So I signed up for a free month of their crap because I wanted to test if it solves novel variants of the river crossing puzzle.

Like this one:

You have a duck, a carrot, and a potato. You want to transport them across the river using a boat that can take yourself and up to 2 other items. If the duck is left unsupervised, it will run away.

Unsurprisingly, it does not:

https://g.co/gemini/share/a79dc80c5c6c

https://g.co/gemini/share/59b024d0908b

The only 2 new things seem to be that old variants are no longer novel, and that it is no longer limited to producing incorrect solutions - now it can also incorrectly claim that the solution is impossible.

I think chain of thought / reasoning is a fundamentally dishonest technology. At the end of the day, just like older LLMs it requires that someone solved a similar problem (either online or perhaps in a problem solution pair they generated if they do that to augment the training data).

But it outputs quasi reasoning to pretend that it is actually solving the problem live.

  • kata1yst@sh.itjust.worksEnglish
    0·
    3 days ago

    I just asked the same question to my preferred onprem model, Mistral-Small 3.1.

    13s on a 7900xtx

    justinledwards/mistral-small-3.1-Q6_K:latest

    To successfully transport the duck, carrot, and potato across the river without losing any of them, you need to plan your trips carefully. Here’s a step-by-step strategy:

    First Trip: Take the duck across the river first.

    • Items on this side: Carrot, Potato
    • Items on the other side: Duck

    Return Trip: Go back alone to the original side.

    • Items on this side: Carrot, Potato (you)
    • Items on the other side: Duck

    Second Trip: Take either the carrot or the potato across the river.

    Let’s say you take the carrot first for illustration.

    • Items on this side: Potato
    • Items on the other side: Duck, Carrot (you)

    Return Trip: Go back with the duck to the original side.

    • Items on this side: Potato, Duck
    • Items on the other side: Carrot

    Third Trip: Leave the duck behind and take the potato across the river.

    • Items on this side: Duck

    • Items on the other side: Carrot, Potato (you)

    Return Trip: Go back alone to get the duck.

    • Items on this side: Duck (you)

    • Items on the other side: Carrot, Potato

    Final Trip: Take the duck across the river. All items are now on the other side.

    By following these steps, you ensure that the duck is never left unsupervised and all items make it safely across the river.

  • BlueMonday1984@awful.systemsEnglish
    0·
    3 days ago

    I’m kinda tired, but this puzzle’s shoved itself into my brain. The obvious solution I can see is, roughly speaking:

    1. Take the duck and carrot across

    2. Take the duck back

    3. Take the duck and potato across

    • shnizmuffin@lemmy.inbutts.lolEnglish
      0·
      3 days ago

      My two solutions:

      1. Eat the carrot. Take the duck and potato across.
      2. It’s a row boat. Take the carrot and potato, supervise the duck as it swims behind you.

      I’m not doing three river crossings, you can’t make me.

      • skillissuer@discuss.tchncs.deEnglish
        0·
        2 days ago

        another solution:

        take duck, carrot and potato at once. if boat is fine if you put duck and carrot in but will sink if you put in duck, carrot and potato then you’re already on horrifyingly narrow engineering margins and probably shouldn’t use it in the first place

    • diz@awful.systemsOPEnglish
      0·
      3 days ago

      Yeah, exactly. There’s no trick to it at all, unlike the original puzzle.

      I also tested OpenAI’s offerings a few months back with similarly nonsensical results: https://awful.systems/post/1769506

      All-vegetables no duck variant is solved correctly now, but I doubt it is due to improved reasoning as such, I think they may have augmented the training data with some variants of the river crossing. The river crossing is one of the top most known puzzles, and various people have been posting hilarious bot failures with variants of it. So it wouldn’t be unexpected that their training data augmentation has river crossing variants.

      Of course, there’s very many ways in which the puzzle can be modified, and their augmentation would only cover obvious stuff like variation on what items can be left with what items or spots on the boat.

      • SGforce@lemmy.caEnglish
        0·
        3 days ago

        It’s just overtrained on the puzzle such that it mostly ignores your prompt. Changing a few words out doesn’t change that it recognises the puzzle. Try writing it out in ASCII or uploading an image with it written or some other weird way that it hasn’t been specifically trained on and I bet it actually performs better.

        • diz@awful.systemsOPEnglish
          0·
          2 days ago

          Not really. Here’s the chain-of-word-vomit that led to the answers:

          https://pastebin.com/HQUExXkX

          Note that in “its impossible” answer it correctly echoes that you can take one other item with you, and does not bring the duck back (while the old overfitted gpt4 obsessively brought items back), while in the duck + 3 vegetables variant, it has a correct answer in the wordvomit, but not being an AI enthusiast it can’t actually choose the correct answer (a problem shared with the monkeys on typewriters).

          I’d say it clearly isn’t ignoring the prompt or differences from the original river crossings. It just can’t actually reason, and the problem requires a modicum of reasoning, much as unloading groceries from a car does.

        • froztbyte@awful.systemsEnglish
          0·
          3 days ago

          oh look it’s a loadbearing “just” in the wild. better hope you can shore that fucker up with some facts

          Try writing it out in ASCII

          my poster in christ, what in the fuck are you on about. stop prompting LLMs and go learn some things instead

          some other weird way that it hasn’t been specifically trained on and I bet it actually performs better

          “no no see, you just need to prompt it different. just prompt it different bro it’ll work bro I swear bro”

          god, every fucking time

        • YourNetworkIsHaunted@awful.systemsEnglish
          0·
          3 days ago

          write it out in ASCII

          My dude what do you think ASCII is? Assuming we’re using standard internet interfaces here and the request is coming in as UTF-8 encoded English text it is being written out in ASCII

          Sneers aside, given that the supposed capability here is examining a text prompt and reason through the relevant information to provide a solution in the form of a text response this kind of test is, if anything, rigged in favor of the AI compared to some similar versions that add in more steps to the task like OCR or other forms of image parsing.

          It also speaks to a difference in how AI pattern recognition compared to the human version. For a sufficiently well-known pattern like the form of this river-crossing puzzle it’s the changes and exceptions that jump out. This feels almost like giving someone a picture of the Mona Lisa with aviators on; the model recognizes that it’s 99% of the Mona Lisa and goes from there, rather than recognizing that the changes from that base case are significant and intentional variation rather than either a totally new thing or a ‘corrupted’ version of the original.

          • SGforce@lemmy.caEnglish
            0·
            3 days ago

            Exactly. It’s overtrained on the test, ignoring the differences. If you instead used something it recognises but doesn’t recognise as the test pattern (having the same tokens/embeddings) it will perform better. I’m not joking, it’s a common tactic to get around censoring. You’re just going around the issue. What I’m saying is they’ve trained the model so much on benchmarks that it is indeed dumber.

            • YourNetworkIsHaunted@awful.systemsEnglish
              0·
              3 days ago

              I don’t think that the actual performance here is as important as the fact that it’s clearly not meaningfully “reasoning” at all. This isn’t a failure mode that happens if it’s actually thinking through the problem in front of it and understanding the request. It’s a failure mode that comes from pattern matching without actual reasoning.

              • diz@awful.systemsOPEnglish
                0·
                2 days ago

                It’s a failure mode that comes from pattern matching without actual reasoning.

                Exactly. Also looking at its chain-of-wordvomit (which apparently I can’t share other than by cut and pasting it somewhere), I don’t think this is the same as GPT 4 overfitting to the original river crossing and always bringing items back needlessly.

                Note also that in one example it discusses moving the duck and another item across the river (so “up to two other items” works); it is not ignoring the prompt, and it isn’t even trying to bring anything back. And its answer (calling it impossible) has nothing to do with the original.

                In the other one it does bring items back, it tries different orders, even finds an order that actually works (with two unnecessary moves), but because it isn’t an AI fanboy reading tea leaves, it still gives out the wrong answer.

                Here’s the full logs:

                https://pastebin.com/HQUExXkX

                Content warning: AI wordvomit which is so bad it is folded hidden in a google tool.

                • YourNetworkIsHaunted@awful.systemsEnglish
                  0·
                  2 days ago

                  That’s fascinating, actually. Like, it seems like it shouldn’t be possible to create this level of grammatically correct text without understanding the words you’re using, and yet even immediately after defining “unsupervised” correctly the system still (supposedly) immediately sets about applying a baffling number of alternative constraints that it seems to pull out of nowhere.

                  OR alternatively despite letting it “cook” for longer and pregenerate a significant volume of its own additional context before the final answer the system is still, at the end of the day, an assembly of sochastic parrots who don’t actually understand anything.