Labcoats and Lairs

To be clear, I use Claude for work so I’m aware of its capabilities, and this isn’t a moralizing “how can you stomach using this stuff” post. It’s more that I’m trying to understand the (ostensibly) blanket excitement I see people have over it, and wonder if anyone has mixed or dulled feelings strictly because of like, the larger global context in which AI was born?

So to answer in 2 parts.

Separating from the odious behavior of leaders in the space

(blithely) I mean I'm still on tumblr, aren't I?

Like Altman would probably personally vivisect his parents to weave their nervous systems into a server rack if it would get OpenAI back to the market position it had in 2024. But looking at the comparison-class I'm not actually sure that makes him below-replacement-rate for American tech oligarchs at the moment. Facebook did and tbh probably still does far more damage than every consumer-facing chatbot combined, and burning billions and billions of dollars on a strictly b-tier model is probably Musk's least bad major move in the last couple of years.

If anything given [gestures] effective and focused regulation of powerful AI seems like it might very plausibly put power in significantly worse hands at the moment!

(partial exception: if everyone mid-level or higher at Palantir was conscripted into a penal battalion and deployed to the front lines of whichever war seems most ironic (Ukraine atm I guess), the world would probably be meaningfully improved. Peter Thiel and Elon Musk are two of the more noxious influences on American politics particularly and those of the 'west' in general and provide two of the more compelling arguments in favor of bills of attainder)

2. Actually answering the question

Oh I'm very often incredibly thrown into a bit of an existential abyss worrying about it - especially military and surveillance-state implications, but secondarily the mass unemployment and resulting social upheaval. Though consider that much less of a danger atm that truly effective mass surveillance and/or killbots. I'm absolutely not uncomplicatedly enthusiastic about it (and that's not even considering having to accept some of the most annoying people online in the 2010s seem to be turning out to be basically right, clearly the greatest injustice here).

Just you know, I still don't think the boosters timelines or expectations are right even if I'm updated from 'the next metaverse' to 'we'll need a new model of higher education and there's maybe 5-10 years of tooling before the equivalent of adobe creative suite for visual design hits every non-legally-protected white-collar profession'. If they are right and they build god and/or the demiurge by 2030 I'm fucked regardless, but if not I tend to view it as basically a continuation (and acceleration) of the same automating trends that have been present since the start of the industrial revolution?

And hey, maybe the third time will be the charm and Trump will actually manage to crash the global economy and take the financing or supply chains for the whole data center buildup out at the news before the midterms or Xi will go for Taiwan and destroy the global semiconductor industry or something. Stuff might happen!