@IntuitMachine But this is a lot and if there are gaps in a local model it only becomes relevant if downstream words show the cracks in the model

@Peregrino1708 @Schuldensuehner Most importantly, it’s not related to German economy but to world Economy as the participating companies are mostly active outside Germany.

@IntuitMachine Don’t they have a locally consistent model within the context?

@gregisenberg People get McKinsey not for their competence, but having someone to blame. Don’t think ai offers that

@weakmath We have a lot of space on the south side. Should not be a problem!

RT @visakanv:

@astuyve … like the interesting part.

@slow_developer Sounds to me more like cope. (And external marketing as far as it is actually based on information from the companies)

@emollick Well, there are ranges of what one would desire. Also with the existing models it is often they not only answer what one asked for, but also like 10 things you could have asked for. Ideally both research as well as normal answering would well respond to length hint

@ChShersh Not having significant rust experience, but curious: is this a problem of the programming language? Is it a lack of optimization of the compiler? Or are code optimizations used that are so expensive that it just slows everything down?

@acherm spoiler: a cubic-millimeter of air has already more configurations than Linux ;-) - No universes needed. BTW: did anyone already look at the relation between Linux configurations and entropy? ;-)

@xpasky Funny, that the crypto thing was apparently with the involvement of Biden in his younger days. - Now apparently takes out pages of this (totally failed) playbook seems crazy.

@DavidSacks Tons of smart takes on obvious stuff that could have been clarified by reading the paper instead of spreading disinformation.

@y_m_asano @aaron_defazio @iclr_conf Indeed ratings from 5 to 10. This looks very random

@kohjingyu been there, done that.. ;-)

I am somewhat ashamed that as a research community we are not able to find a better way than pouring billions into companies for pdfs to be online available.

RT @ctjlewis: .@ollama is awesome. 1. install ollama 2. `ollama run deepseek-r1` that’s it.

@QuanquanGu @percyliang @lm_zheng @ying11231 @zhuohan123 Ideally, a system that people can run locally? The initial system will be too big for many small research groups. Would the 70B distill sufficient in quality?