Ryan1729

@tbortels@infosec.exchange @mhoye@cosocial.ca It's unclear to me exactly what kind of useful patterns the LLM is "inferring" for you, and what useful output it is producing as a result of this "inferring". You mention python and log records being different based on event types. Do you mean the LLM is outputting python code that branches on say a "type" or "kind" field in the JSON?

@tbortels@infosec.exchange @mhoye@cosocial.ca I did a search for jq utilities that work how I described: selecting parts of the text to focus on produces a query, for example, selecting "Alice" in [{"name": "Alice"}, {"name": "Bob"}] gives you .[] | .name, and I didn't find any. Maybe that's not what you meant, or maybe I missed them.

@tbortels@infosec.exchange @mhoye@cosocial.ca For that specific task, it feels to me like a purpose built tool could be made that would ultimately be faster to use than an LLM.

I'm imagining something where you paste in a JSON blob, then select a part you want, and it generates a jq command to extract that from the structure you pasted.

I think I'll add that to my project list, but anyone else can feel free to make that. I'd be interested to know if that ever exists, whether it was in response to this post or not.

@tbortels@infosec.exchange @mhoye@cosocial.ca Do you have an actual example to hand of a copy paste job that takes 4 hours? I can recall cases from coding myself where beating a large template into shape felt like it took forever, but in my cases, the actual wall clock time was nowhere near that.

@cR0w@infosec.exchange Human named Aaron Swartz who pirates scientific papers with the intent to free knowledge and speed up science: Goes to prison. Tech bro who pirates everything to create a machine to rent to everyone, which slows down science because reviewers need to wade through slop: Nearly free marketing (Just one time costs of legal fees and court settlements)