918
Ford had to hire back former engineers to fix mistakes made by its automated systems (www.theverge.com)
submitted 3 months ago by to c/nottheonion@lemmy.world
126 comments fedilink hide all child comments
 

“Mistakenly, we thought that by just introducing artificial intelligence and adjusting the design requirements that we had, that that would produce a high-quality product,” said Charles Poon, VP of vehicle hardware engineering, in a briefing this week with reporters.

you are viewing a single comment's thread
view the rest of the comments
[–] 13 points 3 months ago (4 children)

Yeah, I use it for some programming tasks as well. I’m sick and tired of telling it that it did something wrong or simply omitted something, only to have it apologize and offer to fix its own mistakes.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 17 points 3 months ago (2 children)

    The major problem is that the work of an LLM has a massive number of hidden 'assumptions' that you need to be aware of. If you don't already have a good working knowledge of the task you wont have an intuition about those assumptions. It's annoying.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 5 points 3 months ago

    Yeah, and for those who don't know, the rationalization output of the LLM is just so pursuasive. It sounds quietly confident and rattles off things that sound like real details.

    People are believing the LLM output over actual human experts and the human experts have to expend non-trivial effort trying to disprove an LLM output before they can get on with the business of doing it right.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 3 points 3 months ago

    Yeah, like cards being the best thing ever on front-end. If I want a basic layout I put "avoid cards" usually because then it'll use them sparingly. But yeah, it's just autocomplete so it's always going to get the lowest common denominator code based on what it decided to look at.

    I'm now looking into finding something faster than GPT/Claude where I can ping pong prompt faster and get what I want. Whatever it writes first is just a first draft anyway.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 7 points 3 months ago

    You should be glad it's apologizing. On the occasions I've used it to try to actually write code for me it's had a tendency to blame me for its mistakes.

    It writes a function that gets stuck in an infinitely recursive loop that never exits, I point it out and it's all "Aha! You've fallen for a classic recursion trap!" What do you mean I've fallen for it?

    Between those experiences and seeing the hot garbage some of my coworkers vibe coded, it was enough for me to relegate LLMs purely to the "ask questions that you would have searched for on StackOverflow" role. And it frustrates me that search was made so impotent that it's not a real option to avoid the LLM entirely. The multiple answers and perspectives on SO were often really valuable.

  • permalink
  • fedilink
  • source
  • parent