Login
You're viewing the assemblag.es public feed.
  • Oct 2, 2026, 8:36 PM

    Bluesky is having an odd little debate about use of the phrase "stochastic parrot" and how it appears to be the one phrase that just sends AI advocates into a rage, and I'm really torn, because while I'm in favor of making Ai advocates angry, I sincerely believe that parrots are capable of both understanding concepts and using language to intentionally communicate about their understandings of concepts. And I believe this is not true of LLMs

    💬 38🔄 45⭐ 0

Replies

  • Oct 2, 2026, 8:38 PM

    A character flaw I have is that I will attempt to fit the story about how one time I got to hear Irene Pepperberg give a talk on her research into conversations on subjects that are, admittedly, only tangentially related

    💬 2🔄 0⭐ 0
  • Oct 2, 2026, 8:40 PM

    Anyway I saw a post from Musktwitter where an AI advocate was arguing they can tell they're winning the debate because AI opponents stopped saying "stochastic parrot", and I'm like… is it possible the reason people stopped saying "stochastic parrot" is because "stochastic" is a specialist $5 word and also the phrase overall is incredibly clunky to vocalize

    "spicy autocomplete": cellar door
    "stochastic parrot": every syllable feels like I'm stumbling over it

    💬 4🔄 2⭐ 0
  • Oct 2, 2026, 8:43 PM

    Presence and absence. She claims to have verified in a lab setting that Alex understood the concepts of presence and absence

    💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Leelooleeloo@c.im
    Oct 3, 2026, 8:30 AM

    @onelson @mcc
    Meat proxy is not a term for the LLM, it refers to the human (meat) copy/pasting text back and forth (proxy) between the LLM and e.g. emails.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • fear is not a Weltanschauungwouldinotcallmyselfahumanbeing@zirk.us
    Oct 3, 2026, 1:53 AM

    @mcc an ability all the more impressive in parrots for the fact that no controlled study has been able to replicate the phenomenon in AI advocates

    💬 0🔄 1⭐ 0
  • 💬 0🔄 0⭐ 0
  • JayWhiteCatTamer@mastodon.online
    Oct 3, 2026, 2:50 AM

    @mcc “I won because you gave up” certainly fits with their understanding of, well, everything.

    💬 0🔄 0⭐ 0
  • Oct 2, 2026, 8:44 PM

    @mcc in a former life, I was at the MIT Media Lab while she was there working on developing a web browser for parrots.

    It was called InterPet Explorer.

    I also have a need to bring this up in unrelated conversations.

    💬 4🔄 3⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 1⭐ 7
  • 💬 0🔄 10⭐ 9
  • 💬 0🔄 0⭐ 0
  • Oct 2, 2026, 8:48 PM

    @mcc iirc Emily Bender has expressed regret over the name for that reason, coined (as frequently happens) not knowing it would stick so hard

    💬 1🔄 0⭐ 0
  • Oct 2, 2026, 8:52 PM

    @mcc fwiw while parrots are clearly bright and charming thinking beings with qualia interiority etc, scientists are still a bit conflicted over how similar / complex they are compared to us; as with chatbots their ability to mimic language really muddies our ability to accurately asses what's going on with them, because we hear one utter a complete sentence and our anthropomorphizing tendency tells us we're dealing with a human!

    💬 2🔄 0⭐ 0
  • Oct 2, 2026, 8:56 PM

    @mcc also there are other birds who are astonishingly good mimics who we've determined conclusively have way less going on cognitively than eg african gray parrots, they're just playing back little .flac files for us*, but it still gets us because language is so central to our perceptions of "intelligence".
    * not to say it doesn't "mean something" to them!

    💬 2🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Oct 2, 2026, 10:22 PM

    @jplebreton @mcc Gratuitous shout-out to the lyrebird. What a champ. Absolutely impressive (while utterly depressing to think about the implications of noise pollution and habitat loss that result in the stuff people seem to get most excited about)

    💬 0🔄 0⭐ 0
  • Oct 2, 2026, 9:02 PM

    @jplebreton so an important thing about Irene pepperberg's research is she concludes the reason every researcher before her found zero evidence of parrot cognition and she found significant evidence of parrot cognition is that other researchers were assuming parrot cognition would resemble human cognition whereas she was working from a starting point of analyzing parrot behavior in the wild and the specific way wild parrots seem to approach learning

    💬 4🔄 0⭐ 0
  • Oct 2, 2026, 9:04 PM

    @mcc I'm familiar! (Karla worked with her in the 00s.) Fascinating lovely animals.

    💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Oct 3, 2026, 7:39 AM

    @mcc @jplebreton I mean, that's IMHO one of the major flaws of most research into animals' inner lives. That, and the inverse.

    One particularly enraging study I read was about pain. Scientists electrocuted fish if they got near some place, and observed that when subjected to the electro shocks, the fish behaved much like humans would when subjected to pain. They also observed the fish avoided triggering the sensation.

    Result? "It's inconclusive whether fish experience pain"

    💬 1🔄 0⭐ 0
  • Oct 3, 2026, 7:41 AM

    @mcc @jplebreton I get that you can't project human experience onto animals. But it's also true that "it behaves as if, but inconclusive" is pretty much a psychopath's response to other people.

    So "cognition, but perhaps different" is a *massive* step, IMHO.

    💬 1🔄 0⭐ 0
  • Oct 3, 2026, 9:07 AM

    @jens @mcc @jplebreton I disagree, vigorously.

    This "don't anthropomorphise animals" always smelt like bullshit to me, including when we were indoctrinated with it at university. It's an expression of human supremacism, the religious belief that we're a breed apart from, and above, everything else.

    This is nonsense. We're mammals with opposable thumbs and an overgrown cerebellum. It does give us capabilities that are a quantum leap beyond all others, but even that doesn't make us a different order of being altogether.

    That said, I do agree with the stated premise, just in a way that ought to give its proponents apoplexy.
    Don't look at it in terms of condescendingly acknowledging that a few sparks of divine essence might accidentally have splashed outward and stained another species or two.
    Look at it as recognising feelings and experiences that other species and we share, to whatever extent.

    💬 1🔄 0⭐ 0
  • Oct 3, 2026, 9:25 AM

    @KatS That's pretty much what I mean.

    I think there's a grain of truth in not assuming other animals experience things just like us. But "exactly like us" and "nothing like us" is a false binary.

    We have all evolved to navigate the same world, sometimes the same niches, and we tend to broadly have the organs with the same functions, not to speak of the same DNA. It's highly unlikely that the "nothing like us" end of the spectrum is accurate.

    @mcc @jplebreton

    💬 0🔄 0⭐ 0
  • Oct 3, 2026, 9:11 AM

    @mcc @jplebreton That rhymes exactly with studies that "proved" autistic people don't have empathy, by testing specifically and exclusively for the ways in which allistic people express it.

    💬 0🔄 0⭐ 0
  • Oct 3, 2026, 9:48 AM

    @mcc

    From what I understand, parrots don't mimic other sounds in the wild ( unlike the Lyre Bird) ie. reproducing any sound and in particular human speech is only done in captivity.

    @jplebreton

    💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Oct 2, 2026, 9:25 PM

    @gparker

    Levar Burton waving dismissively: Alexa

    Levar Burton pointing approvingly: Alex

    💬 0🔄 0⭐ 0
  • Oct 2, 2026, 8:58 PM

    @mcc Stochastic mockingbird maybe? Feels like that might be even more offensive to AI boosters.

    💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Oct 2, 2026, 9:15 PM

    @Sassinake That may require a level of language and conceptual understanding skill that is not supported by the scientific literature on parrot cognition

    💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Oct 2, 2026, 9:18 PM

    @Sassinake The parrots have, however, been both vocal and vehement on the "want pistachios" issue

    💬 2🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Oct 2, 2026, 11:25 PM

    @mcc *Fits a log-log curve to Apollo the parrot’s performance*: the pistachio singularity is at hand

    💬 0🔄 0⭐ 0
  • Oct 2, 2026, 9:20 PM

    @mcc How about "large-context Markov model"? Maybe the Markovians would complain the way parrot-people do ;)

    But seriously, these days no one uses a "bare" LLM, and the harness/agents really do make a categorical difference.

    💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • GhostOnTheHalfShellGhostOnTheHalfShell@masto.ai
    Oct 2, 2026, 9:27 PM

    @mcc

    I might be tempted to redirect the conversation if I were to be found on a site like that which I won’t, to a discussion of Microsoft efforts to make sure that the sexually explicit material that can be generated with their systems is faithful to the prompt. Maybe they should be asked to install the value of paying people in Africa slave wages to experience on trauma. Let them work through that.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 1⭐ 0
  • Oct 2, 2026, 10:24 PM

    @mcc

    'I'm in favor of making Ai advocates angry,'

    this really only makes you look like you debate in bad faith. maybe get someone to proof read or edit your work before publishing?

    its a self own.

    💬 0🔄 0⭐ 0
  • Oct 2, 2026, 10:28 PM

    @mcc

    Agree. LLM is a modern day player piano with a bigger drum.

    💬 0🔄 0⭐ 0
  • Oct 2, 2026, 11:01 PM

    @mcc

    i am going to use the phrase 'stochastic parrot' in everything i write from now on.

    thank you!

    💬 1🔄 0⭐ 0
  • 💬 2🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Oct 2, 2026, 11:05 PM

    @mcc the things that come up that I apparently have to catch up on... to catch up, or not to catch up... Dead, that I'm so behind!

    💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Oct 2, 2026, 11:22 PM

    @lilyyllyyllyly If LLMs understand concepts and can communicate about those concepts, and merely use stochastic methods to go about it, then that is just a description of what AI advocates actually believe about their product

    💬 0🔄 0⭐ 0
  • Oct 2, 2026, 11:22 PM

    @mcc have I been misunderstanding the phrase?

    I've always read it as "parrot" being the verb (to mimic) and not like the actual animal.

    💬 1🔄 0⭐ 0
  • Oct 3, 2026, 7:45 AM

    @Plan_A_to_Y @mcc Same difference. The verb comes from a belief that parrots could never understand what they say.

    Thing is, if you keep a pet like that - even budgies - you will notice that they may mimic sounds, but they don't mimic all sounds randomly in all situations. Their mimicry expresses a mood, and you can trace that back to the circumstances under which they learned the sound.

    That does not mean they make sense of the words, but they do make sense.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Oct 2, 2026, 11:40 PM

    @mcc I recall Emily Bender saying the term comes from the idea of mindlessly parroting not parrots.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0