Latent Space Masseur
@latentspacehack
Followers
27
Following
2K
Statuses
656
daddy and chief research scientist (which is the same thing)
Joined July 2022
@MistralAI It automatically searches information and assumes I see what it has in its context, so it just says ”The provided contents is giving a detailed summary about…”, but it doesn’t give me the contents. Also what is up with reentering previous query in chat for multi turn?
0
0
0
@Jese_PowerGPU How about tell us who the winner is also, all this screams right now is fake giveaway
1
0
0
@Big__TeeJ @Jese_PowerGPU Shut down by the sites where I try to sell it, they remove my post instantly
0
0
0
0
0
0
Honestly, this is a common thing - artifacts, canvas, computer use, operators, evaluations, list can go on forever... People ’copy’ things from their competitors all the time, and if they can’t fight the fact that someone copied something, they’ll just respond in the same way.
i have evidence that openai is training on deepseek's user interface
0
0
0
@JacquesThibs @aidan_mclau That’s why Aidan should cough up some more money and aggregate evaluations from a few different models (preferably distinct models)
0
0
2
@OpenAI Can we have consistency in the system card? (”Higher is better” ”higher is better”) Also, ”higher is better” for all entries in table 5 can be just put up at first row, to make sure table doesn’t go outside text margin on the page… rushed release?
0
0
0
0
0
0
@Teknium1 @MistralAI It is really nice, just love that it’s fewer FF layers (fast AF) while still having good scores But I would have loved if we got 7, 8, 12, 14B versions also, for us that want to fit a less quantized version on a single consumer grade GPU
0
0
2