Reaction
Could banning production of training-capable chips slow AI?
The proposed ban would allow inference-only chips, but a critic says inference can substitute for training up to a limit.
TLDR
A coauthor of a working paper on a “hardwired pause” of frontier AI training asks whether such a pause could work globally and for how long. A post discussing the paper describes banning production of training-capable chips while allowing production and use of inference-only chips. A reply on X argues that more inference can substitute for more training, up to a limit, and calls the idea unworkable.
Combined views
2.7K
2 Sources, first seen ago
