This was too good not to reshare here as well.

Even qualia love tacos, as recent research has argued (see fig. 3, below). Oops. Somehow the streams have crossed. This is bad. Total protonic reversal bad.

Made this argument about the importance of distinguishing two different kinds of changes in scientific outputs that arise when (a) new tools are introduced and (b) when new populations of researchers come to the fore in a field. Just wanted to drop it here too.

There has already been a bit of a vibe shift (or gaslighting) on the other site, with several industry insiders suggesting “no one ever really thought scale alone could get to AGI” (see 👇) on the heels of reporting that the next gen of models will only have modest gains.

When the suggested solution to the replication crises in psychology and the social sciences is the abandonment of experimental and observational science.

It’s unlikely to get any response over there, but anyone have any thoughts on this? My intuition is that LLMs require a LOT of structure to grab onto to make any headway on something like this. And structure that is necessarily sequential at that, which doesn’t seem to hold here?

Wanted to cross post this over here for the #metascience community.

I really don’t think this is playing a significant role at all.

The neuroscience is pretty clear: https://pbs.twimg.com/media/Dlr8eseV4AA5OKq.jpg