Not a biologist but my impression is that a lot of progress in biology came from refining and validating existing techniques. Also building up a large library of biological specimens & phenomena, i.e. taxonomy. The esthetic and practice of MechInterp seems in accord with that.
Yeah, but historical biology wasn’t as time-constrained as modern MI, which has alignment to solve.
My point is that for MI now it would be better to do “breadth first” search, trying to throw at problem as many ideas as possible instead of concentrating on small number of paradigms like SAEs.
Not a biologist but my impression is that a lot of progress in biology came from refining and validating existing techniques. Also building up a large library of biological specimens & phenomena, i.e. taxonomy. The esthetic and practice of MechInterp seems in accord with that.
Yeah, but historical biology wasn’t as time-constrained as modern MI, which has alignment to solve.
My point is that for MI now it would be better to do “breadth first” search, trying to throw at problem as many ideas as possible instead of concentrating on small number of paradigms like SAEs.