copygirl@lemmy.blahaj.zone 4 weeks ago
Yeah, this video was really cool, and then the author went “if you’re not familiar with the command line, tell your AI agent to help you set up SSH on your Steam Frame”. Like, from a security standpoint, this is really dangerous. AIs can hallucinate (which can be exploited) or straight up pull malicious information from the internet. Admittedly it’s a little impressive that this was created within a day, but I do question the quality. Like, why exactly does it need 3GB of space…?
Probably about the same danger involved as someone copy pasting random commands from a google search when they understand nothing about the commands they run.
I do dislike advice that boils down to “make ai do it lol”, especially if the solution is just a few commands.
MolochAlter@lemmy.world 4 weeks ago
From experience, LLM code is usually overly reliant on frameworks, because most modern code is, so you’ll get a lot of bloat if you don’t wrangle it in a different direction.