Press Space to continue
Open-weight models are shifting power from closed labs
Press Space to continue
I would give it ~6 months before normal people realize they are not distilling anymore. That’s an old story. Also, the yolo-run was likely not that cheap. The main difference is just open access to decide how the weights should be distributed.
oguzerkan@oguzerkanJul 17
This is highly misleading. Chinese companies aren’t building SOTA models by just investing tenth of what US companies invest. They are distilling US models so it’s basically US capex + their own capex. They aren’t doing some magic that researchers in the Valley can’t solve. https://t.co/0x2KNxzf9C https://t.co/wPAkQjiya1