AI Expert Flags Power Concentration As Top Risk, Urges Open Source

@ClementDelangue @PrimeIntellect Agreed! We also recently wrote a blog about a potential path forward.

https://pgasawa.bearblog.dev/unsafe-ai-or-consolidated-power-ais-false-dichotomy/

@PKUWZP @ClementDelangue @jeremyphoward In the age of internet we can raise capital by crowd sourcing if someone credible (in public eye) is ready to build it, like goal should be open source data, pipeline, model etc etc.

We need free software movement similar to that took place during GNU era.

@ClementDelangue do you have a thorough rationale for this that youve written/can write? would like to know why you think this!

@ClementDelangue @omarsar0 Safety open science/source...

@ylecun @ClementDelangue @Dan_Jeffries1 One Pattern , 100 Languages,Smarter AI

how a universal syntactic cuts model size and ground generation in linguistics structure

@ylecun @ClementDelangue @Dan_Jeffries1 The roadmap is concrete — first base model by end of 2026. The hard part isn't the timeline though, it's federated training across heterogeneous compute matching sync baselines. Curious who's committing early compute.

@ClementDelangue Absolutely — AI should empower everyone, not just a handful of players.