@lawhsw argues AI progress is a theory-free search that recursively accelerates as R&D automates
Critics argue scaled models may become the best theories.
Users dismiss theory-free R&D automation as unable to reach AGI because it lacks essential theoretical foundations.
No Digg Deeper questions have been answered for this story yet.
Most Activity
@sebkrier @lawhsw I like this term "theory free search" but I think it doesn't apply to this situation nearly cleanly as he seems to suggest here. very possible the model itself will be the best theory and that becoming is itself subject to scaling behavior etc
New @lawhsw piece on what automating ML R&D gets you. As usual, worth your time. https://www.learningfromexamples.com/p/theory-free-takeoff
@sebkrier @lawhsw If I may abstract over the arguments of Gary "Dread Pirate" Marcus (@GaryMarcus), theory-free R&D won't get us to AGI, much less SRI. At the moment it looks like it's going to stick the American taxpayers with financial obligations that private industry can't pay.
New @lawhsw piece on what automating ML R&D gets you. As usual, worth your time. https://www.learningfromexamples.com/p/theory-free-takeoff
@sebkrier @lawhsw Staying with the current research sub-paradigm seems safer, then?
We end up with a decentralised ecosystem of supervisors and specialists, something we understand and can maybe work with.