zlacker

>>nyrikk+(OP)
Man that "Unreasonable Effectiveness of ..." pattern is getting a bit overused. With the original paper [1] you could still say that there really is some deeply philosophical mystery. But they now slap that on everything.

[1] https://en.m.wikipedia.org/wiki/The_Unreasonable_Effectivene...

>>ngruhn+0o
What is not unreasonable about intermediate tokens without reason? See the abstract.

>>Meteor+Ox
It's not "unreasonable" if you weren't anthropomorphizing COT, equating it to thinking or "internal dialogue." The results aren't surprising to people in this camp, but I also wouldn't say that makes the work less impactful.

But it would also be more unreasonable to dismiss the fact that a significant portion of the research community (and even greater portion of the public) was operating under these beliefs: that COT was akin to thinking (it's literally there in the name...). It is possible to disagree with something but also not believe someone is being unreasonable by coming to different conclusions.