infrastructure is another huge resource suck, all that copper has to come from somewhere. Nuclear has the nice benefit of being on-demand, so it does at least dodge resources needed for energy storage.
I'm not even saying I put much faith behind those predictions, but in the context of contradicting climate concerns with tech "innovation" requiring such massive amounts of energy it seems pertinent. We can't have it both ways, either most agree that the climate concerns are baseless or we accept that we collectively would be choosing to destroy the planet faster in the name of progress and innovation.
"Progress" itself is such an interesting term. There's no directionality to it, the only meaning is that we aren't standing still. There's nothing baked into progress that would stop us from progressing right off a cliff, I suppose unless we're already off the cliff and progress could change that.
Solar is a whole other can of worms. I wouldn't expect it to be too useful for LLMs demanding such high energy inputs. Solar is only produced for around 5 hours per day depending on latitude. For every megawatt of energy needed 24/7 for a GPU farm you would need around 5 megawatts of solar and 20 megawatts of storage (ignoring losses along the way due to transmission, heat, and conversions).
> Also, it’s not like there is one person in charge of the whole world deciding what happens.
Totally agree and I didn't mean to imply that. The list is surprisingly small though. More importantly, many of those in charge of the main LLM companies have themselves spoken about how important it is to reduce our environmental impact, going so far as setting very specific targets for their companies to reduce or eliminate their net impact. Those goals all but disappeared after they pivoted to LLM products.