Ask HN: Do recent improvements to language AI change the spam/scam game?

I know I shouldn't, but sometimes I engage with a scammer--just to try to understand what kind of scam it is. I get curious.

Recently I got one that started as a wrong number and ended with a pitch involving teaching me how to make money via liquidity hedge trading of cryptocurrencies, whatever that is.

It took them four days to get to the point, until then they just acted like they wanted a friend. It was kind of fun. We exchanged pictures of what we were eating, talked about fiction we like, etc. Had a surprisingly deep conversation about our relationships with our parents...

Every now and then I'd get suspicious that they were a bot (they wrote D&D instead of D&D), but then (presumably) a real person would take over and edit previous posts to fix the error and provide a somewhat reasonable explanation involving switching input methods between languages (they were ostensibly Japanese).

My heuristic had always been: if it's a scam they'll want something from you. But if AI can let one scammer do the work of twenty... it could be a very long time before that happens. Despite going into it with "this is a scam" in mind, there were moments when I was fooled. I wanted to believe that friends could still be made that way.

Is this the final nail in the maybe-the-world-isnt-a-dark-forest coffin? Can we fight back?