Thread

1. dbeley · 1790075630 · focus · HN ↗

   If I'm understanding correctly, each calls to a LLM makes a 3000W GPU run at full speed during the time of inference (let's say from 5 to 60 sec) while Google searches are historically mostly cached so very inexpensive for most searches.

   Isn't the environmental impact 10 to 100 times what we were used to?

   1. mvcosta91 · 1790075901 · focus · HN ↗

      To be honest, the only way to make it more efficient is to use it more, creating an economic incentive to improve the system as a whole. The cognitive impact on future generations is far more important than that.
   2. someguynamedq · 1790078671 · focus · HN ↗

      If we had green energy nobody would care about energy usage

      1. rationalist · 1790083927 · focus · HN ↗

         The Green Energy movement shouldn't have demeaned nuclear. I'm all for not polluting. It reminds me of other extremist groups today that are doing more harm than good about what they claim to care about...
   3. ls612 · 1790086656 · focus · HN ↗

      Not really because queries are batched in production. Your query alongside a bunch of other peoples’ queries are running the GPU at full speed so yours only uses a tiny fraction of the total energy

Unofficial Hacker News client; not affiliated with Y Combinator.