This is really, really cool. I've been dreaming of a system like this for years (with better persistent context and a voice interface) - a proper Star Trek computer!
Pretty impressive. I could see myself using this quite often.
Wonder if it has already scraped all the man pages for the various bash tools or if doing so would improve it even more.
Are we supposed to assume that the only inputs for this model are the prompts shown? Is GPT-3 already trained on shells somehow?
I think the implication is that it has read a _lot_ of information on the web, I think their training data is (in some part) based on reddit links, and that information is then just passively encoded in the model weights. Some of the information comes from the few-shot examples, at least in terms of what meta-form an 'answer' takes, but I think the implication is that GPT-3 is really just that knowledgeable.
How do they separate bad or malicious commands learned from internet? Like
“List all folders” rm -fr * (just a silly example)