Who here uses AI in programming?

[savoita]

For me, the ethical issues around using code-generating large language models make it non-viable for me in its current incarnation. Additionally, I do not trust that I would maintain all of the skills I require to be decent at programming/osdev (to my standards, I guess) if I were to incorporate the use of a large language model. I am comfortable enough with the tools that already exist, and I don't find that writing code down is really what takes the most time in my development.

Hope that makes any sense.

[nullplan]

There is also the issue of copyright, as the copyright status of LLM output has not been contested in court. When I, a human, read code and become inspired to write something myself, as long as it doesn't reach the level of "striking" similarity, it is not considered a derivative work (a fact that has been talked to death in the Katy Perry Dark Horse lawsuit). But when an LLM does the same, it is not yet known if the output will count as a derivative work. Therefore, even though program output cannot be copyrighted and thus the LLM cannot own copyright in its output, the output can still be encumbered by the copyright of others.

For that reason you would have to train the AI exclusively on public domain or WTFPL code. BSDL code cannot be used, since even the BSDL requires attribution that an LLM does not provide. To say nothing of GPL or proprietary code.

[Edenace001]

Personally, I use it to make sure my plans are achievable, and realistic to some extent, and I use it mainly just so I can abstractly convert it to my goals. Does that make sense? Such as currently, I am attempting a simple OSDev Project in which I was trying TLSF for the first time, so I made an LLM explain what in the world it actually does and is. However, I am against using it for "Code Generation" and "Vibe Coding," with the exception of:
- Bug Hunting
- Debugging
- General Fixes or Problem Analysis
Anyways, have a good day!