@GergelyOrosz if they are as amazing as you say they are and they still joined spacex, then you are wrong in your assumption of how musk treats his engineers
X Holdings Corp
โข SpaceX: Starship & Starlink
โข Tesla: EVs, Energy & Optimus
โข XSI: Compute & Superintelligence
Physical infrastructure + space logistics + sovereign compute under one parent.
Our team in FAIR at Meta is hiring contract Research Engineers.
We work on agents automating AI R&D and the science of recursive self-improvement, develop new test-time learning algorithms and apply them to important problems such as alignment.
๐ On-site (Menlo Park / Bay Area) or remote (ideally US timezone).
Some of our work to give a flavor:
Self-Improving Pretraining : https://t.co/b7MA8YtdnJ
Self-Challenging Agents: https://t.co/gpmLgrKvsH
SWEET-RL (multi-turn agent RL): https://t.co/1DRNJUo0dP
Self-Taught Evaluators: https://t.co/C71wLc3EZV
Self-Rewarding Language Models: https://t.co/snuWbNPUR0
Self-Alignment with Instruction Backtranslation: https://t.co/reHdxnyRAw
AI Research Preference Models: https://t.co/OVlgI89VnK
Autodata : https://t.co/z2oxpLCFms
Recycling the Web (pre-training data quality & quantity): https://t.co/soT6XPGSJc
NaturalReasoning: https://t.co/mwr7XyhHeI
Coconut (reasoning in continuous latent space): https://t.co/8eNO3Jd2mv
CoCoMix (LLM pretraining with continuous concepts): https://t.co/Ep7SBbLUgV
RL from Human Interaction: https://t.co/Ifc6zPg3t2
SPICE (Self-Play in Corpus Environments): https://t.co/K4xdShM8XT
Automated LLM Speedrunning Benchmark (NanoGPT): https://t.co/AHMv5ZDmc9
This is not a full-time role. 1-year contract. Interested? Send your resume to xianl [at] meta.
BREAKING: OpenAIโs solution to NavierโStokes does not match its Lean verification.
The most important article to read today is not one of OpenAIโs 700 AI-generated math papers.
It is this other paper, making a deep and worrying point:
A Lean-verified proof does not automatically validate the proof written in natural language, nor does it mean that the formal statement captures the intended theorem.
During translation, an AI can change an assumption, weaken a statement, or replace the argument entirely.
It can hallucinate another theorem.
Lean correctly verifies the result.
But the proved result may no longer be what the paper claims.
This is a general problem. Things get spicy when the authors examine OpenAIโs proposed NavierโStokes solution.
They identify at least two mismatches between the written intermediate results and their Lean counterparts:
One estimate claims that four additional input derivatives suffice. The Lean version requires five: a weaker result.
A pressure-flux estimate is obtained through a different bound, and proved through a different argument.
It is not clear whether these mismatches invalidate the entire proof.
But they raise an important issue.
OpenAI is flooding us with claimed revolutionary breakthroughs. Yet nobody knows whether the proofs are correct or whether they prove what they claim to be proving.
Epistemia at scale.
*
Paper in the first reply