Dear world: if you’d listened to me and kept AI confined it to toys and experiments, none of this crap would be happening now. The lesson is to stop rushing every new tech into production before it’s proven.

you are viewing a single comment's thread
view the rest of the comments
[–] [S] 1 point 1 day ago (9 children)

Based on its actual performance, my opinion is that AI “works” in that it can indeed generate valid code that does more or less what is asked for, under the right conditions. But “works” isn’t good enough to make it the tool people make it out to be. The only place where it’s hypothetically fine to use something that generates insecure code is where security is already impossible. For example, retro systems — but even that is being severely overdone and the crapflood has made it impossible to do good things there.

  • source
  • parent
  • hideshow 9 child comments
  • [–] [S] 1 point 1 day ago (8 children)

    What I mean is, an RCE in a program for DOS is maybe not the biggest deal, because nobody in their right mind would think DOS could even be secure in the first place.

  • source
  • parent
  • hideshow 8 child comments
  • [–] [S] 1 point 1 day ago (7 children)

    But using it in, say, a AAA game or major office tool? Madness. You are setting up the users to get pwned and should be held liable for it.

  • source
  • parent
  • hideshow 7 child comments
  • [–] [S] 1 point 1 day ago (6 children)

    That’s the problem with those vibe coded Adobe clones. Not that they can’t work, of course they can, there’s no magic to Adobe apps, but there’s no way to secure all that generated code properly.

  • source
  • parent
  • hideshow 6 child comments
  • [–] [S] 1 point 1 day ago (5 children)

    Toys and experiments. That’s where AI code shines. Things where security isn’t even an afterthought. Things like devices that aren’t networked at all, or that matter so little as to make vulnerabilities irrelevant.

  • source
  • parent
  • hideshow 5 child comments
  • [–] [S] 1 point 1 day ago (4 children)

    And that thing where people use it to hack their smart TVs to stop them spying and showing ads? That’s pretty good too.

  • source
  • parent
  • hideshow 4 child comments
  • [–] [S] 1 point 1 day ago (3 children)

    I’m very carefully not saying AI doesn’t work or is useless. That would be untrue. What I’m getting at is that its utility is far less, and far more restricted, than boosters claim, simply because we live in a connected world where every RCE is someone’s whole life fust fucked with a spiked glove.

  • source
  • parent
  • hideshow 3 child comments
  • [–] [S] 1 point 1 day ago (2 children)

    Using it to, say, disable the networking on a smart appliance and force it to just work normally, that’d be fine. Using it to make little amusements for retro platforms would be fine if it weren’t being drastically overdone. Using it to bypass the security on ancient software for which no source code is available, thats another good use.

  • source
  • parent
  • hideshow 2 child comments
  • [–] [S] 1 point 1 day ago (1 child)

    When it comes to spyware infested listening devices like smart TVs, I say that fixing them to behave properly falls under “any means necessary” including AI.

  • source
  • parent
  • hideshow 1 child comment
  • [–] [S] 1 point 23 hours ago

    For me, I’m done using it for coding until and unless some things drastically change. But I also have an obligation to myself to stay informed about it, to know what it can and can’t do, because this is tech, where a month is a decade sometimes.

  • source
  • parent