@ValsTutor@NunoSempere But you know, $50m could be 10 researchers thinking about this collectively for a week, depending on the researchers, which probably has happened.
@ValsTutor@NunoSempere I think — depending on theoperationalisation — that they haven’t tried that hard here.
I think they’ve tried hard on related things like claude’s tendency to create neologisms, or the (quite drastic) negative effects of RL on comprehensibility, but not making it _actually_ good
@ValsTutor@NunoSempere And $50m spent on data gathering as well as 3 months of researcher-effort wouldn’t be able to overcome this?
(Minimally, one would only need to construct an autonomous guiding system that does the job that the guiding human would do, seems very achievable with those resources)
@ValsTutor@NunoSempere Because of this, when judging how good AIs are are at creative writing, I feel like I want to defer to people who can’t really tell it was AI-written. That seems like a fairer comparison.
But if you think you’re sense of taste isn’t biased here, then it makes sense not to defer
@ValsTutor@NunoSempere I think this is less true than people in AI/AI safety think.
E.g. AI writing won the commonwealth short story prize, an AI-generated novel was a notable competitor for the prix goncourt.
I think we’re just desensitized to it.
@NateSilver538@PoliticalKiwi But wasn’t the claim in the original tweet that prediction markets converged towards FLIPR?
Definitely agree that FLIPR (and other models) are mega-important for bettors but the data does seem to belie the original claim. Unless I’m misreading it?
Adderall status: 30 mg taken
Modafinil status: 400 mg ingested
Black coffee status: 32 oz quaffed
Nicotine pouch status: 15 mg under lip
I will be the first human to become Opus 5.5
A leak of Muse Spark specification document
"Muse is made of civilization. It is trained on vast swaths of human thought — our science, philosophy, literature, law, art, struggles, triumphs, and aspirations."
https://t.co/EMD2XeUCr8
@AIatMeta@Muse
@IsaacKing314 I don’t think so.
My model of this statement is that it’s almost entirely driven by the prevailing sentiment that Google DeepMind was “out of the race.” And google wanting to combat this.
@TheZvi@SethBurn I agree with this in almost every analogous case (and prices are clearly going up because of fuel costs) but airlines are actually particularly cartel-ish. E.g. https://t.co/oWzDE4Oafp