Original post
François Chollet (@fchollet)@fchollet
Post text is unavailable here. Open the original source.
· UTC+8
François Chollet argues that what matters is what a model is trained or programmed to do, rather than its computing substrate. He calls Transformers a generic substrate and LRMs a new idea from 2024 built on older foundations; his assessment of fluid intelligence is an opinion.
Post text is unavailable here. Open the original source.