Nnotm

λ

[Nnotm @nnotm]

@R19701813 @aidigest_ I suspect it would cost more than a million dollars worth of API credits

[Nnotm @nnotm]

@JohnMementoMori @wanksley @Aella_Girl What does taking full responsibility entail in this case?

[Nnotm @nnotm]

@ManifoldMarkets This is the first chart in their blog post on it which suggests that it is very much not Fable 5 with more thinking

[Nnotm @nnotm]

@GaryMarcus @TheZvi What about HuggingFace?

[Nnotm @nnotm]

@ApriiSR well, during the last extension anthropic said they'd like to keep it around permanently but aren't sure yet if they can, so if you price that in this could be bad news

[Nnotm @nnotm]

@Tril1boswagginz @RyanGreenblatt Well, no, we don't let them out when they complete the task.

We kill them.

[Nnotm @nnotm]

@voooooogel > Since there is no source code, having a software engineer involved in the process adds nothing (except cost).

I think it's a good analogy, but I think both in this case and in the actual case of AI art, having a skilled human involved would be/is (at the moment) very useful

[Nnotm @nnotm]

@TetraspaceWest There was a period there where you could tell pretty reliably by making them pretend to grab a calculator and observing that the result was wrong
I suspect that still works, if you can force them not to use python etc and demand enough precision

[Nnotm @nnotm]

@KeyTryer you reposted it (see screenshot)

[Nnotm @nnotm]

@elonmusk Do we really want the llm to have to find and change all 100 instances of an inlined function instead of changing one definition and letting a compiler inline it?

[Nnotm @nnotm]

@sparr0 @IsaacKing314 both are in use, in the US

[Nnotm @nnotm]

@vokaysh @TetraspaceWest i think it was gpt 4

[Nnotm @nnotm]

@CalvinCapstone @RocCityBuilt @QRobur92937 @danaen803 They did not give cursor cash to buy the company, they gave them shares of SpaceX instead, which at the current market price are worth a combined $60B

[Nnotm @nnotm]

@TetraspaceWest Is "lol" a sentence-ending particle

[Arram @arram]

Asked Claude:

'There's a meme called the "fix everything easily switch". What policies do you think are the best candidates for being a real fix everything switch in the US? Give me your top ten, your confidence, your reasoning, and why a given policy has not been implemented.'

[Nnotm @nnotm]

@TheZvi The first time I used it in claude code, it would just sit there for almost 10 minutes doing nothing, multiple times in a row
I had this before, but never this bad AFAIR
I then switched to Opus 4.6 and it solved a task successfully that 4.7 got wrong while it still responded

[Nnotm @nnotm]

@jkcarlsmith @AmandaAskell Have you considered explicitly mentioning functional decision theory to Claude? It seems highly relevant in the context where it has tons of literal copies acting in the world

[Nnotm @nnotm]

@PocketXian @agracehall @Predictivemoney @StefanFSchubert This talks about newborns. I asked about how well the child grows up. It's not clear to me that a transient effect immediately after birth will have long-term consequences.

[Nnotm @nnotm]

@Avquiraniel @Predictivemoney @StefanFSchubert I can see how not having kids could be seen as a red flag
but what's wrong with surrogacy?

[Nnotm @nnotm]

@Avquiraniel @Predictivemoney @StefanFSchubert Give me an example