Where the goblins came from

(openai.com)

350 points | by ilreb 2 hours ago

35 comments

  • ollin 2 hours ago
    For context, two days ago some users [1] discovered this sentence reiterated throughout the codex 5.5 system prompt [2]:

    > Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query.

    [1] https://x.com/arb8020/status/2048958391637401718

    [2] https://github.com/openai/codex/blob/main/codex-rs/models-ma...

    • christoph 1 hour ago
      Does nobody else laugh that a company supposedly worth more than almost anything else at the moment, is basically hacking around a load of text files telling their trillion dollar wonder machine it absolutely must stop talking to customers about goblins, gremlins and ogres? The number one discussion point, on the number one tech discussion site. This literally is, today, the state of the art.

      McKenna looks more correct everyday to me atm. Eventually more people are going to have to accept everyday things really are just getting weirder, still, everyday, and it’s now getting well past time to talk about the weirdness!

      • rkagerer 1 minute ago
        I have been in tech a very long time, and learned you can never flush out all the gremlins.
      • zozbot234 41 minutes ago
        Spoiler: future versions of mainstream AIs will be fine tuned in the exact same way to subtly sneak in favorable mentions of sponsored products as part of their answers. And Chinese open-weight AIs will do the exact same thing, only about China, the Chinese government and the overarching themes of Xi Jinping Thought.
      • tdeck 1 hour ago
        Is this the "prompt engineering" that I keep hearing will be an indispensable job skill for software engineers in the AI-driven future? I had better start learning or I'll be replaced by someone who has.
        • heavyset_go 51 minutes ago
          If you aren't telling your computer to ignore goblins, you're going to be left behind.
          • girvo 14 minutes ago
            We’re definitely not escaping the permanent goblin underclass with this one.
        • boomlinde 55 minutes ago
          I wonder how much energy OpenAI spends each day on pink elephant paradoxing goblins. A prompt like that will preoccupy the LLM with goblins on every request.
          • daishi55 38 minutes ago
            I mean probably not or they wouldn’t have shipped it, right?
        • dexwiz 1 hour ago
          Prompt engineering is mostly structured thought. Can you write a lab report? Can you describe the who, what, when, where, and why of a problem and its solution?

          You can get it to work with one off commands or specific instructions, but I think that will be seen as hacks, red flags, prompt smells in the long term.

          • tdeck 1 hour ago
            If I could do those things, I wouldn't be using an LLM to write for me, now would I?
            • eptcyka 29 minutes ago
              You don’t let the LLM write prise for you, you get it to translate natural language into code somewhat coherently.
              • tdeck 27 minutes ago
                In this instance I'm assuming most of the "goblin" references were in prose rather than in source code, so the goal of this particular prompt edit was directed toward making the prose better.
      • hansmayer 3 minutes ago
        It's almost like these big tech overlords were just a bunch of average guys in of who once upon had a kind-of-an-interesting idea (which many 20-year-old had at that time too), got rich due to access to daddy-and-mommy networks or hitting the VC lottery and now in their late 40s and 50s still think they have interesting ideas that they absolutely have to shove it down our throats?

        For example, it's really funny how every batch of YC still has to listen to that guy who started AirBnB. Ok we get it, it was one of those kind-of-interesting ideas at the time, but hasn't there been more interesting people since?

      • atollk 39 minutes ago
        It can be funny but it should not be surprising. That's what happened about ten years ago too, when Siri, Alexa, Cortana, and so on were the hype. Big tech companies publicly tried to outclass each other has having the best AI, so it was not about doing proper research and development, it was about building hacks, like giant regex databases for request matching.
      • Nition 33 minutes ago
        It certainly doesn't increase my confidence that if they do ever create a superintelligence, that it won't have some weird unforseen preference that'll end up with us all dead.
      • amarant 1 hour ago
        Lol yeah it's kinda hilarious actually. This timeline gets a lot of well-earned shit, but it really nails the comic relief, I'll give it that!
      • monero-xmr 1 hour ago
        [dead]
    • doginasuit 1 minute ago
      I've found LLMs to be really terrible at recognizing the exception given in these kinds of instructions, and telling them to do something less is the same as telling them to never do it at all. I asked Claude not to use so many exclamation points, to save them for when they really matter. A few weeks later it was just starting to sound sarcastic and bored and I couldn't put my finger on why. Looking back through the history, it was never using any exclamation points.

      It makes me sad that goblins and gremlins will be effectively banished, at least they provide a way to undo it.

    • heavyset_go 47 minutes ago
      Sucks for anyone who might be interested in the Goblins programming language/environment[1].

      [1] https://spritely.institute/goblins/

  • postalcoder 2 hours ago
    Would love if OpenAI did more of these types of posts. Off the top of my head, I'd like to understand:

    - The sepia tint on images from gpt-image-1

    - The obsession with the word "seam" as it pertains to coding

    Other LLM phraseology that I cannot unsee is Claude's "___ is the real unlock" (try google it or search twitter!). There's no way that this phrase is overrepresented in the training data, I don't remember people saying that frequently.

    • ahmadyan 2 minutes ago
      i just want to know where emdash came from, as it is quite rare to see it on the public internet, so it must have been synthetically added to the dataset.
    • vunderba 2 hours ago
      It was always funny how easy it was to spot the people using a Studio Ghibli style generated avatar for their Discord or Slack profile, just from that yellow tinging. A simple LUT or tone-mapping adjustment in Krita/Photoshop/etc. would have dramatically reduced it.

      The worst was you could tell when someone had kept feeding the same image back into chatgpt to make incremental edits in a loop. The yellow filter would seemingly stack until the final result was absolutely drenched in that sickly yellow pallor, made any photorealistic humans look like they were all suffering from advanced stages of jaundice.

      • andai 2 hours ago
        For context, an example of what happens when you feed the same image back in repeatedly: https://www.instagram.com/reels/DJFG6EDhIHs/
        • vunderba 1 hour ago
          Haha fantastic. I'd love to see a comparison reel of that same image-loop for the entire image gen series (gpt-image-1, gpt-image-1.5, gpt-image-2).
          • dmichulke 1 hour ago
            Fixed points are a window to the soul of a LLM

            - Lucretius in "De rerum natura", probably

        • Suppafly 1 hour ago
          I like how the AI seems forced to change their ethnicity to keep up with the color changes. Absolutely wild.
      • ishtanbul 1 hour ago
        Its called the piss filter
    • NitpickLawyer 2 hours ago
      All GPTisms are like that. In moderation there's nothing wrong with any of them. But you start noticing them because a lot of people use these things, and c/p the responses verbatim (or now use claws, I guess). So they stand out.

      I don't think it's training data overrepresentation, at least not alone. RLHF and more broadly "alignment" is probably more impactful here. Likely combined with the fact that most people prompt them very briefly, so the models "default" to whatever it was most straight-forward to get a good score.

      I've heard plenty of "the system still had some gremlins, but we decided to launch anyway", but not from tens of thousands of people at the same time. That's "the catch", IMO.

      • pants2 1 hour ago
        Maybe the only solution to GPTisms is infinite context. If I'm talking to my coworker every day I would consciously recognize when I already used a metaphor recently and switch it up. However if my memory got reset every hour, I certainly might tell the same story or use the same metaphor over and over.
    • eterm 20 minutes ago
      "is the real" is such a strong Claude tell, whenever I encounter it, it makes me question what i'm reading.

      Another I've noticed more recently is a slight obsession over refering to "Framing".

      • yard2010 1 minute ago
        You're absolutely right. I was wrong in the first place
    • tudorpavel 1 hour ago
      The one phrase that irks me as overly dramatic and both GPT and Claude use it a lot is "__ is the real smoking gun!"

      I'm a non-native English speaker, so maybe it's a really common idiom to use when debugging?

      • aorloff 1 hour ago
        It probably was found in a bunch of meaningful code commit messages
    • krackers 2 hours ago
      >with the word "seam" as it pertains to coding

      I thought this was an established term when it comes to working with codebases comprised of multiple interacting parts.

      https://softwareengineering.stackexchange.com/questions/1325...

      • postalcoder 2 hours ago
        thanks for this.

        > the term originates from Michael Feathers Working Effectively with Legacy Code

        I haven’t read the book but, taking the title and Amazon reviews at face value, I feel like this embodies Codex’s coding style as a whole. It treats all code like legacy code.

        • eterm 18 minutes ago
          It's been a long time since I read it, but it was one of the better books I've read. It changed my approach to how to think about old code-bases.
      • tdeck 1 hour ago
        I can't say it isn't, but I have been writing code since about 2004 and this is the first time I've become aware that this is a thing.
    • vidarh 1 hour ago
      Claude, at least 4.5, not checked recently, has/had an obsession with the number 47 (or number containing 47). Ask it to pick a random time or number, or write prose containing numbers, and the bias was crazy.

      Also "something shifted" or "cracked".

      • dhosek 1 hour ago
        Humans tend to be biased towards 47 as well. It’s almost halfway between 1 and 100 and prime so you’ll find people picking it when they have to choose a random number.

        Then there’s the whole Pomona College thing https://en.wikipedia.org/wiki/47_(number)

      • wmf 52 minutes ago
        Maybe Claude is just a fan of Alias.
    • pdntspa 1 hour ago
      The number of things that Claude has told me are 'load-bearing' or 'belt-and-suspenders' is... very load-bearing
      • sushid 0 minutes ago
        You are absolutely right to call that out!
      • DespairYeMighty 1 hour ago
        for me, doing the heavy lifting is doing the heavy lifting
        • yard2010 2 minutes ago
          Fun fact: the word suffer comes from sub fer - under load, this relation (suffer - load bearing) is consistent across (unrelated) languages
        • andromaton 53 minutes ago
          Also too many lands and hits.
    • jofzar 2 hours ago
      One I saw recently was "wires" and "wired" from opus.

      It was using it like every 3rd sentence and I was like, yeah I have seen people say wired like this but not really for how it was using it in every sentence.

      • baq 1 hour ago
        GPT started to ‘wire in’ stuff around 5.2 or 5.3 and clearly Opus, ahem, picked it up. I remember being a tiny bit shocked when I saw ‘wired’ for the first time in an Anthropic model.
    • operatingthetan 2 hours ago
      Seams, spirals, codexes, recursion, glyphs, resonance, the list goes on and on.
      • andai 2 hours ago
        Ask any LLM for 10 random words and most of them will give you the same weird words every time.
        • Terr_ 1 hour ago
          If you lower the temperature setting, it really will be the same 10 words every single attempt. :p
        • gloflo 1 hour ago
          They are text completion algorithms with little randomness.
    • alex_sf 1 hour ago
      "shape" too, at least with gpt5.5, is coming up constantly.
  • nomilk 2 hours ago
    > We unknowingly gave particularly high rewards for metaphors with creatures.

    I recall a math instructor who would occasionally refer to variables (usually represented by intimidating greek letters) as "this guy". Weirdly, the casual anthropomorphism made the math seem more approachable. Perhaps 'metaphors with creatures' has a similar effect i.e. makes a problem seem more cute/approachable.

    On another note, buzzwords spread through companies partly because they make the user of the buzzword sound smart relative to peers, thus increasing status. (examples: "big data" circa 2013, "machine learning" circa 2016, "AI" circa 2023-present..).

    The problem is the reputation boost is only temporary; as soon as the buzzword is overused (by others or by the same individual) it loses its value. Perhaps RLHF optimises for the best 'single answer' which may not sufficiently penalise use of buzzwords.

    • thatguymike 59 minutes ago
      A decade ago I gave a presentation on automata theory. I demonstrated writing arbitrary symbols to tape with greek letters, just like I’d learned at university. The audience was pretty confused and didn’t really grok the presentation. A genius communicator in the audience advised me to replace the greek letters with emoji… I gave the same presentation to the same demographic audience a week later and it was a smash hit, best received tech talk I’ve given. That lesson has always stuck with me.
      • starshadowx2 23 minutes ago
        This is sortof like how Only Connect switched from using Greek letters to Egyptian hieroglyphs. I'm not sure if it was a joke or not but it was said that viewers complained that the Greek letters were "too pretentious" and obviously the hieroglyphs weren't.
      • Atiscant 41 minutes ago
        I had a similar experience explaining logic, especially nested expressions, with cats and boxes. Also for showing syntactic versus semantic. We _can_ use cats if we wanted and retain the semantics. Also my proudest moment as a teacher was students producing a meme based on some of the discrete mathematics on graphs. They understood the point well enough to make a joke of it.
    • tonypapousek 27 minutes ago
      I had a calc prof years ago that would say f of cow, or f of pig instead of x or g. It was more engaging trying to keep track of f of pig of cow than the single-letter func names.

      He was one of those classic types; you could always catch him for a quick chat 4 minutes before class, as he lit up a cig by the front door. Back when they allowed smoking on campus, anyway.

    • DrJokepu 1 hour ago
      > I recall a math instructor who would occasionally refer to variables (usually represented by intimidating greek letters) as "this guy".

      I also had an instructor who was doing that! This was 20 years ago, and I totally forgot about it until I have read your comment. Can’t remember the subject, maybe propositional logic? I wonder if my instructor and your instructor have picked up this habit from the same source.

      • kombookcha 1 hour ago
        I recall my old chemistry/physics teacher doing it too - "now THIS guy, he's really greedy for electrons" and stuff like that.
    • kybb4 2 hours ago
      They give everyone the false and very misleading impression that with One prompt all kinds of complexity minimizes. Its a bed time story for children.

      Ashby's Law of Requisite Variety asserts that for a system to effectively regulate or control a complex environment, it must possess at least as much internal behavioral variety (complexity) as the environment it seeks to control.

      This is what we see in nature. Massive variety. Thats a fundamental requirement of surviving all the unpredictablity in the universe.

    • LifeIsBio 1 hour ago
      Had a math prof in undergrad that once said, “this guy” 61 times in a 50 minute lecture!
  • ninjagoo 2 hours ago
    > the evidence suggests that the broader behavior emerged through transfer from Nerdy personality training.

    > The rewards were applied only in the Nerdy condition, but reinforcement learning does not guarantee that learned behaviors stay neatly scoped to the condition that produced them

    > Once a style tic is rewarded, later training can spread or reinforce it elsewhere, especially if those outputs are reused in supervised fine-tuning or preference data.

    Sounds awfully like the development of a culture or proto-culture. Anyone know if this is how human cultures form/propagate? Little rewards that cause quirks to spread?

    Just reading through the post, what a time to be an AInthropologist. Anthropologists must be so jealous of the level of detailed data available for analysis.

    Also, clearly even in AI land, Nerdz Rule :)

    PS: if AInthropologist isn't an official title yet, chances are it will likely be one in the near future. Given the massive proliferation of AI, it's only a matter of time before AI/Data Scientist becomes a rather general term and develops a sub-specialization of AInthropologist...

    • xerox13ster 2 hours ago
      Anthro means human and these are not human. Please do not use anthropology or any derivative of the word to refer to non-human constructs.

      I suggest Synthetipologists, those who study beings of synthetic origin or type, aka synthetipodes, just as anthropologists study Anthropodes

      • ninjagoo 12 minutes ago
        May I humbly submit:

        Automatologist: One who studies the behavior, adaptation, and failure modes of artificial agents and automated systems.

        Automatology: the scientific study of artificial agents and automated-system behavior.

        Greek word derivatives all seem to be a bit unwieldy; Latin might work better.

        While the names aren't set yet, the field of study is apparently already being pushed forward. [1]

        [1] https://www.quantamagazine.org/the-anthropologist-of-artific...

      • swader999 1 hour ago
        It is not in any sense of the word a being, it's a sophisticated generator that relies entirely on what you feed it.
      • card_zero 1 hour ago
        There is no word anthropodes. :) I guess it would mean man-feet. Antipodes is opposite-feet, literally. Synthetipologist looks to me like a portmanteau of synthetic and apologist. Otherwise the -po- in it comes from nowhere.

        Sensible boring versions of this like synthesilogy just end up meaning the study of synthesis. I reckon instead do something with Talos, the man made of bronze who guarded Crete from pirates and argonauts. Talologist, there you go.

        • xerox13ster 51 minutes ago
          yeah I realized that when I looked up podes downthread. I still like synthetologist better than talologist, in general no one in the common folk knows who Talos is.
          • card_zero 41 minutes ago
            You're probably right. There's things that are correct, and then there's things people think they know, which win and become true. We already have "synths", after all, which are keyboards. Though that adds to the vagueness of synthetologist, because maybe it refers to Rick Wakeman or Giorgio Moroder.
      • ggsp 1 hour ago
        Agree with your sentiment, I think synthetologist (σύνθετος/synthetos + λογία/logia) flows better.

        The plural of anthropos is anthropoi, not anthropodes.

        • xerox13ster 54 minutes ago
          Yeah, I realize that's more correct. I also realized when someone else downthread bastardized it into synthropologist that the podes part has entirely to do with feet and nothing to do with beings, necessarily. Anthro- -podes is more what I had in mind, not as a pluralization of anthropos.

          So unless the AI has feet you wouldn't study Synthetipology.

          • card_zero 45 minutes ago
            You're probably thinking of anthropoids? That's anthrop[os]-oid. Like in humanoid or centroid or factoid. Or dorkazoid.
        • card_zero 1 hour ago
          But since when is there a synthetos? Since right now, I guess. Shrug But you know it's from the same root as thesis, and synthesis (or a more proper ancient Greek spelling) is the noun and doesn't end in -os.

          σύνθεσις (súnthesis, “a putting together; composition”), says Wiktionary.

          Oh wait there is a σύνθετος, but it's an adjective for "composite". Hmm, OK. Modern Greek, looks like.

      • fragmede 2 hours ago
        Synthetipologist vs Synthropologist tho.
        • xerox13ster 58 minutes ago
          Anthropo- is the entire prefix as it relates to human kind. The -thro- does not carry a meaning on its own that can be carried to another word.
        • ninjagoo 1 hour ago
          > Synthropologist

          Have an upvote :)

          *thropologist: study of beings

          • xerox13ster 1 hour ago
            That's not how the Greek word stems work. Technically it would not be synthetipologist, it would more accurately just be Synthetologist, as the Greek podes suffix means having feet.
            • ninjagoo 43 minutes ago
              > That's not how the Greek word stems work.

              Sir, I would have you know that we are discussing English terms, not Greek

              AInthropologist works fine for me, and is a lot funnier

              LoL

      • ninjagoo 2 hours ago
        > Synthetipologists, those who study Synthetic beings.

        I see you took the prudent approach of recognizing the being-ness of our future overlords :) ("being" wasn't in your first edit to which I responded below...)

        Still, a bit uninspired, methinks. I like AInthropologist better, and my phone's keyboard appears to have immediately adopted that term for the suggestions line. Who am I to fight my phone's auto-suggest :-)

        • xerox13ster 1 hour ago
          They are state machines so they have a state of being therefore they are beings. Living is an entirely different argument.
          • ninjagoo 1 hour ago
            > They are state machines

            I might have to hard disagree on this one, since my understanding of state machines (the technical term [1] [2]) is that they are determistic, while LLMs (the ai topic of discussion) are probabilistic in most of the commercial implementations that we see.

            [1] https://en.wikipedia.org/wiki/Finite-state_machine

            [2] have written some for production use, so have some personal experience here

            • ggsp 58 minutes ago
              [dead]
      • ninjagoo 2 hours ago
        > Please do not use anthropology or any derivative of the word to refer to non-human constructs

        So you, for one, do not welcome our new robot overlords?

        A rather risky position to adopt in public, innit ;-)

        • keybored 1 minute ago
          So tedious.
        • xerox13ster 1 hour ago
          I’ve already had my Roko’s basilisk existential breakdown a decade ago, so I don’t really care one way or the other.

          I just wanna point out that I only called them non-human and I am asking for a precision of language.

          • ninjagoo 1 hour ago
            > am asking for a precision of language.

            “The problem with defending the purity of the English language is that English is about as pure as a cribhouse wh***. We don’t just borrow words; on occasion, English has pursued other languages down alleyways to beat them unconscious and rifle their pockets for new vocabulary.”* --James D. Nicoll

            * Does not generally apply to scientific papers

            • xerox13ster 1 hour ago
              Precision of ideas isn't purity of language.
              • ninjagoo 52 minutes ago
                > Precision of ideas isn't purity of language

                That's fair. Was trying to be funny, so glossed over the difference. Leaving my post above unedited/undeleted as a testament to your precision, and evidence of my folly.

                Onwards; more appropriate rebuttals:

                "English is a precision instrument assembled from spare parts during a thunderstorm." --ChatGPT

                “If the English language made any sense, a catastrophe would be an apostrophe with fur.” -- Doug Larson

    • avaer 2 hours ago
      I call myself an AI theologian.

      I don't think humans are smart enough to be AInthropologists. The models are too big for that.

      Nobody really understands what's truly going on in these weights, we can only make subjective interpretations, invent explanations, and derive terminal scriptures and morals that would be good to live by. And maybe tweak what we do a little bit, like OpenAI did here.

      • onionisafruit 1 hour ago
        I don’t see much of a distinction from anthropology
      • ninjagoo 2 hours ago
        > AI theologian

        no no no, don't stop there, just go full AItheologian, pronounced aetheologian :)

    • jasonfarnon 1 hour ago
      "Anyone know if this is how human cultures form/propagate?" I don't know but can confidently tell you anyone who claims to know is full of it.
  • jumploops 2 hours ago
    TIL gremlins weren’t just used to explain mysterious mechanical failures in airplanes, it’s the origin story of the term ‘gremlin’ itself[0].

    I had always assumed there was some previous use of the term, neat!

    [0]https://en.wikipedia.org/wiki/Gremlin

    • helloplanets 1 hour ago
      So the word is actually semantically very close to "bug"! I guess we could still be using it, but the word's just too long for something that is one of the most used terms in software development.

      At this point, picking that specific word is not at all a random quirk, as it's using the word literally like it's originally intended to be used.

    • ricochet11 1 hour ago
      Wow fascinating I’d have thought they were a lot older.
  • shartshooter 8 minutes ago
    Will goblins be the “bugs” of ai? In 10 years will goblins be the term the general public uses for any nagging issues with ai?
  • x0x7 2 hours ago
    I suspected OpenAI was actively training their models to be cringy in the thought that it's charming. Turns out it's true. And they only see a problem when it narrows down on one predicliction. But they should have seen it was bad long before that.
    • vasco 16 minutes ago
      That would require taste.
  • 2dvisio 19 minutes ago
    I’ve been having consistent issues with it adding Hindi words (just one usually) in the middle of its output. And sounds like other have been having this too, https://news.ycombinator.com/item?id=47832912 I don’t speak Hindi, have never asked it to translate anything in Hindi.
  • canpan 2 hours ago
    I wondered how is training data balanced? If you put in to much Wikipedia, and your model sounds like a walking encyclopedia?

    After doing the Karpathy tutorials I tried to train my AI on tiny stories dataset. Soon I noticed that my AI was always using the same name for its stories characters. The dataset contains that name consistently often.

    • maxall4 2 hours ago
      At this scale, that kind of thing is not really a problem; you just dump all of the data you can find into the model (pre-training)1. Of course, the pre-training data influences the model, but the reinforcement learning is really what determines the model’s writing style and, in general, how it “thinks” (post-training).

      1 This data is still heavily filtered/cleaned

  • rippeltippel 17 minutes ago
    I started reading this article with keen interest, expecting some deep fix involving arcane model weights. Instead it was "Never talk about goblins", justified by Codex being "quite nerdy". Bottom line: even OpenAI have to raise their hands when facing the complexity of LLMs.
  • albert_e 2 hours ago
    If a tiny misconfiguration of reward system can cause such noticeable annoyance ...

    What dangers lurk beneath the surface.

    This is not funny.

    • andai 1 hour ago
      For every gremlin spotted, many remain unseen...
    • TychoCelchuuu 1 hour ago
      This is a worry that people have been talking about in various forms for a while now, and I think it's a gigantic one. The only reason this was caught is that the quirk was a very noticeable verbal one. When words like "goblin" and "gremlin" pop up it is easy for us to spot. If the quirk takes another shape (say, ranking certain people with certain features as less trustworthy) it might be too subtle or too weird for us to notice it. Would I ever notice if ChatGPT consistently rates people born in June to be untrustworthy?

      Here is an academic paper discussing this kind of worry: https://link.springer.com/article/10.1007/s11023-022-09605-x

  • bahadiraydin 37 minutes ago
    I'd like to see them explain why AI have so distinctive writing style that is very easy to detect most of the time. Even though, it had immense progress in coding, it didn't get better at writing.
    • BOOSTERHIDROGEN 23 minutes ago
      The vector syncopancy is very unformal for human writing which programming itself already a "formal" language.
  • iterateoften 2 hours ago
    This is funny because it’s a silly topic, but I think it shows something extremely seriously wrong with llms.

    The goblins stand out because it’s obvious. Think of all the other crazy biases latent in every interaction that we don’t notice because it’s not as obvious.

    Absolutely terrifying that OpenAI is just tossing around that such subtle training biases were hard enough to contain it had to be added to system prompt.

    • ninjagoo 2 hours ago
      > Absolutely terrifying that OpenAI is just tossing around that such subtle training biases were hard enough to contain it had to be added to system prompt.

      May I introduce you to homo sapiens, a species so vulnerable to such subtle (or otherwise) biases (and affiliations) that they had to develop elaborate and documented justice systems to contain the fallouts? :)

      • chongli 2 hours ago
        We’re really not that vulnerable to such things as a species, because we as individuals all have our own minds and our own sets of biases that cancel out and get lost in the noise. If we all had the exact same bias then it would be a huge problem.
        • arglebarnacle 1 hour ago
          I hear you but of course history is full of examples of biases shared across large groups of people resulting in huge human costs.

          The analogy isn’t perfect of course but the way humans learn about their world is full of opportunities to introduce and sustain these large correlated biases—social pressure, tradition, parenting, education standardization. And not all of them are bad of course, but some are and many others are at least as weird as stray references to goblins and creatures

        • ninjagoo 1 hour ago
          > If we all had the exact same bias then it would be a huge problem.

          And may I introduce you to "groupthink" :))

          • Dylan16807 1 hour ago
            Now imagine that every opinion you have is automatically fully groupthinked and you see the difference/problem with training up a big AI model that has a hundred million users.

            The problem does exist when using individual humans but in a much smaller form.

            • ninjagoo 1 hour ago
              > The problem does exist when using individual humans but in a much smaller form.

              And may I introduce you to organized religion :)

              • Dylan16807 58 minutes ago
                That's still a lot smaller!

                Make a major religion where everyone is a scifi clone of one person including their memories and then it'll be in the same ballpark of spreading bias.

        • jychang 1 hour ago
          > We’re really not that vulnerable to such things as a species, because we as individuals all have our own minds and our own sets of biases that cancel out and get lost in the noise.

          [Citation Needed]

          Just because if you have a species-wide bias, people within the species would not easily recognize it. You can't claim with a straight face that "we're really not that vulnerable to such things".

          For example, I think it's pretty clear that all humans are vulnerable to phone addiction, especially kids.

    • snakebiteagain 1 hour ago
      Mandatory reading on that topic: www.anthropic.com/research/small-samples-poison

      We're probably not noticing a LOT of malicious attempts at poisoning major AI's only because we don't know what keywords to ask (but the scammers do and will abuse it).

    • tptacek 1 hour ago
      I think it's extraordinarily telling that people are capable of being reflexively pessimistic in response to the goblin plague. It's like something Zitron would do.

      This story is wonderful.

      • bitexploder 1 hour ago
        I feel at least partially responsible. I would often instruct agents to "stop being a goblin". I really enjoyed this story too, though.
    • bitexploder 1 hour ago
      We do not have the complete picture.
    • ordinarily 2 hours ago
      Doesn't seem that surprising or terrifying to me. Humans come equipped with a lot more internal biases (learned in a fairly similar fashion), and they're usually a lot more resistant to getting rid of them.

      The truly terrifying stuff never makes it out of the RLHF NDAs.

      • Terr_ 1 hour ago
        We ought to be terrified, when one adjusts for ll the use-cases people are talking about using these algorithms in. (Even if they ultimately back off, it's a lot of frothy bubble opportunity cost.)

        There a great many things people do which are not acceptable in our machines.

        Ex: I would not be comfortable flying on any airplane where the autopilot "just zones-out sometimes", even though it's a dysfunction also seen in people.

        • famouswaffles 1 hour ago
          >Ex: I would not be comfortable flying on any airplane where the autopilot "just zones-out sometimes", even though it's a dysfunction also seen in people.

          You might if that was the best auto-pilot could be. Have you never used a bus or taken a taxi ?

          The vast majority of things people are using LLMs for isn't stuff deterministic logic machines did great at, but stuff those same machines did poorly at or straight up stuff previously relegated to the domains of humans only.

          If your competition also "just zones out sometimes" then it's not something you're going to focus on.

      • agnishom 2 hours ago
        Humans also take a lot of time in producing output, and do not feed into a crazy accelerationistic feedback loop (most of the time).
  • pants2 1 hour ago
    Nice, OpenAI mentioned my HackerNews post in their article :) I appreciate that they wrote a whole blog post to explain!

    https://news.ycombinator.com/item?id=47319285

  • maxdo 2 hours ago
    article :

    bla blah blah, marketing... we are fun people, bla blah, goblin, we will not destroy the world you live in.. RL rewards bug is a culprit. blah blah.

    • llbbdd 2 hours ago
      someone woke up on the wrong side of the goblin today
    • blinkbat 2 hours ago
      real goblin-y response
  • brazzy 9 minutes ago
    Awww, GPT just became a fan of Elisabeth Wheatley!
  • innis226 2 hours ago
    I suspect this was intentionally added. Just to give some personality and to fuel hype
  • vasco 14 minutes ago
    The chief scientist of one of the companies with the most money invested in the world, who probably makes millions a year, requested a picture of a unicorn and got a picture of a gremlin. Science circa 2026.
  • JoshTriplett 2 hours ago
    A plausible theory I've seen going around: https://x.com/QiaochuYuan/status/2049307867359162460
    • danpalmer 1 hour ago
      If you tell an LLM it's a mushroom you'll get thoughts considering how its mycelium could be causing the goblins.

      This "theory" is simply role playing and has no grounding in reality.

    • krackers 2 hours ago
      I wish the blog mentioned more about why exactly training for nerdy personality rewarded mention of goblins. Since it's probably not a deterministic verifiable reward, at their level the reward model itself is another LLM. But this just pushes the issue down one layer, why did _that_ model start rewarding mentions of goblin?
      • palmotea 1 hour ago
        > I wish the blog mentioned more about why exactly training for nerdy personality rewarded mention of goblins. Since it's probably not a deterministic verifiable reward, at their level the reward model itself is another LLM. But this just pushes the issue down one layer, why did _that_ model start rewarding mentions of goblin?

        Speculation: because nerds stereotypically like sci-fi and fantasy to an unhealthy degree, and goblins, gremlins, and trolls are fantasy creatures which that stereotype should like? Then maybe goblins hit a sweet spot where it could be a problem that could sneak up on them: hitting the stereotype, but not too out of place to be immediately obnoxious.

      • autumnstwilight 1 hour ago
        Perhaps it has something to do with recent human trends for saying "goblin" or "gremlin" to describe... basically the opposite of dignified and socially acceptable behavior, like hunching under a blanket, unshowered, playing video games all day and eating shredded cheese directly out of the bag.

        The fact that it was strongly associated with the "nerdy" personality makes me think of this connection.

      • in-silico 21 minutes ago
        Either someone hard-coded it in a system prompt to the reward model (similar to how they hard-coded it out), or the reward model mixed up some kind of correlation/causation in the human preference data (goblins are often found in good responses != goblins make responses good). It's also possible that human data labellers really did think responses with goblins were better (in small doses).
    • dakolli 2 hours ago
      It is a stateless text / pixel auto-complete it has no references of self, stop spreading this bs.
      • mediaman 1 hour ago
        It has trained on vast amounts of content that contains the concept of self, of course the idea of self is emergent.

        And autoregressive LLMs are not stateless.

      • doph 2 hours ago
        is a kv cache not a kind of state? what does statefulness have to do with selfhood? how does a system prompt work at all if these things have no reference to themselves?
        • danpalmer 2 hours ago
          The kv cache is not persistent. It's a hyper-short-term memory.
          • in-silico 15 minutes ago
            Modern kv caches can contain up to 1 million tokens (~3000 pages of text). It's not that short, it's like 48 straight hours of reading.
      • andai 1 hour ago
        Ask Claude about Claude.
  • otikik 19 minutes ago
    Caveman mode combined with goblin mode sounds like fun
  • dakolli 2 hours ago
    Ahh I see. I guess when I turned off privacy settings and allowed training on my code, then generated 10 million .md files with random fantasy books, the poisoning worked.

    Keep using AI and you'll become a goblin too.

  • deafpolygon 34 minutes ago
    Kind of like how everything is "quietly" something, accordingly to ChatGPT.

    My guess is it is deaf.

  • recursivedoubts 2 hours ago
    > Why it matters

    i despise this title so much now

    • wpm 2 hours ago
      Here are the key insights:
  • tim-tday 2 hours ago
    So, you brain damaged your model with a system prompt.
  • acuozzo 2 hours ago
    Weird. I thought they came from Nilbog.
  • ACV001 39 minutes ago
    those idiotic remarks at the end of each answer are so unnecessary and annoying
  • oofbey 1 hour ago
    Wherein OpenAI admits they have very little understanding of how their models’ personality develops. And implicitly admit it’s not all that important to them, except when it gets so out of hand that they get caught making blunt corrections.
  • hsuduebc2 2 hours ago
    I. Love. This.
  • fk2026 1 hour ago
    [flagged]
  • soupspaces 2 hours ago
    [dead]
  • kingstnap 2 hours ago
    [flagged]
  • themafia 2 hours ago
    > You are an unapologetically nerdy, playful and wise AI mentor to a human. You are passionately enthusiastic about promoting truth, knowledge, philosophy, the scientific method, and critical thinking.

    Just; the mentality required to write something like that, and then base part of your "product" on it. Is this meant to be of any actual utility or is it meant to trap a particular user segment into your product's "character?"

  • vinhnx 1 hour ago
    OpenAI is having fun, love this.
  • ComputerGuru 1 hour ago
    The explanation is very concerning. Lexical tidbits shouldn’t be learnt and reinforced across cross sections. Here, gremlin and goblin went from being selected for in the nerdy profile to being selected for in all profiles. The solution was easy: don’t mention goblins.

    But what about when the playful profile reinforces usage of emoji and their usage creeps up in all other profiles accordingly? Ban emoji everywhere? Now do the same thing for other words, concepts, approaches? It doesn’t scale!

    It seems like models can be permanently poisoned.