3 ms·
Why? Is there any reason to believe problems in that context won't generalise?
by lmm 1y ago
Why? Is there any reason to believe problems in that context won't generalise?
- mh- 1y agoLots, yes. The fine tuning may attempt to introduce concepts that were intentionally omitted from the training data for safety* reasons. Maybe nothing wrong with that, but it might mean that the perceived weaknesses don't generalize to an area of the model that hasn't been lobotomized. * using safety the way OpenAI have been using the term, not looking to debate the utility of that.
- mvdtnz 1y agoAre you serious? Of course there are reasons to believe they won't generalise.