So they call the competitive players "toxic" and consign those decks to being "netdecks", partially as a way to separate them out from the True, Real players who build decks as a creative outlet

There's a pervasive cultural tendency among less experienced players that deck construction is a form of self-expression as opposed to making a competitive game piece, and when their expressive deck loses repeatedly to similar looking decks that are built to be competitive it feels bad

harder to say that just doesn't work yknow

ftr I do think AI is dogshit for the information ecosystem in an active way and deserves to be rejected on face for this.
i just think that this tendency to run to the extremes re: use case is why we aren't getting the same scrutiny on like. DL fucking things up elsewhere.

bc nuance is hard I guess and people seem to think that giving any ground (admitting use cases) is tantamount to losing

Does things faster/more efficiently/at scale even in cases where that does not apply or where the model itself obscures the information we would want to get. Imo the reaction is often equally strong and ungrounded. If they're saying "AI is necessarily better" we must say its necessarily worse

Like I wrote on proposed use cases for AI models in ecology recently and the primary thing that I've found is that while (DL, not LLM, notably) AI is useful for some applications the problem is that there seems to be some ungrounded techno-optimist belief that AI is necessarily better

Which like, sure, you really shouldn't be using LLMs to research or anywhere that fact checking matters. But to say that they're completely without a functional use case is incorrect. The problem is that the actual use case is extremely narrow and applied, which doesn't market well

The problem I think is that their use case is vastly overstated compared to what they're actually good for, which tends to be the case with the way AI is marketed broadly.
That makes the "they suck and aren't useful" criticisms easy if only because you're comparing use case to marketing line

pip no