3 ms·
What I've always been confused on, why can't we train LLMs to code without ever seeing source code? If it understands human language enough, it should be able
by FloatArtifact 1y ago
What I've always been confused on, why can't we train LLMs to code without ever seeing source code?
If it understands human language enough, it should be able to understand the logic laid out in the documentation mapped to symbols to construct code.
- dtagames 1y agoWe have this already. You can ask Cursor to go read the doc on syntax it may not have ever seen and write something that conforms. I used this recently to support a new feature in Lit which I'd never seen before and I doubt is in the training set much, if at all. You can also describe your own app's syntax, architecture, function signatures, etc. in markdown files or just in chat and Cursor will write code that conforms to your desired syntax, which definitely doesn't exist in the training set.
- FloatArtifact 1y agoYes, but that's not how they're primarily trained.
- baijum 1y agoThis project could be one option for new languages: https://genlm.org/genlm-control/ https://genlm.org/genlm-control/