3 ms·
That could be. If so, it would be interesting to know how Copilot does that. For me, just asking LLMs "Can the following function be improved" for a function I
by mg 3y ago
That could be. If so, it would be interesting to know how Copilot does that.
For me, just asking LLMs "Can the following function be improved" for a function I just wrote is already pretty useful. The LLM often comes up with a way to make it shorter or more performant.
- nilsherzig 3y agoI just tried the gpt4, without any modifications it's impressively worse than the current chat model
- mg 3y agoWhat did you try?
- nilsherzig 3y agoRunning some queries in a new chatgpt session and via the API. I tried adding the same system prompt on both. I can run one for you, if you want :)
- mg 3y ago"some queries"? Show them, so we can discuss?
- spenczar5 3y agoYes, the official plugin sends context from recently opened other buffers. It determines what context to send by computing a jaccard similarity score locally. It uses a local 14-dimensional logistic regression model as well for some decisions about when to make a completion request, and what to include. There are some reverse-engineering teardowns that show this.