Guardrails removed spam, off-topic, unclear, or duplicate replies.
Ask a question below.
Published answers will appear here.
@NathanpmYoung @Dr_Atoosa If their goal was a "moratorium on AI training" (as we both know it is) and some state level DC construction moratoriums get passed (with dubious counterfactuality), which probably have a tiny impact overall on the global supply of compute pre-2030, I wouldn't count that success.
via X@NathanpmYoung @Dr_Atoosa see of course many things will happen which have some slowdown effect (including the end of the Iran ceasefire (!)), but "and therefore AI pause folks were successful" goes too far in claiming credit IMO
One of my disagreements with the AI safety community is that I don't think an explicit pause or slowdown is a good idea. Part of my reasoning is that there are so many world events that implicitly trigger a slowdown by raising the cost of capital — as yesterday demonstrated.
@herbiebradley @Dr_Atoosa Successful at their aims or successful in your view? Like if they work with polticians on AI moratorium regulations which pass.. then uh.. sounds like they were successful by their own aims. At least in the short term.
via X@NathanpmYoung @Dr_Atoosa No. Though I expect state level datacenter moratoriums in places due to economic anxiety. I think there is some low probability of a slowdown coordinated between OAI/Ant/GDM but not with government involvement.
via X@herbiebradley @Dr_Atoosa And if those moratoriums make AI more expensive then that probably slows it down a bit right? Like I think AI pause folks probably think this is good. Surely we judge them at success by their own aims?
via X@herbiebradley @Dr_Atoosa Sure, but if they do somehow manage to significantly raise the cost of training runs in the US I think that might be a win for them.
via X@herbiebradley @Dr_Atoosa @David_Kasten, do you find @herbiebradley's argument's compelling?
via X@herbiebradley @Dr_Atoosa Successful at their aims or successful in your view? Like if they work with polticians on AI moratorium regulations which pass.. then uh.. sounds like they were successful by their own aims. At least in the short term.
via X@NathanpmYoung @Dr_Atoosa No. Though I expect state level datacenter moratoriums in places due to economic anxiety. I think there is some low probability of a slowdown coordinated between OAI/Ant/GDM but not with government involvement.
via X@herbiebradley @Dr_Atoosa And if those moratoriums make AI more expensive then that probably slows it down a bit right? Like I think AI pause folks probably think this is good. Surely we judge them at success by their own aims?
via X@herbiebradley @Dr_Atoosa Sure, but if they do somehow manage to significantly raise the cost of training runs in the US I think that might be a win for them.
via X@herbiebradley @Dr_Atoosa @David_Kasten, do you find @herbiebradley's argument's compelling?
via XGuardrails removed spam, off-topic, unclear, or duplicate replies.
Ask a question below.
Published answers will appear here.