[esafak]

That is not true; first he says it's "tens of millions of dollars well spent," followed by "you never know". I don't think he knows.

[lyjackal]

I’ve wondered whether they use thanks as a signal to a conversation well done, for the purpose of future reinforcement learning

[eru]

I'd speculate that they use slightly more complicated sentiment analysis. This has been a thing since long before LLMs.

[esafak]

I don't know if they do or if it is efficient but it is possible.