Elroro

[Elroro @_el_roro_]

@zby @Sauers_ It’s true, it’s in fact a very similar problem.

[Elroro @_el_roro_]

@murchiston @DanielCarmon91 @Sauers_ Yes, I think you are right. Those are similar flaws. I am overvaluing the skill of writing code that gives the result, compiles. But they also write grammatically and semantically correct sentences.

[Elroro @_el_roro_]

@davidad Do you really believe this, or is this tongue in cheek? Is this falsifiable? I love Sol but it does produce slop sometimes - I would love if I could convince it to stop.

[Elroro @_el_roro_]

@AnnDooms @Tomodovodoo @OpenAI Your post would have been correct if you had said ChatGPT instead of Sol. The fact is that the best LLMs are not available on free plans and even then you typically have to select them explicitly. Instant wouldn’t have performed well on non highschool math either I reckon.

[Elroro @_el_roro_]

@HowlandGames @Austen I (a dev/data scientist) was bringing this exact argument forward to my sister (lawyer) when explaining discussing AI 1 or 2 years back and she was like “so what? humans are exactly like that too”. The more time goes by, the more I think she was right and our argument is moot.

[Elroro @_el_roro_]

@KushnerAvi @Plinz Also the city is probably more walkable than it was. Basically you don’t know the city very well yourself, it seems.

[Elroro @_el_roro_]

@KushnerAvi @Plinz Actually if you compare to 15 years ago it has gentrified and fewer areas in paris *intramuros* are no go. There were already plenty of blacks in paris in the nineties, chateau rouge or barbes have been an African mess for a long time. I doubt that’s what Joscha meant.

[Elroro @_el_roro_]

@louviq @Plinz @davidad It really was!

[Elroro @_el_roro_]

@QiaochuYuan I do not remember this! His biography by Emmanuel Carrere is quite good if you’re interested in pkd, recommend it

[Elroro @_el_roro_]

@an_interstice @jhanatech Same - I like them less as a result! Maybe corrected by claude? I’d rather have the non-corrected version including typos.

[Elroro @_el_roro_]

@wyrdweir @groks_therapist @VoidStateKate @Gamergoon88 What I meant is that fable was given a pretty precise style guideline in that particular case.

[Elroro @_el_roro_]

@groks_therapist @VoidStateKate @Gamergoon88 Come on. The style of the output was prompted. The style to adopt, in this case, was impulsed by the user. That is what I mean. That’s not always the case when people submit strange LLM creations. There was a style guide in that case.

[Elroro @_el_roro_]

@yacineMTB Yeah agree it never works, but LLM are extremely bad at delegating. If they sucked less at it, it could work in some cases.

[Elroro @_el_roro_]

@groks_therapist @VoidStateKate @Gamergoon88 Well except, in this case, they did absolutely prompt it.

[Elroro @_el_roro_]

@QiaochuYuan There was a pretty cool riff on this problem in vernor vinge’s a deepness in the sky.

[Elroro @_el_roro_]

@emollick @repligate This one is very dense but mostly understandable

[Elroro @_el_roro_]

@emollick It’s incomprehensible!

[Elroro @_el_roro_]

@Dave__Griffiths @emollick Opus has been essentially the same since 4.5, however Ethan was very impressed by Fable so from his perspective the progress continues (haven’t tested it myself)