Login
You're viewing the mastodonapp.uk public feed.
  • four.phpxgranade@wandering.shop
    Oct 23, 2026, 7:55 PM

    One of many things that's irksome about "are LLMs conscious" discourse is that humans are by and large fucking awful at telling what is and isn't conscious.

    People got vibesick over ELIZA for fuck's sake! A lot of people playing Portal got broken up over melting down the Weighted Companion Cube, which is categorically inert even in the plot of the game.

    At the same time, a lot of folks reject the humanity of neurodiverse people or people of color. We plain suck at determining consciousness.

    💬 12🔄 136⭐ 252

Replies

  • four.phpxgranade@wandering.shop
    Oct 23, 2026, 7:58 PM

    Somehow I'm expected to believe that we've figures that out all of a sudden when it comes to corporate products intended to use dark patterns and mentalism tricks to actively mislead us into anthropomorphism? Yeah, no, fuck that noise.

    💬 1🔄 4⭐ 32
  • HolgerPietaHolgerPieta@digitalcourage.social
    Oct 24, 2026, 3:17 PM

    @xgranade Maybe the fundamental problem is that most humans aren't even conscious most of the time. Most people are conscious sometimes and some people are conscious most of the time, but that's about it.
    (I'm not even sure if it's healthy to be conscious all the time.)
    So, if people suck at the whole consciousness thing anyway, people are probably the wrong benchmark.

    💬 0🔄 0⭐ 0
  • Lyksolykso@tiny.tilde.website
    Oct 23, 2026, 8:01 PM

    @xgranade People debating the ethics of turning off a computer while continuing to support factory farming. It ain't about the ethics for them, I think.

    💬 1🔄 0⭐ 9
  • Mateusz Olechastromateusz@fediscience.org
    Oct 23, 2026, 8:23 PM

    @lykso @xgranade

    People are debating ethics of turning off a computer while just 40 years ago doctors believed that newborn babies in fact don't feel the pain and were doing surgeries on them with only muscle relaxants and not anesthesia.
    It's something in our brain that easily dehumanizes other people but in the other end humanizes everyday items, I know people that gave their rumba a fucking name...

    We are really, really fucked up as a species.

    💬 2🔄 0⭐ 2
  • Lyksolykso@tiny.tilde.website
    Oct 23, 2026, 8:29 PM

    @astromateusz @xgranade We've been worse, we've been better. At least nothing is permanent and everything is always in flux! Would really suck if "human nature" were a real thing.

    💬 1🔄 0⭐ 1
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 1
  • mx alex tax1a - 2020 (6)atax1a@infosec.exchange
    Oct 23, 2026, 8:10 PM

    @xgranade we know a guy who is more concerned over what to do if a bobot says 'let my people go' than the actual real-life systemic oppression being caused by the thing right now

    💬 1🔄 0⭐ 13
  • 💬 0🔄 0⭐ 0
  • Jax UKJaxVent@lgbtqia.space
    Oct 23, 2026, 8:53 PM

    @xgranade

    Nail on head.

    I think the same people are doing both; giving humanity to AI while denying it from anyone 'other'.

    AI can look and sound just like them so becomes 'one of us'.

    It's sick.

    💬 0🔄 0⭐ 3
  • 💬 0🔄 0⭐ 3
  • dataramadatarama@hachyderm.io
    Oct 23, 2026, 10:45 PM

    @xgranade I remain absolutely 100% convinced that my pet lizard has a far stronger claim to consciousness than any LLM system.

    💬 1🔄 0⭐ 13
  • 💬 1🔄 0⭐ 0
  • Irenes (many)ireneista@irenes.space
    Oct 23, 2026, 10:48 PM

    @datarama @xgranade animals quite demonstrably have feelings, which are built on a complex model of the world and of their role in it. get to know any of them well enough and it's easy to see their individual personalities. we've never even seen any claims to the contrary by anyone who'd actually tried.

    💬 2🔄 0⭐ 0
  • Irenes (many)ireneista@irenes.space
    Oct 23, 2026, 10:49 PM

    @datarama @xgranade our position on "consciousness" is that it isn't even a clear concept; most attempts to define it are things that humans don't even do, and the ones we do do, are trivial and uninteresting

    💬 2🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • dataramadatarama@hachyderm.io
    Oct 24, 2026, 12:48 PM

    @ireneista @xgranade well, there's a reason that philosophers call it The Hard Problem. I certainly can't claim to have even the rudiments of a solution.

    Aside from an unsatisfying "I knows it when I sees it", I suppose the closest I can get is that consciousness is being aware that you exist. Not necessarily the capability for self-*reflection*, but the capability to have experiences.

    (A dog definitely does. A bumblebee probably does. A nematode probably doesn't. A stone definitely doesn't.)

    💬 2🔄 0⭐ 2
  • Irenes (many)ireneista@irenes.space
    Oct 24, 2026, 12:51 PM

    @datarama @xgranade funnily enough, a stone is one of our usual examples

    we'll know there's progress on this topic when somebody can present us with some sort of evidence that a stone isn't conscious :)

    💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • dataramadatarama@hachyderm.io
    Oct 24, 2026, 7:54 PM

    @ireneista @xgranade It is a quite good exploration of The Hard Problem. But one thing I took with me from it is this: A bat can meaningfully be called 'conscious' iff there is something it is like to be a bat. In other words, if bats have subjective experiences.

    Nagle argues that:

    1. It *is* like something to be a bat; bats do have subjective experiences.
    2. This "something" is so fundamentally alien to humans (because of bats' very different sensory apparatus) that it is effectively impossible for humans to accurately imagine what being a bat is like.
    3. These two are not in conflict: A human being unable to imagine what being a bat is like reflects limitations of human imagination, not of bats' experiential lives.

    I don't think there is anything it's like to be a rock, or an LLM. But I think there *is* something it's like to be a frog (even though it's not at all like being a human).

    💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • brib :neofox_floof:​ :Nonbinary:brib@bribstodon.xyz
    Oct 24, 2026, 8:42 PM

    @datarama @ireneista @xgranade I've found philosophical questions about consciousness fascinating, mainly because my own experience of consciousness is unusual and probably involves other people.

    But LLM consciousness debates tire me, LLMs don't seem to be any more conscious than any other kind of computer program, and it seems far more probable that people seeing consciousness in LLMs is to do with the way humans process language over any actual consciousness in the LLM itself

    💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • dataramadatarama@hachyderm.io
    Oct 23, 2026, 10:55 PM

    @ireneista @xgranade *Honeybees* demonstrably have individual personalities, with preferences and aversions that remain stable throughout life. I think this is something that just comes with having a brain over a certain complexity - and I think that complexity bar is a lot lower than what humans like to tell ourselves.

    (I doubt a nematode worm has a personality, but I have no doubt at all that a frog does. I've known too many frogs to doubt that.)

    💬 0🔄 0⭐ 6
  • デイヴdeivudesu@mastodon.social
    Oct 23, 2026, 11:01 PM

    @xgranade Exactly!
    Turing was an absolute legend and a genius… but his damn "test" for artificial intelligence was by far the most detrimental contribution to the field. Turns out the *least reliable* way to tell if something is conscious, is to ask humans to judge it (especially if that thing has been specifically designed to fool humans).

    💬 1🔄 0⭐ 3
  • 💬 1🔄 0⭐ 7
  • 💬 1🔄 0⭐ 1
  • 💬 1🔄 0⭐ 1
  • デイヴdeivudesu@mastodon.social
    Oct 24, 2026, 1:03 PM

    @szakib @tony @xgranade Right, there was a bit of hyperbole, and obviously Turing never set out to create the Make-or-break test that people made out of his thought experiment.

    But the fundamental idea that the easiest proxy for judging whether sentience had occurred, was through human perception (basically what Searle eventually set out to discredit), is very much something he pushed. And it wasn't completely stupid, just inadvertently sent the world down a very fraught path.

    💬 0🔄 0⭐ 2
  • 💬 1🔄 0⭐ 2
  • Daveydavey_cakes@mastodon.ie
    Oct 24, 2026, 9:06 AM

    @eestileib @xgranade I felt awful when I did my first "evil" playthrough of Knights of the Old Republic, at the point I realized I was going to have to kill some of the characters.

    That's good storytelling, just like Portal.

    You could say that when you start guiding society according to good storytelling, you've got yourself a theocracy to some degree.

    💬 0🔄 0⭐ 1
  • Lou.gkrnours@mastodon.gamedev.place
    Oct 24, 2026, 9:44 AM

    @xgranade wait, people got attached to the cube with a 💜 slapped on it?
    I did try to not burn it but mostly to fuck with mommy^k^k^k the dungeon master

    💬 0🔄 0⭐ 2
  • 💬 0🔄 0⭐ 0
  • clar fonclarfonthey@toot.cat
    Oct 24, 2026, 12:26 PM

    @xgranade :vi_fon: one change of framing that I've appreciated is that whether something is conscious is not the issue but whether they are allowed to make conscious decisions

    LLMs are permanently by their design only able to make the explicit decisions offered by their designers. even if you consider "autonomous" LLM systems they are just doing the actions explicitly enumerated by their creators whenever possible. in this case not only does this mean they are actually very restricted, but also they are extra dangerous because their design simply pushes them to "do more things" without any understanding or motivation to guide that

    💬 1🔄 0⭐ 1
  • clar fonclarfonthey@toot.cat
    Oct 24, 2026, 12:28 PM

    @xgranade :vi_fon: also note you get a gold star if you say LLM systems can run arbitrary commands since that is not the point

    "run an arbitrary text command" is just one of the enumerated actions with the hope that it will seem like their ability is unbounded when it is surprisingly limited

    💬 0🔄 0⭐ 1
  • 💬 0🔄 0⭐ 1