[jzymbaluk]

You'd still need those giant data centers for training new frontier models. These Taalas chips, if they work, seem to do the job of inference well, but training will still require general purpose GPU compute

[amelius]

Yeah but you need even bigger factories to fabricate those inference chips, so what is the point?

[bonoboTP]

Next up: wire up a specialized chip to run the training loop of a specific architecture.