Splinter is your home for news and opinions that challenge power in our political and economic system that's becoming more unhinged each and every day.
Plants that take ten years to build don’t seem like a very good response to a boom that measures dataceneter build time in months. Battery backed solar and wind on the other hand are both cheaper, and can be built faster than the ‘AI’ datacenters they are ment to power.
Don’t get me wrong, I think nuclear power is important to the energy transition, and will find its use in certain use cases like large scale marine transport or places near the artic circle, but the window to build it was 1970 to 2010. At a point when the biggest thing slowing down green energy if finding financing for it, it makes sense to go with the lowest cost option available, which is battery backed solar.
10 years is unrealistic, today it’s closer to 15 years. Both time frames would be too late. Doesn’t matter either way though since the US elections. The world is not going to be able to compensate for the US emissions under Trump’s fascism when we were already on our way to 3 degrees globally BEFORE that.
Not to the degree in which your random smart lightbulbs company has now packaged a poorly shipped “AI” product for the sake of capitalizing on the hype. However, leveraging LLMs has its rewards too. Anecdotally, it has 10x my ability at rapid prototyping. In other examples, I’ve seen it help discover new pathways to better medicines and diagnose diseases more accurately. There’s pros and cons, like anything.
In domain-specific applications, yeah, agree. But building massive datacenters so randos on the internet can ask it to draw a picture of the Spice Girls if they were all anthropomorphic top-hats is a waste of resources, IMO lol.
And that’s why we need nuclear
Plants that take ten years to build don’t seem like a very good response to a boom that measures dataceneter build time in months. Battery backed solar and wind on the other hand are both cheaper, and can be built faster than the ‘AI’ datacenters they are ment to power.
Don’t get me wrong, I think nuclear power is important to the energy transition, and will find its use in certain use cases like large scale marine transport or places near the artic circle, but the window to build it was 1970 to 2010. At a point when the biggest thing slowing down green energy if finding financing for it, it makes sense to go with the lowest cost option available, which is battery backed solar.
10 years is unrealistic, today it’s closer to 15 years. Both time frames would be too late. Doesn’t matter either way though since the US elections. The world is not going to be able to compensate for the US emissions under Trump’s fascism when we were already on our way to 3 degrees globally BEFORE that.
OR…“and that’s why we don’t need AI”
Not to the degree in which your random smart lightbulbs company has now packaged a poorly shipped “AI” product for the sake of capitalizing on the hype. However, leveraging LLMs has its rewards too. Anecdotally, it has 10x my ability at rapid prototyping. In other examples, I’ve seen it help discover new pathways to better medicines and diagnose diseases more accurately. There’s pros and cons, like anything.
In domain-specific applications, yeah, agree. But building massive datacenters so randos on the internet can ask it to draw a picture of the Spice Girls if they were all anthropomorphic top-hats is a waste of resources, IMO lol.
Ok but… now I wanna see that.
lol, right? Alas, some things are best left to the imagination.
Lmao agreed
The same thought can be applied to gaming. It’s just an another convenient scape goat for the oil industry.
and it only needed the power supply of Germany to do it.
No, we already needed nuclear power.