Sholto Douglas: belief

Transcript context

…Wait, so did the Mistral people just not do a good job of identifying those? It's hard. It's entirely possible that in some respects, there's almost no reason that all of the different archive features should go to one expert. I don't know what buckets they had in their paper, but let's say they had arXiv papers as one of the things. You could imagine biology papers going here, math papers going here, and all of a sudden your breakdown is ruined. But that vision transformer one, where the class separation is really clear and obvious, gives I think some evidence towards the specialization hypothesis. I think images are also in some ways just easier to interpret than text. There’s Chris Olah’s interpretability work on AlexNet and these other models. In the original AlexNet paper, they actually split the model into two GPUs just because GPUs were so bad back then relatively speaking, they were still great at the time. That was one of the big innovations of the paper. They find branch specialization. And there's a Distill Pub article on this where colors go to one GPU and Gabor filters and line detectors go to the other. Like the floppy ear detector, that was just a neuron in the model that you could make sense of. You didn't need to disentangle superposition. So just different data set, different modality.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.