https://www.cs.dartmouth.edu/~dfk/research/aslam-kerf-news/a...
which had a machine learning component that automatically helped refine syslog SQL queries to find anomolies. More info here:
https://www.cs.dartmouth.edu/~dfk/research/project/kerf/inde...
Before LLM's.
AGI doesn't have to do everything better than every human ever born to be generally useful and constructive.
If makes you feel better, think of it as dumbAGI - a generally capable AI that's not that smart.
GPT4o is a dumbAGI. You and I are dumbGIs.
I wouldn't ever do this in an automated way. However in person, it not only makes me solve problems maybe five times faster, but there are also many errors that I would give up on, which I can now solve easily.
Just the ability to turn a log into a human readable narrative does something remarkable. After I've done it once or twice, then my brain provides the narrative itself, and I can read the raw log without the need for assistance.
This suggests that not only does the AI assist me, but it assists my learning as well.
I also struggle with Man pages, in particular finding an option to do something that I need to do. I end up having to read through dozens of options getting more and more confused, sometimes skipping the option that I needed and not finding it. Now I can just dump the entire man page into an llm, describe what I need to do, and ask it to highlight the sections that I need to read. Or I can just ask it questions.