Just saying this since I know some folks have engaged me about this on here; I've uninstalled/unsubscribed from all AI tools for now. I think they have a place in some workflows; but the (maybe justified?) FUD around use of them is too significant to risk use of them until it's cleared up.
And frankly, the bigger concern is that my use of them responsibly will encourage more larger organizations to use them irresponsibly... and I'm not onboard with that.
@jay I worry about your second point there a lot: I spend a lot of time trying to teach people how to use these tools in what I hope is a responsible way, but it concerns me that my efforts might be enabling irresponsible use too - and does the good outweigh the bad?
My optimistic hope is that by encouraging positive uses I can help tip the scales in that direction
@jay something that's influencing me a lot is the huge boost I've had to my own personal productivity from adopting these tools, in particular my ability to quickly explore and learn new things
It's given me such an advantage (especially as a software engineer) that at this point it feels unethical for me NOT to try and help people understand how I'm using them
@simon @jay Do you ever do live-streams of your usage of LLMs?
Your experience is so much the inverse of mine that I do want to understand what your usual workflow is like. I'm still not sure I'm comfortable using GPT output in ways that I might claim authorship over it, given that I don't want to be an accomplice to plagiarism, but there's no risk of that right now because for me, its output is largely garbage, even when asking basic questions like how to write some part of a pyproject.toml.
@glyph @simon @jay FWIW, when i tried using ChatGPT to ask it about Python stuff I was like, “This is all nonsense!” but when I had projects using stuff like JS / HTML / CSS / SQL, I’ve found it surprisingly useful, because the gaps in my knowledge are basic enough that I am mainly trying to get it to tell me the broad concepts I should be looking into.
@iokiwi @pganssle @glyph @simon I'm a little bit nervous even using it for that, because I don't trust it to point me in the correct direction. I used it as a tool to help jog my memory on libraries I was generally familiar with but couldn't always recall the exact method names on them or to help me generate boilerplate such as python docstrings.
@pganssle @glyph @simon @jay yeah I like to use it like a reverse dictionary or to get a 1000ft view so I can then go off and know what to google for or search for in the docs.
Sometimes it helps me discover libraries and frameworks I didn't know about.
I sometimes generate code but it often gets reworked so much that It ends up mostly mine anyway. It's a bit easier to refactor/rewrite someone (something) else's code than to start from a blank canvas.
I feel like it works okay as tool if you already sorta know what your doing