Jj McMc

[Jj McMc @AccountMus629]

they pay me to post here

[Jj McMc @AccountMus629]

@podesta_aldo I do really wish you the best and that you will be able to do a pretrain run at some point. Starting from a chinese base doesnt really get us anywhere.

[Jj McMc @AccountMus629]

@podesta_aldo The problem isnt that you started from a pretrain, the problem is that you are selling this as "Europe is back in the race". When we clearly arent - we cant even get our own pretrain (please dont say "Mistral").

[Jj McMc @AccountMus629]

@LexnLin Also a benchmark where the LOWEST score is 99.1% is toooootally not saturated

[Jj McMc @AccountMus629]

@elshayib_ @notjazii Even stupider chart: max being consistently cheaper

[Jj McMc @AccountMus629]

@PussyBlaster10k @EERandomness @insane_analyst "Windows is great! You just gotta install a linux VM on it!"

[Jj McMc @AccountMus629]

@maxhanet @archived_videos That simply means the answer is "No" - there is no posttraining FROM SCRATCH. Youre just being needlessly obtuse because you know that just posttraining a chinese model isnt actually of any value

[Jj McMc @AccountMus629]

@Nightshadessx @maxhanet Just look at whatever Kimi scores - this model scores that

[Jj McMc @AccountMus629]

@maxhanet @Litesout86 Oooh, thats brilliant! Its crazy that all the other AI companies havent come up with doing more with less compute. Someone should tell them!

[Jj McMc @AccountMus629]

@maxhanet @quentin_hk Kimi is the thing that is scoring on those benchmarks. Go ahead and hit those numbers without it

[Jj McMc @AccountMus629]

@maxhanet @alfaenger Translation for anyone not coping: Yes, it is

[Jj McMc @AccountMus629]

@maxhanet @synquid "Only a small part of the recipe" it is literally the pretraining base. Thats like saying potatos are "only a small part of the recipe" when youre making french fries.

[Jj McMc @AccountMus629]

@kimmonismus yeah but how else would Microsoft make money? They need the valuable income from these ads, or they would go under!

[Jj McMc @AccountMus629]

@merman1974 @823jjd AI objectively has a value, it can get work done

NFTs dont provide value at all

[Jj McMc @AccountMus629]

@kwargq That is not at all how this works, artificial analysis doesnt give a shit who wins

[Jj McMc @AccountMus629]

@Call_me_Ra1n @tina_trautmann @Paddy_film Aber die Regel macht Sinn, das Gewicht ist der entscheidene Faktor bei einem Unfall (Impuls = Masse × Geschwindigkeit)

Blöd wenn man so was nicht versteht

[Jj McMc @AccountMus629]

@synthwavedd It is fucking 10 points behind on Artificial Analysis - i dont care how cheap this model is, they could pay me for using it and i still wouldnt

[Jj McMc @AccountMus629]

@dummerspast39 Wdym "What the fuck is going on at Anthropic"?

If the chinese WERENT trying to steal their weights, id be more confused

[Jj McMc @AccountMus629]

@krawallburscht @tinitinsch Boomer ist ein Mindset, keine Generation.

[Jj McMc @AccountMus629]

@MONKE2525E @theo Gemini: Time to fully use my 1,048,576 token output limit! That still leaves some space for the user prompt, as gpt-6.1-sol has a context window of 1,050,000 tokens!

[Jj McMc @AccountMus629]

@theo no