Just by chance I saw your request - and of course I wondered whether #ChatGPT could do this. It can do many transformations as long as you can pass them through some tokenized abstraction, and for structural formulas the abstraction would be SMILES.
--------
Prompt:
Do you know the SMILES string for caffeine?
ChatGPT:
Yes, the SMILES string for caffeine is "CN1C=NC2=C1C(=O)N(C(=O)N2C)C". [plus a bonus explanation]
---------
You can confirm that easily with one of the common free, online tools such as the one here:
https://www.cheminfo.org/flavor/malaria/Utilities/SMILES_generator___checker/index.html
... and you can also generate a 3D-model, eg. through the "build" command of the amazing(!) free ChimeraX visualizer.
-----
I haven't tested this extensively regarding its boundaries of capability and fact ... but just at first glance this looks like an interesting angle that I haven't seen before.
It also works the other way around:
---
Can you identify this one: "OCC(O)C(O)C(O)C(O)C=O" ?
Yes, I can identify the molecule represented by the SMILES string "OCC(O)C(O)C(O)C(O)C=O". This is the SMILES string for D-glucose, which is a type of sugar and one of the most important carbohydrates used by living organisms for energy.
----
This is also correct - but not completely: it could also be L-Glucose ... and even though ChatGPT knows about Isomeric SMILES, it can't generate or parse them correctly. That may not be the end of the story, perhaps only a different approach is needed to phrase the request so it can be properly sequenced. I'll probably revisit this at some point.
---
Finally: for checking the ground truth, I send my students to PubChem (e.g. https://pubchem.ncbi.nlm.nih.gov/compound/Caffeine#section=CAS)
Enjoy!
#SentientSyllabus #ChatGPT #HigherEd #AI #Chemistry #Education #SMILES