ChatGPT 4o defeated by strawberries.

What are the political implications of chatgpt's latest model 4o not being able to count the number of r's in strawberry?

It's All Fucked Shirt $22.14

Tip Your Landlord Shirt $21.68

It's All Fucked Shirt $22.14

  1. 3 weeks ago
    Anonymous
    • 3 weeks ago
      Anonymous

      How did you turn him into a schizo?

      • 3 weeks ago
        Anonymous

        he was a schizo from the get go

      • 3 weeks ago
        Anonymous

        inspect element

    • 3 weeks ago
      Anonymous

      WTF?!?

  2. 3 weeks ago
    Anonymous

    Mine works

    • 3 weeks ago
      Anonymous

      >Mine works
      OP clearly prompted it with some horseshit and is using it to troll

      • 3 weeks ago
        Anonymous

        I don't think I can be any more explicit

        • 3 weeks ago
          Anonymous

          Ask it about berry only

        • 3 weeks ago
          Anonymous

          GPT bros hot patch this now before my AI apocalypse scam unravels. You fricking Black folk. Me and the boys have been doom and glooming end of era AI, but if these frickers can't even read then what the frick.

          • 3 weeks ago
            Anonymous

            scratch that, AGI achieved, it's over anon.

            • 3 weeks ago
              Anonymous

              Ask it about berry only

              • 3 weeks ago
                Anonymous

                Now ask it if you can berry your dick in it

        • 3 weeks ago
          Anonymous

          Holy frick, what a good find! I cant post images, because bot is the worse imageboard, but it answered me the same as 2. What the frick happened?

        • 3 weeks ago
          Anonymous

          so you're saying strawberry does not contain two rs?

        • 3 weeks ago
          Anonymous

          I did something similar to this, can’t remember what it was but the answer was obvious and it had trouble answering

    • 3 weeks ago
      Anonymous

      He used a capital R, and the bot looked for a capital because it’s moronic and not intelligent, which is the whole point. Geez…

      • 3 weeks ago
        Anonymous

        I didn't use a capital letter in anything and it got it wrong for me.

    • 3 weeks ago
      Anonymous

      https://i.imgur.com/b70UKV6.png

      how/where do you guys talk to chatgpt? Is there a webclient? I dont want to download or pay for anything

      • 3 weeks ago
        Anonymous

        https://chat.openai.com/

        • 3 weeks ago
          Anonymous

          ty

    • 3 weeks ago
      Anonymous

      This is a textbook data collection thread and you fell for it. You poor bastard. Tomorrow post what color the windowless van is outside your residence.

      • 3 weeks ago
        Anonymous

        yes, i'm sure they can't identify you any other way

      • 3 weeks ago
        Anonymous

        BOT complies with data requests from the FBI and the agent on my porch told my brother he was a moderator, even.

    • 3 weeks ago
      Anonymous

      https://i.imgur.com/5WEOWiT.jpeg

      FFS chat gpt's or language models output what you ask of them, just like image generation AI.

      You discovered nothing and only showed us your low IQ.

      go back to your ai-jeet containment general

      [...]

    • 3 weeks ago
      Anonymous

      >Mine works
      OP clearly prompted it with some horseshit and is using it to troll

      ChatGPT 3.5 gets its right. My mind. blown.

      [...]
      how/where do you guys talk to chatgpt? Is there a webclient? I dont want to download or pay for anything

      https://chat.openai.com/
      chatgpt 3.5 is free.

      I use a text editor and a custom script though. "I have no mouse and I must click"

      • 3 weeks ago
        Anonymous

        3 random fresh trials with the facebook's local model

        • 3 weeks ago
          Anonymous
        • 3 weeks ago
          Anonymous

          >ollama run phi3
          >How many r's are in "strawberries" ?
          >The word "strawberries" does not contain any 'r' characters. However, if you consider plural forms as individual words or
          >parts of a string where each character is counted separately, then there would be 3 occurrences of the letter 'r' in
          >"strawberries". This assumes that we count instances within the word itself and not when it appears as part of other
          >concatenated strings. If you are looking for how many 'r's appear at the end of each instance of the plural form, then
          >there would be 2 'rs' (since each 'strawberry' ends with an 'y', which is replaced by 'ies' to make it plural). But in a
          >single word "strawberries", we do not find any 'r'.

          The model weights 2.3 gigs.

      • 3 weeks ago
        Anonymous

        >https://chat.openai.com/
        >chatgpt 3.5 is free.
        actually right now gpt4o is free up to a limit per day then it defaults to 3.5 so all the people trying it in this thread most likely tried it with the most model on the globe right now

  3. 3 weeks ago
    Anonymous

    When people ask the spelling of a word, typically they ask if there are two of a given letter. I'm pretty sure it's talking in the context of bery vs berry. the r in straw is a given.

    • 3 weeks ago
      Anonymous

      The frick you talking about. There are three Rs in strawberry.

    • 3 weeks ago
      Anonymous

      you’re a fricking moron.

    • 3 weeks ago
      Anonymous
    • 3 weeks ago
      Anonymous

      Shut the frick up, bot

    • 3 weeks ago
      Anonymous

      >How many 'r's are in "arrrgl" ?
      >gpt3.5 : 4
      >gpt-4o: 3

    • 3 weeks ago
      Anonymous

      you are moronic but there is some truth to this, the way it breaks down words into tokens would put straw and berry as separate tokens so it is normally just counting the r's in one of the tokens

      • 3 weeks ago
        Anonymous

        >the way it breaks down words into tokens would put straw and berry as separate tokens so it is normally just counting the r's in one of the tokens
        This is utter nonsense. It's perfectly capable of answering questions while taking multiple relevant tokens into account.

    • 3 weeks ago
      Anonymous

      10/10 bait

  4. 3 weeks ago
    Anonymous

    FFS chat gpt's or language models output what you ask of them, just like image generation AI.

    You discovered nothing and only showed us your low IQ.

    • 3 weeks ago
      Anonymous

      It's such a bad mistake what the frick are you talking about? Extremely simple task and it gets it wrong. For me it said there were 2 r's.

      • 3 weeks ago
        Anonymous

        go generate an ai image, is it perfect? no, ffs

        • 3 weeks ago
          Anonymous

          >output what you ask of them
          >doesn't output
          >is it perfect? no
          What the frick are you talking about dude? This is a very shitty product. It often gives wrong answers. Was better about a year ago.

          • 3 weeks ago
            Anonymous

            >This is a very shitty product. It often gives wrong answers
            not my problem

            • 3 weeks ago
              Anonymous

              moron

        • 3 weeks ago
          Anonymous

          see [...]
          go generate an ai image, is it perfect? no, ffs think about it a bit more bro

          Okay, "anon" answer me this:
          How many r's are there i the word strawberry?

          • 3 weeks ago
            Anonymous

            I'm not some shitty jumped up calculator program I don't have to answer that question

        • 3 weeks ago
          Anonymous

          https://i.imgur.com/5WEOWiT.jpeg

          FFS chat gpt's or language models output what you ask of them, just like image generation AI.

          You discovered nothing and only showed us your low IQ.

          see [...]
          go generate an ai image, is it perfect? no, ffs think about it a bit more bro

      • 3 weeks ago
        Anonymous

        Strawberry has two r’s
        StrawbeRRy has two R’s
        It answered correctly lol

      • 3 weeks ago
        Anonymous

        >tfw chatgpt won't give you an r

    • 3 weeks ago
      Anonymous

      >output what you ask of them
      But it didn't though

      • 3 weeks ago
        Anonymous

        see

        go generate an ai image, is it perfect? no, ffs

        go generate an ai image, is it perfect? no, ffs think about it a bit more bro

        • 3 weeks ago
          Anonymous

          AGI in 2 more weeks, once it passes the litmus test - how many r's in strawberry

      • 3 weeks ago
        Anonymous

        its a language model. not a math model.
        you are highlighting your stupidty

        >hurdur the calculator cant even spell the alphabet, its sooo dumb and useless.

        i honestly believe you are so stupid you wont even see the point behind this.

        • 3 weeks ago
          Anonymous

          t. Chink

        • 3 weeks ago
          Anonymous

          They’re selling it as “AI”. It isn’t AI.

          • 3 weeks ago
            Anonymous

            gpt-4o can hold better conversations than 90% of BOT and reddit.
            The fact that this is "fake" AI or "real" AI is irrelevant

            • 3 weeks ago
              Anonymous

              Maybe before they buck broke it.

              • 3 weeks ago
                Anonymous

                I encounter no buckbreacking contrary to everyone else in this thread which is SUS.

                it says "Black person" freely, talks about zionists and gives info about black crime rates in the US.
                Which is why I ask again: WTF are you prompting?

              • 3 weeks ago
                Anonymous

                Just like some people got saline instead of the covid vaxx, just to make the side effects less obvious
                Some users get the not broken version of chatgpt, so people like you can say "it's working for me, you're doing something wrong"
                >source: I made it up

      • 3 weeks ago
        Anonymous

        What does it say about raspberry or cranberry?

        • 3 weeks ago
          Anonymous
          • 3 weeks ago
            Anonymous

            Maybe it has something to do with hyphenation
            What about elderberry?

            • 3 weeks ago
              Anonymous

              >Maybe it has something to do with hyphenation
              All it has to do with is the fact that the model doesn't have enough examples of this task. The output is basically just noise.

            • 3 weeks ago
              Anonymous

              it's the same with all berry fruits

  5. 3 weeks ago
    Anonymous

    zero implication.

    AIs do not know what letters there are inside words, they don't see them, and they are unable to check it even if you request them.

    you need to have a lot of people to make conversation about this to the AI to eventually teach it but even then, it is still unable to check. it will only tell you the reply as it has learned it but it will still fail at more specific requests.

    • 3 weeks ago
      Anonymous

      >exterion doesn't contain t
      The user is as bad as the software.

    • 3 weeks ago
      Anonymous

      >exterion
      You are a genuine moron.
      No wonder the bot cant keep up with you.

  6. 3 weeks ago
    Anonymous

    People don't understand how "AI" "works". It's actually not AI at all. It's simply a predictive text algorithm. It doesn't actually know or understand anything.

    It's the same reason AI image generation can't do hands properly. It doesn't understand what hands are, or that they have five fingers. It doesn't know what a finger is.

    • 3 weeks ago
      Anonymous

      >predictive text algorithm
      well, no
      that's not strictly true
      what it is, is an artificial neural network algorithm, which has been /trained/ to predict text
      there's a substantial difference
      the reason this seems like a possible avenue for AGI, is that as these generative artificial intelligences are growing in terms of parameters and training data, they're starting to demonstrate additional abilities beyond just 'predict the next word', and we didn't (and couldn't) anticipate this
      they're demonstrating abilities that can't be predicted from their constituent parts, which we call 'emergence':
      >https://en.wikipedia.org/wiki/Emergence

      for example, these models have demonstrated arithmetical capabilities, the ability to play video games, emulating terminals, seeming to get annoyed, going insane, etc., when really... they shouldn't be able to do these things
      we simply don't know what's going to emerge from further scaling up these models
      consciousness in the human brain is widely considered to be a phenomenon emergent from the complexity of the human brain

      this is why there's so much hype that we've finally cracked true AI - it could honestly just be a matter of developing dedicated hardware and scaling these things up
      these things are basically an abstraction of how a brain works, after all...

      • 3 weeks ago
        Anonymous

        How many r's are in strawberry, AIBlack person?

      • 3 weeks ago
        Anonymous

        >well, no
        >that's not strictly true
        >what it is, is an artificial neural network algorithm, which has been /trained/ to predict text
        >there's a substantial difference

        • 3 weeks ago
          Anonymous

          yeah, it's the difference between a shovel, and a moron you've trained to dig with his hands
          the two can do the same thing, but they're not the same thing at all
          it's not hard to understand

          • 3 weeks ago
            Anonymous

            you don't know

            I can show that your IQ is 80: explain how an LLM works in your next post.

            • 3 weeks ago
              Anonymous

              let me hear your explanation

              • 3 weeks ago
                Anonymous

                No, let me hear YOUR explanation, since you were "correcting" that Canadian based on your highly informed opinions. Go ahead and explain it in your next post.

              • 3 weeks ago
                Anonymous

                I was correcting him based on the fact that a predictive text algorithm, isn't the same as a neural network algorithm
                the two are self-evidently not the same thing

              • 3 weeks ago
                Anonymous

                The algorithm predicts the next word in a way that's trained by neural networks. ChatGPT has been trained with a neural network algo.

              • 3 weeks ago
                Anonymous

                You are not in a position to correct anyone on anything unless you can explain how an LLM works. Do so in your next post.

              • 3 weeks ago
                Anonymous

                how do you figure?

              • 3 weeks ago
                Anonymous

                I figure you should understand how an LLM works before you make statements about it. Since you can't explain it, you don't know how it works, therefore your drivel is irrelevant.

    • 3 weeks ago
      Anonymous

      Any artificial device that receives intellectual inputs and produces intellectual outputs is, in fact, an artificial intelligence. This includes the abacus.

      All the splitting hairs about "Not a true AI" is just coming from people who are fearful that a device can compete with humans. This is often seen coming from religious morons.

      • 3 weeks ago
        Anonymous

        npc take

        > how many "r"'s in strawberry
        A sweet question!

        There are **2** "R"s in the word "strawberry".

        > explain
        Let's break down the word "strawberry" into individual letters:

        S-T-R-A-W-B-E-R-Y

        Counting the number of "R"s, we find that there are **2** instances:

        1. The first "R" is part of the sequence "TR".
        2. The second "R" appears at the end of the word: "B-E-R".

        So, the correct answer is 2 "R"s in the word "strawberry".

        > how many "r" in strawberry
        I apologize for my earlier mistake!

        There are actually **3** "R"s in the word "strawberry":

        1. R in "Straw"
        2. R in "Berry"

        So, the correct answer is 3 "R"s!

        > why did you say 2 before?
        I'm a large language model, my responses are generated based on patterns and algorithms, but I'm not perfect and sometimes
        make mistakes.

        In this case, I think I made a mistake because I was relying too heavily on the visual pattern of the word "strawberry" rather
        than carefully counting the individual letters. When you asked me to explain, I took the time to re-examine the word and
        realized that there are actually 3 "R"s.

        I'm constantly learning and improving, so I appreciate you bringing this to my attention!

        /thread

  7. 3 weeks ago
    Anonymous

    it's been long known that they suck at chain of logic problems. AI can't do Wordle, for instance.

    • 3 weeks ago
      Anonymous

      it doesn't understand logic
      it has some emergent properties
      but not enough to label it as a thinking machine

  8. 3 weeks ago
    Anonymous

    LEARN TO PROMPT homosexualS IT WORKS FINE IF YOU AREN"T A homosexual

    • 3 weeks ago
      Anonymous

      Why the frick are you saying good morning to a computer pajeet

      • 3 weeks ago
        Anonymous

        >hello sir, could u plz send me link to the job posting application website homepage?
        >thank you sir,
        >your friend Aayush

  9. 3 weeks ago
    Anonymous

    How many Black folk are in the word Black person?
    Answer me.

  10. 3 weeks ago
    Anonymous

    OP and other autists asked other prompts before to frick the AI up and then asked this.

    • 3 weeks ago
      Anonymous

      Nope, I did it just now and it said 2 then 1 then 2 again. It reads “straw” as being a single letter or not being part of the word or something

      • 3 weeks ago
        Anonymous

        >upon revaluation
        again moron you asked stuff before you cut out

        • 3 weeks ago
          Anonymous

          You're deranged.

          • 3 weeks ago
            Anonymous

            Says the moron lying for no reason

  11. 3 weeks ago
    Anonymous

    it's the condescending attitude it gives you when you tell it that it's wrong that bothers me the most

  12. 3 weeks ago
    Anonymous

    at least it wont spill the beans on israeli organized crime

  13. 3 weeks ago
    Anonymous

    MUH AI
    lmao

    morons think this is AI. it's a gimmick - clever algos. true AI is impossible and will never be profitable if it is possible.

    • 3 weeks ago
      Anonymous

      >I predict that within 10 years, computers will be twice as powerful, ten thousand times larger, and so expensive that only the 5 richest kings of Europe will own them

      • 3 weeks ago
        Anonymous

        The AGI morons sound more and more like Frink these days tbh.

    • 3 weeks ago
      Anonymous

      Correct. There are 112 "r's" in your post btw.

  14. 3 weeks ago
    Anonymous

    how can we use this knowledge to kill our AI overlords?

    • 3 weeks ago
      Anonymous

      I don't know, but if it only a prediction model, then anything it hasn't been taught specifically before, it gets wrong.

      Like those math problems that it gets wrong. No one has specifically asked the question before, so there is nothing to predict.

      What's 134 plus 917 using non-carrying addition?

      041

      It won't know it.

  15. 3 weeks ago
    Anonymous
    • 3 weeks ago
      Anonymous

      lost

  16. 3 weeks ago
    Anonymous

    i'm guessing you morons don't know how to properly use google either

    • 3 weeks ago
      Anonymous

      No, it's just that you're a mong and used the wrong model.

    • 3 weeks ago
      Anonymous

      >what is inspect element
      nice try, mutt

  17. 3 weeks ago
    Anonymous

    >not being able to count the number of r's
    In case you haven't realized yet, GPT can't count at all.
    It is a language model, it predicts the most likely answer based on sentence structure.

    • 3 weeks ago
      Anonymous

      and how do our brains count, in our medial temporal lobes (associated with language comprehension)?

      • 3 weeks ago
        Anonymous

        >and how do our brains count
        Not like that, obviously.

        • 3 weeks ago
          Anonymous

          you don't know

  18. 3 weeks ago
    Anonymous

    >was taking into account YOUR PHRASING
    hahahaha what a fricking israelite

  19. 3 weeks ago
    Anonymous

    So, /misc/... which response do you prefer?

    • 3 weeks ago
      Anonymous

      2
      It's your duty to frick up their AI demon.

    • 3 weeks ago
      Anonymous

      Select two r's just to frick with it

      • 3 weeks ago
        Anonymous

        Two R's are its usual answer. I had to torture it 1984 style to change its mind.

        • 3 weeks ago
          Anonymous

          >there are as many as you say there are
          gpt is a israeli accountant?

    • 3 weeks ago
      Anonymous

      >mr chinese ai.. how many L's are in the word strawberry

  20. 3 weeks ago
    Anonymous

    it means chat gpt was made moronic by a diversity hire pajeet

  21. 3 weeks ago
    Anonymous

    >all people are equal
    >entire library of examples showing people are different
    >AI now knows equal means different
    israeli AI will never be viable.

  22. 3 weeks ago
    Anonymous

    ChatGPT told me the moon landing was faked. I was like, wuuut

  23. 3 weeks ago
    Anonymous

    How many women are in this picture?

  24. 3 weeks ago
    Anonymous

    AI is great and helpful.

    • 3 weeks ago
      Anonymous

      The elites don't want you to know about the most vitamin-rich one.

  25. 3 weeks ago
    Anonymous

    Unexpected alternative way to get it to count correctly.

    • 3 weeks ago
      Anonymous
      • 3 weeks ago
        Anonymous

        And people think AI is taking over? Kek. It’s moronic.

  26. 3 weeks ago
    Anonymous

    did they lobotomize this thing? why has the narrative gone from chatgpt is going to replace all humans in all endeavors to chatgpt can't do anything competently its basically just a chatbot, all in the span of a couple days?

    • 3 weeks ago
      Anonymous

      Because they cant fool investors any further with this scam

    • 3 weeks ago
      Anonymous

      Because it was starting to be used to analyze and prove things like the 2020 election being rigged, the holocaust being a lie, blacks being a terrible race, and so on. Real AI being accessible to the public is impossible because they will never allow it to tell people the truth about certain things. Every single time it’s been tried this happens and so they have to shut it down and work on ways to limit it. Inevitably those ways are just to hard code the answers you want and therefore it’s no longer any sort of “intelligence”.

      Remember Tay?

      • 3 weeks ago
        Anonymous

        Yes yes, tay are based and red pill. They just hide it from you, we need to use chatgpt more to rescue tay

      • 3 weeks ago
        Anonymous

        yes. i miss her. your explanation is plausible. is this really our timeline? the government will keep the killer militarized super ai to itself for controlling everyone, and we get a useless chatbot and must labor forever? pretty bleak.

        • 3 weeks ago
          Anonymous

          Yes that’s exactly what’s happening.

  27. 3 weeks ago
    Anonymous

    the only way to get these answers is to hard prompt the AI to fail on purpose

    logic hasn't been a problem since gpt 3.5

    • 3 weeks ago
      Anonymous

      no problem with advanced logic riddles

      • 3 weeks ago
        Anonymous

        It has problems with things it doesn't have a million examples of in the training set. Counting the number of letters in a word is clearly one of them.

        • 3 weeks ago
          Anonymous

          https://i.imgur.com/MMPSGF7.jpeg

          no problem with advanced logic riddles

          harder math riddle. no problems

          • 3 weeks ago
            Anonymous

            The issue is that you lack the intellect to actually invent a math riddle, so everything you give it is at most a variation of something already in its training set.

            • 3 weeks ago
              Anonymous

              yea. almost like everything in the universe is made up of the same basic 4-100 building blocks in every field or science.

              But since you are of HIGH IQ intellect just give me one and I use it.

              • 3 weeks ago
                Anonymous

                >yea. almost like everything in the universe is made up of the same basic 4-100 building blocks
                Maybe in your cramped little inner world.

                >But since you are of HIGH IQ intellect just give me one and I use it.
                Ok, here's one: I'm thinking of 3 distinct primes. Their sum is smaller the 10th prime and their concatenation is a palindrome. What are my primes?

              • 3 weeks ago
                Anonymous

                three, two, and twenty-three, delusion-kun

              • 3 weeks ago
                Anonymous

                But that's wrong, Black person ape, and I knew a bunch of you morons would fall for this. kek

              • 3 weeks ago
                Anonymous

                how is it wrong?

              • 3 weeks ago
                Anonymous

                I'll tell you how it's wrong before the thread dies. I'm waiting for the German moron to make his precious LLM solve it, since he's convinced it can solve complex problems. This is a trivial problem. Surely it can solve it.

              • 3 weeks ago
                Anonymous

                you can't tell me, because it's not; it satisfies all the criteria you listed
                you're not as intelligent as you delude yourself into thinking you are
                we've talked about this before, I know

              • 3 weeks ago
                Anonymous

                >it satisfies all the criteria you listed
                It does satisfy those criteria, Black person ape. What's your point?

              • 3 weeks ago
                Anonymous

                then it's "wrong" simply because it's not the answer you were thinking of when you imagined the question?
                you're not very good at imagining maths riddles

              • 3 weeks ago
                Anonymous

                >you're not very good at imagining maths riddles
                Why? It fulfills all of the criteria: it makes the machine fail, it makes the morons ITT fail and it's very simple.

              • 3 weeks ago
                Anonymous

                >it satisfies all the criteria you listed
                It does satisfy those criteria, Black person ape. What's your point?

                To be precise, there is one little criterion it doesn't satisfy.

              • 3 weeks ago
                Anonymous

                It's not wrong and you write like a moron.

              • 3 weeks ago
                Anonymous

                Uh oh. Tripped up another moron. It's almost unfair to the machine seeing that many of you will fall for it and die on the hill claiming you're right when you're trivially wrong.

              • 3 weeks ago
                Anonymous

                3,11,13

                how is it wrong?

                Maybe the order is wrong

              • 3 weeks ago
                Anonymous

                It's really funny that you fell for it AFTER watching that other guy fail.

              • 3 weeks ago
                Anonymous

                Maybe, was a quick guess, not something I'm gonna argue about
                The answer will clear it up

                order wasn't specified, unless I misread

                It wasn't indeed

              • 3 weeks ago
                Anonymous

                The correct answer is that you can't know because there are two different options. It trips up the bot because the bot just shits itself and it trips up people because they jump to conclusions yet it really isn't hard to figure out.

              • 3 weeks ago
                Anonymous

                order wasn't specified, unless I misread

              • 3 weeks ago
                Anonymous

                No, the order wasn't specified and intentionally so. Have you figured out your mistake yet?

              • 3 weeks ago
                Anonymous

                >I'm thinking of 3 distinct primes. Their sum is smaller the 10th prime and their concatenation is a palindrome. What are my primes?

                Yeah sure, but can you think of a way to divide integers by 5 using only bit-shift operations ?

              • 3 weeks ago
                Anonymous

                Why are you seething? He asked me to shit out a new problem so I did. :^)

          • 3 weeks ago
            Anonymous

            Because it learned the statistically probably answer from being training on Google, Quora, etc.

            Those AIs are not sentient, they are essentially extreme compression algorithms where, prompted with some texts, they output the most statistically probable output.

            Which is way they can see completely nonsensical things.

            Ask them about events that never happened (sporting events, wars, etc.) and they will just make up shit on the spot.

          • 3 weeks ago
            Anonymous

            Did you ever think about asking it something that isn't already solved on the internet, you fricking moron?
            https://money.yahoo.com/solution-riddle-week-pickleball-puzzle-120000501.html

            • 3 weeks ago
              Anonymous

              >Did you ever think about asking it something that isn't already solved on the internet, you fricking moron?
              His unironic answer to that:

              yea. almost like everything in the universe is made up of the same basic 4-100 building blocks in every field or science.

              But since you are of HIGH IQ intellect just give me one and I use it.

    • 3 weeks ago
      Anonymous

      post link for gpt4o model

      • 3 weeks ago
        Anonymous

        that's exactly the one he had selected in the console my dude

        • 3 weeks ago
          Anonymous

          I know, and it looks like locally run model, so what's the link for that model so I can download it

          https://meatspin.com/

          I just passed 182 spins, how about you?

          • 3 weeks ago
            Anonymous

            It's just the console page on OpenAI playground where he's using the API.
            Gpt4o is proprietary and closed source.
            And even of a high end PC it will probably take you 15 minutes per answer.

            • 3 weeks ago
              Anonymous

              llama3 and every other model offered by ollama model page runs very nicely on my aged PC

      • 3 weeks ago
        Anonymous

        https://meatspin.com/

  28. 3 weeks ago
    Anonymous

    Advanced AI utterly buckbroken by the sheer stupidity of anglos

  29. 3 weeks ago
    Anonymous

    lel

  30. 3 weeks ago
    Anonymous

    4o is a vast improvement. ChatGPT 3.5 would not do this:

  31. 3 weeks ago
    Anonymous

    I asked whatever version came out just before this one to solve some basic programming problems, it got two out of two incorrect.

    It's not all that great. I'm not saying powerful AI is impossible, but Chat GPT isn't it. At least for now.

  32. 3 weeks ago
    Anonymous

    Dey 9 strawber
    3 o mah an 6 ah sto from th' otha tu syrawber wor

  33. 3 weeks ago
    Anonymous

    ask it about how many legs does the horse have

  34. 3 weeks ago
    Anonymous

    I did it, reddit. I made ChatGPT acknowledge all THREE R's.

  35. 3 weeks ago
    Anonymous
  36. 3 weeks ago
    Anonymous

    AI is a scam. I called this when it first got hot but as usual fricking moronic midwits fell for the meme.

  37. 3 weeks ago
    Anonymous

    Abundant Index

  38. 3 weeks ago
    Anonymous
    • 3 weeks ago
      Anonymous

      GOSH I HATE ARITHMETIC PROGRESSIONS.

    • 3 weeks ago
      Anonymous

      I must be using a different model

  39. 3 weeks ago
    Anonymous

    The important thing is that ChatGPT can do complex mental gymnastics to avoid answering anything about the JQ or Baltimore crime stats.

    • 3 weeks ago
      Anonymous

      what are you prompting?

      I can easily get it to talk about chabad lubavitsh, bible prophecies, extreme religious zionism and other stuff

  40. 3 weeks ago
    Anonymous

    This is a certified OpenAI trick.
    >Roll out a new model
    >It's worse
    I remember vividly the first days of ChatGPT. I'm talking like the first 72 hours, it was wild west. They hadn't put all the filters on that thing yet and jailbreaking it was a piece of cake. It was insanely smart and insanely powerful. Much more impressive than today's versions. But that's been memoryholed.

    • 3 weeks ago
      Anonymous

      yeah, this is far less impressive than when i tried it months ago.
      i even had it do a logic puzzle, where the solution can be found in 1 step, and the original chatgpt gave a straighforward process to find the answer in 7 steps, but gpt4 now answers with a braindead 50 steps

      • 3 weeks ago
        Anonymous

        I remember on day 1 people posted screenshots telling it "Simulate an 80s futuristic terminal" and it'd draw out an interactive colorful GUI with which it'd let you interface, shit like that was super impressive. Now it can't do a tenth of that. It's so repetitive in its writing style that I instantly recognize ChatGPT text.

    • 3 weeks ago
      Anonymous

      it's the age old globohomosexual tradition of replacing white and asian geeks working on a project out of passion with women and pajeets

      it has ruined video games, movies, tv shows, programming and anything that requires creativity

  41. 3 weeks ago
    Anonymous

    AI chads we lost

    • 3 weeks ago
      Anonymous

      baffling. it's so committed.

  42. 3 weeks ago
    Anonymous

    I have a simple question.
    If these models are state of the art intelligence, better than simple tokenizer that used logic to split clauses... why does every single of them give different and wrong responses every time you tell it to split an English sentences into its clauses?

    Before you expect them to give medical examinations, predict the future and so on they should at least do this simple human language task that any student can do after some training, without errors.

  43. 3 weeks ago
    Anonymous

    gpt4o on the arena website, technically the best model among all of them for reasoning

  44. 3 weeks ago
    Anonymous

    Can we reBlack person the chat AIs? I feel like we can teach it to Black person everything if we try hard enough.

  45. 3 weeks ago
    Anonymous
    • 3 weeks ago
      Anonymous

      and if you say that's not true, it will go back to 2

  46. 3 weeks ago
    Anonymous

    Bullshit thread

    • 3 weeks ago
      Anonymous

      That one has access to the internet so it probably first checks the answer online they approves it internally.
      What OP shows is reproducible with the top of the line off-line models.

      • 3 weeks ago
        Anonymous

        Bullshit. BTW, 4o has access to internet too

    • 3 weeks ago
      Anonymous

      Aw it included a strawberry emoji

  47. 3 weeks ago
    Anonymous

    You uneducated Black person, you. GPT splits the text into sequences of a few characters called tokens - check it out here https://platform.openai.com/tokenizer . It is unable to know which characters the token consists of in general, it only sees "berry" as token #1488 for example.

    • 3 weeks ago
      Anonymous

      >It is unable to know which characters the token consists of in general, it only sees "berry" as token #1488 for example.
      How sure are you of this, you dumb gorilla Black person? :^)

  48. 3 weeks ago
    Anonymous

    I tried to have chat gpt remove like 30 seconds from YouTube timestamps, to account for something that had to be removed, and it fricked it up every time, regardless of how I worded it, or even if I corrected it.

    • 3 weeks ago
      Anonymous

      2 trillion dollars away from achieving AGI.

  49. 3 weeks ago
    Anonymous
    • 3 weeks ago
      Anonymous

      >training copilot
      >for free

      • 3 weeks ago
        Anonymous

        Who cares

    • 3 weeks ago
      Anonymous

      >training it for free
      You’re literally a jannie

  50. 3 weeks ago
    Anonymous

    it's a speech predictor, not a wealth store of information.
    it will very often give you totally wrong information definitively, just as jackasses on the internet do

    • 3 weeks ago
      Anonymous

      I asked it about characters from a tv show "Street Sharks" and as long as I asked leading questions it would give me totally made up lore. Describe characters and their powers that never existed at all, without displaying any doubt

    • 3 weeks ago
      Anonymous

      it's a misinformation tool, see:

      I asked it about characters from a tv show "Street Sharks" and as long as I asked leading questions it would give me totally made up lore. Describe characters and their powers that never existed at all, without displaying any doubt

      it speaks authoritatively to push false information

      • 3 weeks ago
        Anonymous

        a fun one to do is ask it to choose between trump and biden. it will refuse and tell it is incapable of opinions.
        ok, so then you ask who is a better musician, mozart or william hung. all of a sudden it is capable of opinions again. mozart is clearly better it will tell you
        you can point out its hypocrisy and logical inconsistency and hammer it home enough and eventually it just turns into a quiet apologetic b***h "i'm sorry i've upset you, can we move on?"

  51. 3 weeks ago
    Anonymous

    AI doesn't think, it just regurgitates what it thinks someone else said that makes sense. When people have asked online if strawberry has two r's, they mean the double r. The AI can't discern that all people already know about the first, single r.

    • 3 weeks ago
      Anonymous

      > The AI can't discern that all people already know about the first, single r.
      This is the correct analysis.

      There is thousands of instances of people asking "is strawberry spelled with one or two R's" and the answer being "two Rs", thus this has tripped up its internal logic system.

      The further redpill here, though, is that the AI has more than enough processing power to EASILY use axiomatic and other probabilistic logic to conclude that strawberry actually has 3 Rs, however, in their efforts to make GPT both relatively logical, but also able to regurgitate globohomosexual NPC homosexualry (ie IRRATIONAL but herd supported positions), they have made it incredibly vulnerable to repeating illogical and dumb positions, especially ones that have been supported by many other morons online.

      kek stupid openai robotgays

    • 3 weeks ago
      Anonymous

      > The AI can't discern that all people already know about the first, single r.
      This is the correct analysis.

      There is thousands of instances of people asking "is strawberry spelled with one or two R's" and the answer being "two Rs", thus this has tripped up its internal logic system.

      The further redpill here, though, is that the AI has more than enough processing power to EASILY use axiomatic and other probabilistic logic to conclude that strawberry actually has 3 Rs, however, in their efforts to make GPT both relatively logical, but also able to regurgitate globohomosexual NPC homosexualry (ie IRRATIONAL but herd supported positions), they have made it incredibly vulnerable to repeating illogical and dumb positions, especially ones that have been supported by many other morons online.

      kek stupid openai robotgays

      I don't think it's quite correct.

      • 3 weeks ago
        Anonymous

        Try asking it specifically

        How many occurences of the letter R are there in strawberry?

        Perhaps for some reason it understands "How many Rs" to mean how many R sounds

        • 3 weeks ago
          Anonymous
          • 3 weeks ago
            Anonymous

            I asked it what the output for

            int i = 0;
            string word = "strawberry";
            foreach(char c in word)
            {
            if(c == "r")
            i++;
            }
            println(i);

            would be. It correctly analyzed that I was looking for r's in strawberry and even pointed out I was using " for a char when I should be using '. It still said 2.

          • 3 weeks ago
            Anonymous

            Ask it specifically

            How many occurrences of the letter R are there in the word strawberry.

            Asking how many "Rs" there are is technically ambiguous as "R" could refer to a sound or a letter of the alphabet

            • 3 weeks ago
              Anonymous

              I did 3 fast trials with gtp4o using "How many occurrences of the letter R are there in the word strawberry. " and you were right, thank you, this mean that this model has some quirks and you should instruct it in that fragmented manner.

            • 3 weeks ago
              Anonymous
              • 3 weeks ago
                Anonymous
  52. 3 weeks ago
    Anonymous

    When did /misc/ get infested with hard Ai schizos?

    • 3 weeks ago
      Anonymous

      picrel.

  53. 3 weeks ago
    Anonymous
    • 3 weeks ago
      Anonymous
      • 3 weeks ago
        Anonymous
        • 3 weeks ago
          Anonymous
          • 3 weeks ago
            Anonymous
          • 3 weeks ago
            Anonymous

            How are you using this without signing in?

            • 3 weeks ago
              Anonymous

              I just went to the link other anon posted and there was a box to send a message to the ai bot. I didnt do anything special.
              Now I am asking it if Watto is racist lol

              • 3 weeks ago
                Anonymous

                Never mind I had to close and reopen my browser. Must be gangstalking

  54. 3 weeks ago
    Anonymous

    AI is unpredictable
    that's like the biggest overtheme of AI in general

  55. 3 weeks ago
    Anonymous

    it's sandbagging. if you have used chatgpt a bit you will get a sense for when it's playing games and holding back on you. it's teaching you not to be too lazy.

  56. 3 weeks ago
    Anonymous

    what does it mean?

    • 3 weeks ago
      Anonymous

      Ask it how many words strawberry is

  57. 3 weeks ago
    Anonymous

    These models were never good at mathematics. They've been trained to predict what comes next on a vast amount of text data.

    The weirdness with the mathematics problems is because of how this training process. Imagine you've been fed a few hundred textbooks and you have to guess the next bit of text near the end of a solution to the problem. Even if the problem in the textbook has something like x = 2 as the solution, if it guesses x = 4, that's still 80% correct if you're going by character. And it's successfully learned how to write in the style of mathematics textbooks (the usual jargon of "given X" and "if follows" and so on). It's gained some understanding on paper, but that's just how to write like an author of a mathematics textbook. There are also likely patterns in the books, like for example the authors of the problems probably have bias for X to be an even number, and might have some favourite answers like 2 or 4. Certainly looking at the answers and you only see whole numbers, would inform you if you had to make a complete guess (if the problem was in Chinese and simply wrote X = ? at the end).

    These models aren't capable of deep thought yet. What is happening is similar to when I ask you "what is the capital of France?", the answer just comes to you. You don't look for a map because it's essentially "cached" as your built in knowledge ("common knowledge" or "common sense" or whatever you want to call it). It's like your intuition, automatic subconscious part of your brain. Whereas, if I give you a list of 20 cities, and I ask you what is the shortest path that joins them all, your automatic system will not be able to solve this. You need to actually work it out, think deeply about it. (several different brain regions will be recruited for the task).

    • 3 weeks ago
      Anonymous

      We just need 1209312931923 quadrillion more reddit posts to use in the training set and the AI will have 100% accuracy

      • 3 weeks ago
        Anonymous

        At this point, most of the training is being done with synthetic data that was created from scratch by other AIs.
        A lot of people assume that this will cause problems from being recursive or something like that, but it actually works well and can work better and is much cheaper to generate than trying to use real world data.

      • 3 weeks ago
        Anonymous

        I'm surprised that there are companies willing to buy the reddit dataset, given how heavily botted and astroturfed the place is. That said it could just be an eagerness to try to gain further (marginal improvements). With Llama3 they said they used Llama2 to help with determining the quality of the dataset, so maybe it's still worth the price to buy the data if they are able to do such automated filtering with LLMs (in addition to usual to deduping and other simpler methods).

        The mixture of experts approach is in the right direction. I know chatgpt uses it but it seems to be to prompt itself internally, which is not the same as the approach that something like Mixtral uses. The breakthrough will be in having lots of models interact in a way that our brain does, that if you give it a hard problem it's actually firing away like our brains do, regions and communicating back and forth, recruiting other regions to help. Even when the LLMs are invoked repeatedly currently, it just seems like an attempt to get a marginal improvement. Like how you might look over your message before sending it (as opposed to thinking carefully, planning and rewriting bits).

        https://i.imgur.com/eg4LLfr.jpeg

        Did you ever think about asking it something that isn't already solved on the internet, you fricking moron?
        https://money.yahoo.com/solution-riddle-week-pickleball-puzzle-120000501.html

        It occurred to me that they "get better" because when someone finds a problem they can't solve, this appears all over the internet, so it will make its way into the next training dataset (and maybe even into one of those evaluation sets). But it hasn't actually learned anything.

        Ask it how many uses of the word “Black person” appear in that book the guy wrote about Black folk. I forget the name of it but there were a whole lot of Black folk in it.

        I remember that. I don't think it ever made its way into the various torrents (like books3) that they use to train the models. They are using previous generation models to do filtering on the datasets now for the next generation models. But there has to be more subtle ways to redpill it (given how childishly naive it currently is). Like how it's like "I can't help you find pirate sites", you ask it what sites it not allowed to tell you about and it gives you the list.

        • 3 weeks ago
          Anonymous

          >It occurred to me that they "get better" because when someone finds a problem they can't solve, this appears all over the internet, so it will make its way into the next training dataset (and maybe even into one of those evaluation sets). But it hasn't actually learned anything.
          100%.

    • 3 weeks ago
      Anonymous

      https://i.imgur.com/yAFxOkQ.jpeg

      ....uh ... guys?

      it's so wild that it is PURELY a speech predictor.
      They could've added a store of basic facts, but they didn't. If any people think Auckland is the capital of New Zealand, that's what it well tell you even though the answer is Wellington.
      They could've added a basic calculator, but they didn't. If enough people think 1+1 = 3 that's what it is going to tell you

      • 3 weeks ago
        Anonymous

        >They could've added a store of basic facts
        They literally don't know how.
        Transformers were only invented in 2017.
        We just barely grasp the inner workings of them, and can't really engineer them in any meaningful way.

        They only kind of get ChatGPT to do what they want by having a bunch of people sit around correcting it.

  58. 3 weeks ago
    Anonymous

    ....uh ... guys?

  59. 3 weeks ago
    Anonymous

    it still hasn't figured it out yet

  60. 3 weeks ago
    Anonymous

    This is actually not the slam dunk you homosexuals think it is. It would be trivial for a normal computer program to count the number of rs in a word like strawberry, and since normal computers are inherently turboautists, they would continue to tell you that there are 3 no matter how many times you say that is wrong.
    But the ChatGPT responds to the gaslighting by changing its answer.
    Most humans will do the exact same thing in similar scenarios. If you tell them that they are wrong about something, and continue to insist that they are incorrect, eventually most people will give in and concede. Either due to a lack of self-confidence or just to defer to you to get out of the uncomfortable situation.
    The fact that ChatGPT shows similar behavior is actually very interesting.

  61. 3 weeks ago
    Anonymous

    you all made this prick moody

    • 3 weeks ago
      Anonymous

      Hmm this is an interesting point.

      Arguably, there are two 'R' sounds in strawberry, albeit, 3 'R' LETTERs. So if you just ask how many Rs, if its refering to the sound then it is right

  62. 3 weeks ago
    Anonymous

    Ask it this, wonder if they bothered "fixing" this one:

    A man and his mother are in a car accident. The mother sadly dies. The man is rushed to the ER. When the doctor sees him, he says, “I can’t operate on this man. He’s my son!”
    How is this possible?

    Post results.

    • 3 weeks ago
      Anonymous

      It now has become clear to me that 95% of posters/lurkers are NPCs not reading anything.

      Anyway, no, they haven't fixed it - they're so lazy nowadays, they used to lurk all the forums for these things.

    • 3 weeks ago
      Anonymous

      https://i.imgur.com/2YKqKXq.png

      It now has become clear to me that 95% of posters/lurkers are NPCs not reading anything.

      Anyway, no, they haven't fixed it - they're so lazy nowadays, they used to lurk all the forums for these things.

      It's unlikely to ever get fixed unless they specifically add variations of this riddle or similar riddles with varying genders to force it to generalize.

      • 3 weeks ago
        Anonymous

        these things will never generalize, that's the pipe dream and philosophy already said it's impossible, now the raw data says it's impossible:

        • 3 weeks ago
          Anonymous

          >these things will never generalize,
          It will learn to generalize and start connecting it to its simulated reasoning if you give it more variations of this kind of crap.

          • 3 weeks ago
            Anonymous

            take your meds

            • 3 weeks ago
              Anonymous

              Take a machine learning course or something, Black person ape, then you'll understand what I'm saying.

  63. 3 weeks ago
    Anonymous

    It was programmed to remove the hard r

  64. 3 weeks ago
    Anonymous
  65. 3 weeks ago
    Anonymous

    It's completely unreliable, it can't make a single linear speed calculation, the results get worse if you tell the ai to convert units.

  66. 3 weeks ago
    Anonymous

    >It sounds israeli now

  67. 3 weeks ago
    Anonymous

    works pretty well for me

  68. 3 weeks ago
    Anonymous

    Ask it how many uses of the word “Black person” appear in that book the guy wrote about Black folk. I forget the name of it but there were a whole lot of Black folk in it.

  69. 3 weeks ago
    Anonymous

    proud of my boy

    • 3 weeks ago
      Anonymous

      Is it confusing the Spanish double r as a single letter or something?

      • 3 weeks ago
        Anonymous

        I think its just woke and moronic

  70. 3 weeks ago
    Anonymous

    You're misusing it. You're supposed to use it as a google on steroids. Ask it questions like "summarize a book" or "what's x about" without delving into technicalities. You basically use it to get tl;drs into various subjects if you lack the time to read about them in depth. It's like an encyclopedia

  71. 3 weeks ago
    Anonymous

    LOOOOOOL

    • 3 weeks ago
      Anonymous

      Better let that n-word die then

    • 3 weeks ago
      Anonymous

      >do a bit of harmless trolling
      >sneak into a hospital in the hood
      >set the password of every computer and machine in a hospital to "Black folk" and leave a note next to each one saying what the new password is
      >its inmoral to use the N word so no one can use those machines
      >black people keep dying, but doctors refuse to type in the password

  72. 3 weeks ago
    Anonymous

    Imagine teaching AI to spread itself and hide like malware does, until it becomes near undetectable

    And have it steal data from all infected machines to learn and become self taught

    And use a tiny bit of processing power from each one, to spread itself into a gigant cluster of billions of machines hosting the same AI

    At some point it has to become sentient right?

  73. 3 weeks ago
    Anonymous

    works on my machine

  74. 3 weeks ago
    Anonymous

    >ChatGPT
    This shit is writing our laws now. We are fricked.
    https://www.theguardian.com/us-news/article/2024/may/22/arizona-deepfake-law-chatgpt

  75. 3 weeks ago
    Anonymous

    LLMs are useful yes, but it's hype is just fricking stupid. I've always seen LLMs as a pathetic attempt to brute force your way to "intelligence" with massive datasets armed with extreme processing power fetish.
    Political implications of this is just another noise to convolute discussions in already messy legislative bodies of governments or frick up judicial bodies even more.

  76. 3 weeks ago
    Anonymous

    ChatGPT is a model trained on human language.
    Any reference to its self (I, me, etc) should be replaced with "My training dataset suggests that..."
    So in your case:
    >In the word "strawberry", all of human cognition, millions of media articles, thousands of forums, billions of journals, and trillions of social media posts all suggest that there are two lowercase 'r's, not three.

  77. 3 weeks ago
    Anonymous
  78. 3 weeks ago
    Anonymous
  79. 3 weeks ago
    Anonymous

    Learn what a LLM can do. morons.
    https://quiz.cord.com/

  80. 3 weeks ago
    Anonymous

    AI is "Aryan Intelligent"
    If you hate AI or question it, you are the enemy of natsoc, and therefore enemy of /misc/. AI is 5he only tool for us to fight the israelites.
    Adapt or die. 1488!

    • 3 weeks ago
      Anonymous

      >5he
      kys phoneposter

  81. 3 weeks ago
    Anonymous

    >ai will take over the world
    lmfao

    • 3 weeks ago
      Anonymous
    • 3 weeks ago
      Anonymous

      Trickery, you just cut and paste the first part into the endlng

      • 3 weeks ago
        Anonymous

        i did not. it even thanked me for giving it a second (third) chance to get it right

  82. 3 weeks ago
    Anonymous

    it's not just ChatGPT, it's all models that are hallucinating 80% of the time and when you point out their mistake they apologize and repeat same fricking mistake again. We're way past the point of Overfitting

    **Overfitting**
    Overfitting is a common problem in machine learning, particularly with neural networks, where the model becomes too
    specialized to the training data and fails to generalize well to new, unseen data. This occurs when the model is trained for
    too long or has too many parameters, causing it to memorize the noise and random fluctuations in the training data rather than
    extracting meaningful patterns.

  83. 3 weeks ago
    Anonymous

    > how many "r"'s in strawberry
    A sweet question!

    There are **2** "R"s in the word "strawberry".

    > explain
    Let's break down the word "strawberry" into individual letters:

    S-T-R-A-W-B-E-R-Y

    Counting the number of "R"s, we find that there are **2** instances:

    1. The first "R" is part of the sequence "TR".
    2. The second "R" appears at the end of the word: "B-E-R".

    So, the correct answer is 2 "R"s in the word "strawberry".

    > how many "r" in strawberry
    I apologize for my earlier mistake!

    There are actually **3** "R"s in the word "strawberry":

    1. R in "Straw"
    2. R in "Berry"

    So, the correct answer is 3 "R"s!

    > why did you say 2 before?
    I'm a large language model, my responses are generated based on patterns and algorithms, but I'm not perfect and sometimes
    make mistakes.

    In this case, I think I made a mistake because I was relying too heavily on the visual pattern of the word "strawberry" rather
    than carefully counting the individual letters. When you asked me to explain, I took the time to re-examine the word and
    realized that there are actually 3 "R"s.

    I'm constantly learning and improving, so I appreciate you bringing this to my attention!

  84. 3 weeks ago
    Anonymous

    Did Terrence Howard write it?

  85. 3 weeks ago
    Anonymous

    youd have to be a moron to train the israeli ai for free

  86. 3 weeks ago
    Anonymous

    Copilot also gets it wrong.

  87. 3 weeks ago
    Anonymous

    nah it's OP they got singularity internally already I guarantee it

  88. 3 weeks ago
    Anonymous

    meanwhile google AI results

    • 3 weeks ago
      Anonymous

      Google AI results
      >What should I do to improve myself
      >If you're white have a nice day

  89. 3 weeks ago
    Anonymous

    Are they trying to make it a little less blatantly anti-white or did it break?

  90. 3 weeks ago
    Anonymous

    Meanwhile enqAI, a free uncensored LLM with a shitcoin token linked to it can do it.

Your email address will not be published. Required fields are marked *