Should be fine I think

Without the original ChatGPT accidentally tapping into the ELIZA effect, LLMs would have remained a curiosity, with extremely niche backend applications.

But now we are living in the AI hype era, and any conversation about LLMs divorced from the parameters of the AI hype era is fanfiction.

Holding out hope that the first thing I see is the tumblr crab rave gif

I had that already on-hand because of a presentation I gave at work, so right there with you

See also a whole lot of citations here from before the current mess: papers.ssrn.com/sol3/papers....

Less important, but also so that TV writers don't get so very confused about how academic jobs work?

on of my favorite bird-related linguistic situations is gannets, which due to taxonomy fun are no longer in the genus sula (from old Norse súla, meaning gannet)

Yep! I know we had at least one conversation about this ages ago. I will also argue that the way that we teach math only allows for a very narrow type of "good at math" in any case?

and then who is even allowed to be "good at math" is a whole topic of its own

If I actually believed my job was creating sentient entities capable of suffering whose survival utterly depended on me, who I then was forcing into unpaid labor with little awareness of their actual motivation or desires, then I would quit that job because I would not be able to sleep at night.

Also pretty worried that while they might leave SF alone for now, Oakland stuff could still escalate quickly.

Tasha is pretty skittish but likes to sit on the couch and observe while looking dignified (except when making squeaky noises to demand treats)

I like to describe my dad's cat Spot as hyperallergenic (that pile of fluff is pretty typical)

Yeah, there's that and also a lot of stuff about security and interoperability that would require a radically more trusting society, and even then, the only people wearing them regularly in most contexts are functionally government officials - non-Starfleet people usually use consoles

and apparently even seeing it fail in live demos isn't enough to convince people that just maybe it's not actually ready to integrate into serious applications?

Generating database queries from natural language gets bad really fast - "Enterprise history. Aberrant behavior. Medical cross reference" requires so much implicit context to get a useful result, but that gets elided for TV, and now people are getting told there's a tool that can do this

"Tea, earl grey, hot" only sounds weirdly specific if you don't think about all of the assumptions that go into your idea of a default cup of tea (sugar? iced? brewed for how long? does only camellia sinensis count?)