← Back to context

Comment by malchow

6 hours ago

Well-run autolabs recursively training discipline-specific models are becoming very valuable assets. I write more about this here:

https://x.com/jmalchow/status/2067298271647904061

8 comments

malchow

Reply
wasabi991011  5 hours ago

Is there a way to read this without a Twitter account?

  • WithinReason  4 hours ago

    https://xcancel.com/OpenAI/status/2067293746556027376

    GPT-5.4 reviewed scientific literature, generated and ranked research proposals, helped design experiments, analyzed results, and proposed follow-up studies.

    Human chemists steered the work, selected proposals for testing, and validated the final result.

    Maria [AI] tested the idea across 10,080 reactions, and human chemists later validated representative results by hand.

    Under the optimized conditions, yields improved for 88% of the boronic acids and 83% of the sulfonamides tested.

    Human chemists then repeated 14 representative reactions by hand: 11 showed higher yields, including 8 with a more than twofold improvement.

    The full process took about 2.5 months, plus another half month for human chemists to write up the results.

  • CamperBob2  2 hours ago

    It kills me how many scientific researchers in various fields still rely on X. Every time they post or read a tweet, they are subsidizing the guy who wants more than anything else -- having already gained everything else -- to put them out of a job.