David

[David @djidjienne]

@BRAZZSKILLS @iExplosiveRage @R3DD34D2 @Khalid2c7u @BrutalSam_ @Arabic_Hesham @saudigamer what is the size of the astrobot iso or whatever ?

[David @djidjienne]

@ConfidntRetards Claude or Astra ate that for breakfast

[DSOGaming - DarkSideOfGaming @D_S_O_Gaming]

First Look at the PC Native Port of Super Mario Odyssey
https://t.co/VtMePMcGLF

[David @djidjienne]

@thsottiaux Je pense il a le seum vu que les Belges ont perdu hier

[David @djidjienne]

@maxjendrall which is in contrast with 5.6 which was forgetting a lot of important info afetr compaction

[David @djidjienne]

@maxjendrall yes above 256k or so it uses double the quota per token. I don't know if there are more thresholds at ~500k and ~750k . But compacting at 400K or so is a good idea, and lower depending on your needs. Astra is very good at compacting and little info is lost.

[David @djidjienne]

@BrianMRey waiting

[David @djidjienne]

@Tech2Wild If it is true they will have to reduce drastically the price of Astra

[David @djidjienne]

@Kappaemme1926 64 definitely to be confortable for a professional

[David @djidjienne]

@somehowali 🫡

[David @djidjienne]

@0xSero because you need 50k+$ hardware to run it at decent speed. And one session at a time max. Much cheaper to pay 20$ or 100$ a month.

[David @djidjienne]

@VectorCrossProd anyone that tried at 27B model know it is dum. can do good result only if orchestrated by a much bigger model.

[Yann LeCun @ylecun]

@PessimistsArc Right. Dario was already claiming that GPT2 was too dangerous to open source back in 2019.
I made fun of them then.
Everyone should make fun of them now.

[David @djidjienne]

@techwith_fahim pure scam. fake data. gives a little lag so the model can win, but does not work with real money. you should be banned for sharing this scam.

[David @djidjienne]

@ai_for_success these are stochastic machines that can show large run to run variance. nothing new.

[David @djidjienne]

@Mr_Salio these are stochastic machines that can show large run to run variance. nothing new.

[David @djidjienne]

@chetaslua these are stochastic machines that can show large run to run variance. nothing new.

[David @djidjienne]

@buildwithrajath these are stochastic machines that can show large run to run variance. nothing new.

[David @djidjienne]

@wholyv that's the play. Go back to Claude because it is not nerfed anymore now since all user moved to Astra.
Always use the one that is less hyped.

[David @djidjienne]

@OpenAI @thsottiaux reset to celebrate ?