Comment by rybosworld
19 days ago
Hey Brendan - first time listener, first time caller.
Inferring the overall tone from the comments, I think the folks here are struggling with what sounds like a logical fallacy from someone who is certainly a logical thinker.
> how I could lead performance efforts and help save the planet.
The problem on the face of it being: Performance gains will not translate to less energy usage (and by extension less heat released into the atmosphere). Rather, performance gains will mean that more effective compute can be squeezed from the existing hardware.
If performance gains translate to better utilization of the hardware, it also follows that it will translate to more money for the company, allowing for the purchase of more GPUs. Ad infinitum.
My stance is that this is just businesses doing what they do. It's always required regulation to slow down the direct/indirect negative byproducts (petro companies being the most obvious example). I don't see how AI would inherently be different.
Is there another angle that I (we) am (are) missing where the performance efficiencies translate to net benefits for the planet?
Given the widely reported issue of AI environmental cost, and being among the top people in the world who can fix it, I don't feel comfortable with _inaction_. I can reduce datacenter sizes, so I'll do that.
As a performance engineer I'm familiar with Jevons paradox, but it does not discourage improving efficiency.
But that's just it,you're not going to reduce sizes. You're deluding yourself.