ALMOST PERFECT! ChatGPT-o1 Tested--And Explained!

Поділитися
Вставка
  • Опубліковано 8 лис 2024

КОМЕНТАРІ • 40

  • @tjrs1618
    @tjrs1618 Місяць тому +7

    I appreciate all your tests on LLM/LMMs. Keep up the amazing work, but don't forget to take time for yourself, especially during the school year. :)

  • @jimwhitehead1532
    @jimwhitehead1532 Місяць тому +2

    One might argue that ChatGPT-01 got the driving question correct for the real world, by using rules of driver fatigue that exist. When we take an actual trip, we never use Google or Tesla's estimates of trip time. We stop for breaks!

  • @tm-uz7md
    @tm-uz7md Місяць тому +16

    What you're not disclosing : about 4 generations of GPUs from now, we must leave earth to lasso a star for its energy.

    • @GrigoriZhukov
      @GrigoriZhukov Місяць тому +3

      LOL, we will make a black hole. And have a budding Dyson sphere around it to harvest energy from just outside the edge of the event horizon. With sufficient gaps in the sphere to allow the material to fall into the black hole. Way more, any silly homemade star.

    • @alvaroluffy1
      @alvaroluffy1 Місяць тому +4

      No. Look up the numbers. They are literally marginal for the moment. And you're assuming AI models will keep getting bigger and bigger to infinity. Eventually they just get smaller and smaller and smaller because we'll have AGI or even ASI that will help human researchers to make these models many orders of magnitude more efficient. Its like saying in the 1960s that computers had no future because if you want them to be more capable they will have to grow to the size of a whole building while the opposite happened, computers were more and more miniaturized and now the equivalent of thousands of those room-sized computers fits on a single smartphone. Just so you know, the human brain requires just 20 watts of power. We are trillions of times below the theoretical maximum limit of efficiency. Dont worry, we'll have ASI executing locally on a smartphone-sized device, you just wait

  • @RickPierce7
    @RickPierce7 Місяць тому +7

    I actually think holding back the chain of thought will keep all these llm's from acting like each other. Like a patent, it causes the competitors to come up with its own solutions and theoretically better solutions. It is not like they have not shared the concept of how the build is different. Great video!

  • @MrRandomPlays_1987
    @MrRandomPlays_1987 Місяць тому +2

    The craziest thing is that such AI models that think through their question currently for minutes would all get much faster in the future and would do it even in realtime at some point.

  • @mrlaine1666
    @mrlaine1666 Місяць тому

    Great vid! Don't hold back on the math/physics questions. Go ahead and see if you can find ANY questions it can't handle - stress-test it a bit - I think you'll be surprised.

  • @BrianBull
    @BrianBull Місяць тому +1

    haha that's so funny regarding the 3 rs in strawberry.... Chatgpt struggled with that one for a while lol

  • @darylfortney8081
    @darylfortney8081 Місяць тому +1

    Those suits are the greatest innovation of all. They almost just look like something you would put on to go skiing. Super comfortable looking, not at all bulky and heavy like the og

  • @stevenbarrett7648
    @stevenbarrett7648 Місяць тому +6

    The BBC here in the UK mentioned it briefly, didn't mention Space X much and Elon Musk not at all !

    • @UnCoolDad
      @UnCoolDad Місяць тому

      BBC don't like Musk. He's not woke.

  • @jeff15011
    @jeff15011 Місяць тому +3

    As an old guy, I appreciate that you pronounced it "Gemenee" 😊.

    • @DrKnowitallKnows
      @DrKnowitallKnows  Місяць тому +1

      If you watch the historical films that’s how they say it 🚀

    • @jeff15011
      @jeff15011 Місяць тому

      I remember!

  • @JaredFarrer
    @JaredFarrer Місяць тому +1

    It is pretty cool!

  • @JorgePacker
    @JorgePacker Місяць тому +5

    Can you test if it can solve a sudoku? I tested with other llms, but none of them were able to... Not even a simple sudoku puzzle

  • @roger_is_red
    @roger_is_red Місяць тому +5

    Unbelievably, the small minded, US government can't congratulate the incredible feat of a space walk by an American company. Jeannine

  • @Michael-jd5vf
    @Michael-jd5vf Місяць тому +2

    John. Ask the driving question in a few years and it’ll give you the same answer and this time be correct, why? Robotaxi

  • @rogerc7960
    @rogerc7960 Місяць тому +2

    Level 1 achieved

  • @edwinungerer7989
    @edwinungerer7989 Місяць тому

    Lovely video!

  • @stewartciesla8142
    @stewartciesla8142 Місяць тому

    Pretty cool!

  • @timower5850
    @timower5850 Місяць тому +3

    15/5 = 3 trips is incorrect? What am I missing? Do we have a driver (person number 6) or a Robotaxi or what?

    • @joechughtai3155
      @joechughtai3155 Місяць тому +2

      I think he went over that, but the driver has to drive the car back, so only 4 per trip till the last trip.

    • @timower5850
      @timower5850 Місяць тому

      @@joechughtai3155 I thought maybe that was it. That said, I try to listen to this guy, but he just gets too wordy; I have trouble sticking around.

    • @joechughtai3155
      @joechughtai3155 Місяць тому

      @@timower5850 ha ha

    • @alexanderpoplawski577
      @alexanderpoplawski577 Місяць тому +1

      You have to divide by 4 because the driver needs to drive back and only arrives finally on the 4th trip.

  • @angloland4539
    @angloland4539 Місяць тому

    ❤️

  • @JosephSilv4
    @JosephSilv4 Місяць тому +2

    💪🧠

  • @jason_v12345
    @jason_v12345 Місяць тому

    But it wasn't a simple substitution cipher, and it even talks about how it's not one earlier in its thought process. Anyone can see that instantly by looking at the plaintext and cyphertext. Seems like its thinking was inconsistent.

  • @jabster58
    @jabster58 Місяць тому

    He's just reading everything

  • @SycamoreAfternoon
    @SycamoreAfternoon Місяць тому

    Miller Timothy Davis Patricia Brown Robert

  • @keepcalm7453
    @keepcalm7453 Місяць тому

    ❤❤🙏❤❤🙏❤❤🙏❤❤

  • @paulryaninhi
    @paulryaninhi Місяць тому +1

    Ask it if Donald Trump is the kind of person who should be reasonably considered for the office of President of the United States?

  • @tedmoss
    @tedmoss Місяць тому +1

    I still don't see any proof that LLM's or any AI can be smarter than humans.

    • @glengray8467
      @glengray8467 Місяць тому

      You said it can't analysisdo images, but can it do speech to text? If so, could you give it a utube video to expand its knowledge?

    • @craigington73
      @craigington73 Місяць тому +4

      It outperformed PhD holders.... that's pretty smart!

  • @zen1tsu-sam
    @zen1tsu-sam Місяць тому

    The deficiencies they got now will be what they're good at in the future.