NHacker Next
- new
- past
- show
- ask
- show
- jobs
- submit
login
▲Using AI to improve a challenging reaction in medicinal chemistry (self.__VINEXT_RSC_CHUNKS__=self.__VINEXT_RSC_CHUNKS__||[];self.__VINEXT_RSC_CHUNKS__.push("2:I[\"aadde9aaef29\",[],\"default\",1]\n3:I[\"6e873226e03b\",[],\"Children\",1]\n5:I[\"bc2946a341c8\",[],\"LayoutSegmentProvider\",1]\n6:I[\"6e873226e03b\",[],\"Slot\",1]\n7:I[\"3506b3d116f7\",[],\"ErrorBoundary\",1]\n8:I[\"a9bbde40cf2d\",[],\"default\",1]\n9:I[\"3506b3d116f7\",[],\"NotFoundBoundary\",1]\na:\"$Sreact.suspense\"\n:HL[\"/assets/index-BLEkI_5r.css\",\"style\"]\n")target="_blank">openai.com)
Rendered at 18:30:47 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
https://x.com/jmalchow/status/2067298271647904061
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.
requires Twitter sign in
https://xcancel.com/jmalchow/status/2067298271647904061
does not, although sometimes spotty
It's basically high through put screening plus an AI engine to map out the "variable space".
Back in 1990 when robotics became more reliable we did the same thing. The only difference is a trained chemist would determine what variables would be altered.
It's not that hard to do, it doesn't take that much brain power, just an understanding of what variables may impact the yield. Claiming AI can now do this isn't all that impressive.
It could be as unimpressive as motorcycling across the USA in 33 hours vs 7 days
One thing about the high throughput screening when it came out, it was immediately obvious that it produced so much data that one person could never "delve" into the results of any single run to more than just a limited percent. Before moving on to the next run.
I knew at the time, decades ago, that I would have to wait for AI if I wanted to make the most of it for that reason alone. And it would likely be accomplished only by those having much bigger resources than I was likely to have access to.
So I decidedly went the other way where we basically avoid very much bulk reaction at all, especially not producing novel products in the analytical chemistry labs.
Still expected to wait for AI until at least 1999, since for what I need AI would have to be deterministic to a 100% reliable degree. Ab. So. Lutely.
Or it will not replace the work of a 98 to 99% reliable human.
https://www.anthropic.com/research/making-claude-a-chemist
I don't know why they think this is okay, any more than it would be to call their models "AI doctors" or "AI lawyers".
Ironically, where I live, most actual medicinal chemists are not allowed to call themselves chemists since you have to pay dues to a professional association to use that title.
It made me wonder what other professions require an associate's degree or better to be able to claim the profession without some sort of modifier, such as licensed physician, or Master plumber...