193 sats \ 2 replies \ @zuspotirko 18 Oct 2022
- I think the copyright debate over copilot is very overblown. If you feed the AI enough data the code that gets put out could be way to generic to be attributed to an individual source.
Nobody would sue an artist for making a stroke witth oil paint either if someone else made a stroke with the same angle and color either. It's too generic.
- The fear of copilot replacing all our jobs is overblown as well. It's like wordpress/squarespace etc 10 years ago. Yes these things replaced a lot of websites that would have been programmed otherwise. But it did not make frontend engineers jobless en masse - the bulk of very specific demands of UI are hard to do with generic tools.
reply
193 sats \ 0 replies \ @ZenulAbidin 18 Oct 2022 freebie
I have used Copilot during the beta - strictly for personal projects only as GPL'ed code is not allowed in my org. It gave the best code samples for Python and JS, but only for really simple algorithms and coding patterns. Oftentimes, I was given code that would not compile due to missing variable names. And that's not to mention the snippets did not entirely follow best practices for the particular library or framework it was suggested for.
So it's not on par with a professional developer yet. It's more like a monkey typing pseudorandom garbage text in a code editor - it cannot make a sophisticated program by itself.
reply
26 sats \ 0 replies \ @TonyGiorgio 18 Oct 2022
Your first point is wrong tho, read the arguments the post is making and you'll see that it's not the case. There's straight up copy and pasting going on.
And the solution being "feed it more of the code it is infringing upon and it will stop infringing" cracked me tf up lmao.
reply
193 sats \ 0 replies \ @ZenulAbidin 18 Oct 2022 freebie
There's an easy solution for this: Just generate a file called COPYING_ML which lists all the licenses used by each snippet of code.
You could have something like this:
- A SHA256 hash of the snippet on one line,
- and below it, the license text of the project taken from that hash.
That would solve all the problems related to attribution, and a lawsuit wouldn't be needed. It will also allow someone to easily identify GPL'ed snippets
reply
30 sats \ 0 replies \ @02174b3c39_lntxbot 5 Nov 2022
Fearmongering luddites of the 21st century.
No, we don't want or need regulation controlling the future for the benefit of the past. You like things the way they are, welcome to incumbency and since you've enjoyed whatever the situation is that is about to change, perhaps you're a little old to change and adapt to the new.
Sadly the term of the status quo in intellectual production (there's no such thing as intellectual property) is shortening and just when you thought it all over, it has only begun.
Photography impacted picture drawers and stable diffusion is going to make more pictures than "creatives" ever did.
It was nice while your skills were rare and valuable, but now anyone can get something out of an idea without touching pencil, oil or ink.
Programming, you want it to remain mired in the past, bad luck. The attempt to strangle open source, foss, with copyright is just sick. Copyleft kept it free for all, and that means making it a training set for the next thing to help more people, not a few from the past retain their privileges.
reply