@aledeniz@spandrell4 (Google here employee who has not worked on Gemini nor TPUs) If I understand Google TPUs and what OP have said these are not the same thing. Google TPUs could in practice work on many different models, what OP has done requires burning the model directly in the chip
@RafaelChia_CM@carlosfm75 Eso da igual, como si Allianz tiene un 0.1%. La compañía que te tiene que atender es la entidad jurídica con la que tienes el contrato. No una de sus propietarias. No califico la mala atención que puedas haber recibido, pero si insistes en algo tan erróneo es difícil ayudar.
@nntaleb Transitivity does not always hold up on deterministic too right? And on correlative it could be valid if there is a causal chain? Am I missing something?
@HarmonicMath Hi there. I have a Q1 publication accepted. A certain set of theorems could use Lean proofs for my second publication on Q2-Q1. I have already applied. Anything else that O should do?
@sama SAT/MaxSAT knowledge would be pretty cool. It's pretty efficient and proper modelling is as complex as MILP. I could assist with help in this area.
@willdepue Lot's of people saying login with OpenAI to connect external apps to their memories. I don't think this is the future. Too much risk of data leaking. The way is a login with OpenAI that allows apps to use the pro account (aka: API keys for users with subs) Imagine Alexa with this
@DevLatra Creo que hay un cambio en el sistema de tributación de empresas de Estados Unidos. Me ha parecido leer que ahora los costes de desarrolladores de software no se pueden desgravar como gasto si no que se amortizan al 7-20%. Eso sería MUCHA pasta que hay que sacar de algún lado.