Chemist suggests retraction of DeepMind robotic synthesis paper
twitter.comJust for clarity, the linked paper in the twitter thread is "An autonomous laboratory for the accelerated synthesis of novel materials" (https://www.nature.com/articles/s41586-023-06734-w) which does have two authors from DeepMind but seems to be mostly from material science researchers at UC Berkeley. This thread is not about the recent Nature paper "Scaling deep learning for materials discovery" (https://www.nature.com/articles/s41586-023-06735-9) from Deepmind which made news a few days ago.
The linked paper (https://www.nature.com/articles/s41586-023-06734-w) in the quoted tweet appears to be from the Ceder Group at UC Berkeley, not DeepMind. Is there a different link I'm missing?
As the paper notes, two of the authors are from DeepMind, the chemicals they are synthesizing are those predicted to be stable by a DeepMind model, and the entire work arises out of an extensive collaboration between DeepMind and the Berkeley lab: https://www.nature.com/articles/d41586-023-03745-5
Missed that -- thanks!
It appears that the OP misread the tweet. This has nothing to do with DeepMind.
For reference (in case the title gets changed), this post is currently titled "Chemist suggests retraction of DeepMind robotic synthesis paper".
I didn't see him explicitly suggest retraction in that thread, although he's certainly very upset at how bad the study is and thinks it should never have been published.
Yeah the lab replied defending their paper with some more details and he started a whole new thread where he did actually explicitly recommend retraction:
https://nitter.net/Robert_Palgrave/status/173088441028562563...
Thank you, appreciate the link
A somewhat related question: let's assume that a super duper ChemGPT has discovered new, heretofore unknown, molecules.
What happens next?
Is there lots of work still to be done before the molecules are "built"? How would a lab determine the properties of truly novel compounds? How would you even figure out how to synthesize a new molecule?
You should read the paper because they claim to have done it already with robots mixing powdered forms of these chemicals.
From the paper (but read it, there's even pictures):
> Researchers initialized the A-Lab by proposing 58 target materials, which were successfully realized at a rate of >2 new materials per day with minimal human intervention
Going from a target molecule to prospective synthetic pathways is often done by a process known as retro synthetic analysis.
Even a novel molecule spewed out by a computational model is going to be able to be broken down into chunks that look like other molecules, with plausible ways to stick em together or rearrange them.
So you use knowledge of existing molecules, reactions, etc to figure out a bunch of plausible pathways.
This can be done with a bunch of grad students, or, with the use of modern computer models - pretty much all of which are proprietary.
I’ve often compared the process to software reverse engineering, and I’m hopeful that maybe soon there will be open source AI/ML models enabling automated retro synthesis and constraint solving
Headline is wrong. The chemist in question is suggesting redaction of a paper published by Ceder Group, not by DeepMind.
i work alongside this group ( one of Ceders PhD students), but not on this project. I believe OP is referencing a different paper to that of the tweet author
Seems like a pattern with these DeepMind papers...