Logic Puzzles with OpenAI o1

Поделиться
HTML-код
  • Опубликовано: 11 ноя 2024

Комментарии • 158

  • @mihirvd01
    @mihirvd01 2 месяца назад +113

    WHAT A TIME TO BE ALIVE!

  • @Srindal4657
    @Srindal4657 2 месяца назад +274

    If they are selling this type of technology, imagine what they are working on now

    • @FOX_ON_THE_RUN
      @FOX_ON_THE_RUN 2 месяца назад +18

      Yup! I would give anything to be able to work with the new systems !✨

    • @bcak611
      @bcak611 2 месяца назад +9

      maybe some tech that can invent new AI xD Terminator coming soon!

    • @ValidatingUsername
      @ValidatingUsername 2 месяца назад +7

      The military is about 50 years more advanced than the civilian release of the technology they control and the civilian roll out directly correlates with mass surveillance of the technology.

    • @SB-cw2mg
      @SB-cw2mg 2 месяца назад +4

      God

    • @harshitbhatt3243
      @harshitbhatt3243 2 месяца назад

      Right ✅️ 😮

  • @ugurergun9816
    @ugurergun9816 2 месяца назад +115

    you put us on hold again for the new sound mode.

    • @MacGuffin1
      @MacGuffin1 2 месяца назад +4

      Yeah wtf, idk why I keep subscribing they promised it ages ago ...

    • @maxc9432
      @maxc9432 Месяц назад +2

      I think it's because of the legal issues with Scarlett Johansson

    • @ababiya_worku
      @ababiya_worku Месяц назад

      I hate to say this but, They will never release it bro! Caz it’s against privacy policy, anyone can use this to do bad things , including the API

    • @ugurergun9816
      @ugurergun9816 Месяц назад

      @@ababiya_worku Maybe they gave up as you said, they don't give much explanation, we don't know.

  • @mfpears
    @mfpears 2 месяца назад +45

    Since the answers are so big, start it as a list of collapsed steps with labels so we only see the details if we expand it. It would be much faster to get to what we're interested in seeing.

    • @happyjohn1656
      @happyjohn1656 2 месяца назад +1

      The response is being streamed so we'd have to wait as long anyway

    • @mfpears
      @mfpears 2 месяца назад +2

      @@happyjohn1656 I mean faster to read the response after it's done

    • @happyjohn1656
      @happyjohn1656 2 месяца назад +2

      @@mfpears ohhh

    • @NotHumant8727
      @NotHumant8727 20 дней назад

      lazy ass future kids not showing any interest for the process, just results. bravo

  • @wavey61
    @wavey61 2 месяца назад +54

    Still waiting on 4o voice mode...

  • @ryanmarcshaww
    @ryanmarcshaww 2 месяца назад +28

    You gotta try the problem for yourself to understand how impressive this is. I have been thinking about it for like 20 minutes and gave up. Maybe I’m just dumb lol

    • @Thompson_James
      @Thompson_James 2 месяца назад +10

      math major graduate here, let x be the prince's current age and y be the princess' current age. First, we know the princess is older than the prince, since the prince 'will be' as old as she is now at some point given the conditions.
      So we have: (x+y)/2 is half the sum of their present ages, when the princess was that age the prince was y-x years younger, and we want twice the age of the prince when she was half the sum of their present ages, so thats 2((x+y)/2 - (y-x)). When the princess is that age, the prince will be again y-x years younger, so he will be 2((x+y)/2 - (y-x)) - (y-x), which is how old the princess is now. So the full equation is 2((x+y)/2 - (y-x)) - (y-x) = y. We solve this equation, first distribute the negative signs to get 2((x+y)/2 - y + x) - y + x = y, distribute the 2 to get x + y - 2y + 2x - y + x = y, combine terms to get 4x - 2y = y, add 2y on both sides to get 4x = 3y, divide both sides by 4 to get x = (3/4)y where x,y ∈ Z+.
      ...however (x+y)/2 ∈ Z+ too since that must be a valid age, so x+y must be even, so either x,y are both even or x,y are both odd. Let's assume x,y are both odd, then we have 4x = 3y, but we arrive at a contradiction, since the left side is even (even times odd is always even) and the right side is odd (odd times odd is always odd), hence x,y are both even. Therefore we know x = (3/4)y where x,y ∈ Z+ and x,y are even. Since we must divide y by 4 and get an even integer x, y must have factors 4 and 2, hence y is a multiple of 8. Finally we arrive at our solution of y = 8n and so x = (3/4)y = 6n where n ∈ Z+.
      Interestingly enough the trivial solution of both of them being zero years old is in the solution space. Also practically n should be less or equal to 15 considering human lifespans.
      o1 got this right, though I don't like its proof.

    • @wenhanzhou5826
      @wenhanzhou5826 2 месяца назад

      It is quite difficult, I would say it is easily an exam problem at a university level math course. Solvable for a math major, but requires significant effort.

    • @Dashman100
      @Dashman100 Месяц назад +1

      No, you are Nerd enough to think 20 minutes!

    • @lukashora5993
      @lukashora5993 Месяц назад +2

      university level math course maybe for history majors

    • @circle_line
      @circle_line Месяц назад +2

      Let's say Q represents age of the princess and K represents the age of the prince
      For "the princess is as old as the prince will be", that can be represented by
      Q(present) = K(at time 2)
      For "when the princess was twice as old as the prince was", that can be represented by
      Q(at time 2) = 2 * K(at time 1)
      For "when the princess's age was half the sum of their present age", that can be represented by
      Q(at time 1) = (Q(present)+K(present))/2
      I think the main thing you have to realize is that the difference between the prince and the princess remains the same throughout the different stages.
      So, then Q(present) - K(present) = Q(at time 2) - K(at time 2) = Q(at time 1) - K(at time 1).
      Using those you can solve for the different variables in a variety of ways. i solved for k1 and k2 and then was able to come up with the 6 and 8 possible solution

  • @deletedytacc-b4l
    @deletedytacc-b4l 2 месяца назад +26

    ai really is the future

  • @D43vil
    @D43vil 2 месяца назад +4

    It'll be a different day when these systems can reliably do math, that day is a lot sooner then I thought.

  • @dorukilhan4329
    @dorukilhan4329 2 месяца назад +4

    It took me an hour to solve it myself, maybe because I am not a native english speaker and i gave my soul to understand the question, the translators doesnt work well on this complex question (that makes the question really hard xd). I am assuming it solved the question faster than every human being in the world anyways.

    • @richikdadhich733
      @richikdadhich733 Месяц назад

      I know a lot of people (engineers, math majors, or people preparing for competitive examinations) who could solve this under 5 minutes. I have mentioned, the solution in a previous comment, this could be done in 2 minutes if you are good with algebra.

    • @dorukilhan4329
      @dorukilhan4329 Месяц назад +1

      @@richikdadhich733 well , it solved faster than 2 minutes so this means am i still right ?

    • @richikdadhich733
      @richikdadhich733 Месяц назад

      @@dorukilhan4329 the last part of your original statement is not correct.

    • @Rudzani
      @Rudzani Месяц назад +2

      @@richikdadhich733Unlikely there’s anyone solving it at this speed.

  • @delirium2944
    @delirium2944 Месяц назад +1

    Graphical solution is the easiest one!

  • @spin4team4096
    @spin4team4096 2 месяца назад +16

    For anyone about to ask why it's so "slow":
    It's slow because it's very new technology. AI in the past never actually "thought" or "reasoned". It was just predicting the answer based on its data, so if you asked a question that was never asked before, it may not get it right.
    GPT-4o did have some minor capabilities with this, as it could "work" and code an algorithm to solve certain types of problems.
    But GPT-o1 is actually reasoning and not just predicting. Which makes the capabilities even larger. Yet it's a very new technology that has never yet been seen before, so of course it's going to be slow. But you can expect it to get faster in the future, of course ;)
    And keep in mind that it's still very fast compared to the average human :P

    • @kingfirebone2000
      @kingfirebone2000 2 месяца назад +1

      I don't seem to quite understand, what differentiates it from regular chain of thought though? Did they figure out how to use more compute per token or something? From what I've seen in the news blog its just a hidden chain of thought. Its not super clear what part of it is reasoning.

    • @mka17_
      @mka17_ 2 месяца назад +1

      @@kingfirebone2000 more compute on inference

    • @kshitijbhattarai9887
      @kshitijbhattarai9887 2 месяца назад

      Do you mean to tell that it like AGI than ai

    • @spin4team4096
      @spin4team4096 2 месяца назад +2

      ​@@kshitijbhattarai9887 It's not AGI ...yet

    • @turnt0ff
      @turnt0ff 2 месяца назад +2

      It’s not AGI, but it’s closer than we’ve ever been, and that’s saying everything. I predict a few more months and boom, we’ll have it. People don’t understand how fast things are progressing 🤯

  • @Usrnet
    @Usrnet 2 месяца назад +1

    I was so hoping for them to eventually have come of age.

  • @PhantomAyz
    @PhantomAyz 2 месяца назад +33

    we are reaching the event horizon

    • @Ghostrider-ul7xn
      @Ghostrider-ul7xn 2 месяца назад

      The military already has. The civilian population hasn't yet.

  • @gdiab
    @gdiab 2 месяца назад +1

    OpenAI regains its crown!

  • @anastasia.2007.
    @anastasia.2007. Месяц назад

    You need to write down four equations with five unknowns. You can then simplify them down to 3x = 4y (if x is the current princess's age).

  • @KristianTheDesigner
    @KristianTheDesigner Месяц назад

    I find it interesting that in this specific comment section people seem more impressed and not scared compared to the videos about it generating a game. The comments in there is filled with anxiety-boosted wannabe programmers telling every junior-engineer they are no longer needed and doomed. I like this comment section alot more because, well, you seem more level-headed and quite frankly, smarter than those in that commentsection 😄

  • @trucid2
    @trucid2 2 месяца назад +4

    Where is Sora? Voice?

  • @adityakhanna113
    @adityakhanna113 2 месяца назад +2

    To solve this, work backwards. Start at the end of the sentence and define your variables+equations
    They all simplify and the answer falls out. It's incredible this model can do it but like I'm sure cg4 with a push can do it too

  • @lith8163
    @lith8163 2 месяца назад +1

    1:48 Aw man! I was just going to say that!

  • @PSpace-j4r
    @PSpace-j4r 2 месяца назад +1

    This system hopefully will be implemented in all hospitals accross north America and the EU

    • @metakron
      @metakron 2 месяца назад

      OMG THE DISTOPYAN CYBERBIOPUNK FUTURE IS REAL

  • @Teslabull
    @Teslabull 2 месяца назад +2

    I watched all o-1 videos, still confuse why they keep coming to the same sofa

    • @fixapp1775
      @fixapp1775 2 месяца назад +1

      bc where they live its all matrix, we just get some demo Ai versions to test our reactions how we adapt to technology

  • @khoslaaccount
    @khoslaaccount 2 месяца назад +1

    this goes crazy in middle school math club

  • @adityakhanna113
    @adityakhanna113 2 месяца назад +15

    Also how do we know this problem wasn't in the training set? If chatgpt is trained on everything

    • @AncientPrayers
      @AncientPrayers 2 месяца назад +7

      Because it provided a detailed step by step how it arrived to the solution. It actually thought about it.

    • @flutteredlearning
      @flutteredlearning 2 месяца назад +1

      @@AncientPrayers This applies only if this problem has never been publicly solved or talked about in YT videos / on twitter. Do you think that is the case?

    • @CoolIcingcake3467
      @CoolIcingcake3467 2 месяца назад +1

      @@flutteredlearning yeah, i dont trust this presentation, this problem is already contaminated to begin with. i will see this AI models performance on SIMPLE Bench to see the real result

    • @citizen3000
      @citizen3000 Месяц назад

      @@CoolIcingcake3467well you got that lol

  • @GNARGNARHEAD
    @GNARGNARHEAD 2 месяца назад

    are wee finally going to get some good Riddler stories in Batman comics😮

  • @h-e-acc
    @h-e-acc 2 месяца назад +6

    This is really awesome, but still waiting for the live vision and voice features you teased a couple months ago

    • @citizen3000
      @citizen3000 Месяц назад

      You’ve been told when to expect it. Calm down.

  • @AIeks1729
    @AIeks1729 Месяц назад

    I think that 4k for Princess and 3k for Prince where k is a natural number is the correct answer

    • @azai.mp4
      @azai.mp4 Месяц назад

      There's another number in the problem that must be an integer, which is why o1 says 8k and 6k. (i.e. 4k and 3k with k being even) I missed this too.

  • @Theguywithspectacles
    @Theguywithspectacles 2 месяца назад

    THE IDES OF MARCH HAS COME

  • @aragon5956
    @aragon5956 Месяц назад

    Can neural networks derived from mineral chemistry be used to create neuromorphic chips for resistive memory?

  • @user-kl8vr8io3b
    @user-kl8vr8io3b 2 месяца назад +1

    Впечатляет ..

  • @lronSausage
    @lronSausage 2 месяца назад +2

    James May

    • @donsurlylyte
      @donsurlylyte 2 месяца назад

      or he may not, who knows.

  • @ripplecutter233
    @ripplecutter233 Месяц назад

    Ok yeah I get it, it's smart, geez. Just rip off the band-aid and release AGI already so I can stop wondering whether I'll still have a job

  • @ramsey2155
    @ramsey2155 2 месяца назад

    This is gona replace thinkers

  • @pile333
    @pile333 Месяц назад

    Next video could be totally done with a stable AI animation.

  • @Vlican
    @Vlican Месяц назад

    What grade math is this? Seems like quite the high level algebra...

  • @richikdadhich733
    @richikdadhich733 Месяц назад +4

    It unnecessarily complicated the solution. I solved it in under 2 minutes. The essence is that the difference of ages prince and princess would be the same in the two scenarios. Assume princess age to be x and prince age to be y. Hence, we get the equation x - y/2 = y - (x + y)/2. Solving this, we get the ratio of x : y = 4 : 3, which is the correct answer.

    • @big_mac_love
      @big_mac_love Месяц назад +2

      Not bad, it took me like 30 minutes to come up with it...

    • @esveann
      @esveann Месяц назад

      Your equation is giving a 2 : 3 = x : y ratio tho

    • @ZEPHYRZHANG-mg8zi
      @ZEPHYRZHANG-mg8zi Месяц назад +3

      It's just a simple system of equations, nothing crazy. CHatgpt does way more complicated math I dont know why they dont show case that.

    • @citizen3000
      @citizen3000 Месяц назад

      Nobody cares what you solved in 2 minutes. You’re not an LLM.

  • @user-br9js4el2f
    @user-br9js4el2f 2 месяца назад

    I think things will start to become Crazy in in the 2030s

    • @spol
      @spol 2 месяца назад

      try 2025

  • @metakron
    @metakron 2 месяца назад

    COGITO ERGO SUM🗣️🗣️🗣️🗣️🔊🔊🔊🔊🔥🔥🔥🔥

  • @rafiqulhaque1189
    @rafiqulhaque1189 Месяц назад

    Is O1 giving the correct answer here? I get the princess age to be 1.4 times of prince

  • @universemaster
    @universemaster 2 месяца назад

    I may each month, and yet I still don't have advanced voice mode.

  • @vishnuprathish
    @vishnuprathish 2 месяца назад +7

    can it spell Strawberry? Show us

    • @AncientPrayers
      @AncientPrayers 2 месяца назад +1

      They did. Check the other videos. (ChatGPT 1o Reasoning Counting)

  • @manjuverma3342
    @manjuverma3342 2 месяца назад +1

    I couldnt solve the problem . ITs difficult.

  • @tejpalnayak5033
    @tejpalnayak5033 Месяц назад

    Now I really feel like 'dumb' or like a 'chimpanzee' in front of that A.I.

  • @trucid2
    @trucid2 2 месяца назад +1

    It's a well known problem that's in the training data set.

  • @lokidoki471
    @lokidoki471 2 месяца назад

    i just asked this model to word count its response and it got it wrong..

  • @derek_mckay
    @derek_mckay 2 месяца назад

    Shouldn't it read "their present ages", and not "present age"... ?

  • @DrEnginerd1
    @DrEnginerd1 2 месяца назад +3

    How many 'r's are in strawberry?

  • @galailliz
    @galailliz 2 месяца назад

    Money is going to go away

  • @furuf3756
    @furuf3756 2 месяца назад

    remarkable

  •  Месяц назад

    BG2

  • @ShpanMan
    @ShpanMan 2 месяца назад

    What percent of humans can solve this? And no human can solve it this quickly, or write out the solution this quickly.

    • @adityakhanna113
      @adityakhanna113 2 месяца назад +2

      i don't know how to say this politely but this is a problem a 9th grader can solve. You have to set up the equations and it works. I can for sure do it on par with the system and the solution is barely 4 lines

    • @flutteredlearning
      @flutteredlearning 2 месяца назад

      I think the better question is to ask this: consider the number of resources that went into creating this model. The entire thing. GPUS, datacenters, electricity, manpower. Now get a group of 3 highschoolers together in math class, watch them come up with a solution (slower), and tell me this is that impressive.

    • @ndjarnag
      @ndjarnag 2 месяца назад +2

      @@flutteredlearning Good point. But think about the scale: deploy this model across the whole world. How many problems can it solve over time? compare it to how many problems a group of high schoolers can solve over the same time.

    • @ShpanMan
      @ShpanMan 2 месяца назад

      @@adityakhanna113 @ I don't know how to say this politely, but you need to get out and explore the world more and talk to the majority of people in the world - in 3rd world and developing countries.
      You're so very smart, but ironically you used 4o thinking instead of o1 and didn't actually answer the question 🤣

    • @MacGuffin1
      @MacGuffin1 2 месяца назад

      @@ShpanMan Bro, respectfully... In my country this is 100% of 2nd graders. The point of the demo is that previous AI can do basic: 'If I have 4 strawberries and...' for something outside of it's training data, but once the loops became convoluted/nested etc, LLMs had a difficult time, even with scratchpad memory. Being able to do this effectively is a big deal, although I have yet to like this new model, it seems to come up short in other areas, and most of what it does is just giving you a loading banner for stuff that 4 was doing in the background anyway...

  • @ujjawaldiwan5370
    @ujjawaldiwan5370 2 месяца назад

    Hey I thought gpt-5 is the next. Wtf is o1

    • @Ori-lp2fm
      @Ori-lp2fm 2 месяца назад +1

      Edgy minimalists

  • @manjuverma3342
    @manjuverma3342 2 месяца назад

    wow

  • @onur_eren48
    @onur_eren48 2 месяца назад +1

    Türkiye 🇹🇷

  • @timaka46
    @timaka46 2 месяца назад +2

    WE ARE SO BACK

  • @AphexHenry
    @AphexHenry Месяц назад

    Sam Altman was acting pretty tense lately when people asking for voice mode. it's all on OpenAI man, you're being absolutely opaque about something you got us excited about. we would stop asking if you would start explaining what's up with it. was it just a bluff? are you waiting for us to forget about it?

  • @thechadeuropeanfederalist893
    @thechadeuropeanfederalist893 2 месяца назад

    That puzzle isn't that easy. Try to solve it on your own first.

  • @qweasdzxcrfv1
    @qweasdzxcrfv1 2 месяца назад

    metanle subtitulos ya pe causa

  • @CameronLestagez
    @CameronLestagez Месяц назад

    what?

  • @MountMatze
    @MountMatze 2 месяца назад +2

    I bet 20 bucks that it still fails in simple logic

  • @alinmathuo4018
    @alinmathuo4018 2 месяца назад +2

    Where is the new fast voice?

  • @giochkhaidze7234
    @giochkhaidze7234 2 месяца назад

    O(1)

  • @ryzikx
    @ryzikx 2 месяца назад

    🥱🥱🥱🥱🥱

  • @ClitGPT
    @ClitGPT 2 месяца назад

    There are things I can do better.

  • @grady_young
    @grady_young 2 месяца назад

    It’s good. But it still doesn’t know 9.11 < 9.9

  • @QuantumVoid-ro3hi
    @QuantumVoid-ro3hi 2 месяца назад +2

    "half the sum of their present age"? Stop right there, because you can't have a "sum" of a singular "age." The question doesn't even make sense.

    • @cajbajthewhite4889
      @cajbajthewhite4889 2 месяца назад +5

      Given the context it clearly means the sum of the ages of the prince and princess at the present time. Those are the only two integers that apply given the specified time, and it's asking for the sum. Maybe they should have said "ages" but I think that's a linguistic edge case and the model was able to figure it out anyway.

    • @tyrone_music
      @tyrone_music 2 месяца назад

      ​@@cajbajthewhite4889Came to say this

    • @QuantumVoid-ro3hi
      @QuantumVoid-ro3hi 2 месяца назад +1

      @@cajbajthewhite4889 When performing math, you don't make assumptions without stating them upfront. That's a reasonable assumption, but it should state it explicitly or ask for clarification before spitting out an answer.

    • @QuantumVoid-ro3hi
      @QuantumVoid-ro3hi 2 месяца назад

      @@cajbajthewhite4889 Also, I haven't worked it out yet, but just glancing at it, this looks like a basic Algebra level 1 system of equations that it has grossly overcomplicated.

    • @adityakhanna113
      @adityakhanna113 2 месяца назад

      ​@@QuantumVoid-ro3hithe equations aren't even that complicated tbh

  • @v.svishnu2380
    @v.svishnu2380 2 месяца назад

    It's super slow tho

    • @ahsookee
      @ahsookee 2 месяца назад +8

      That's the point. Use a different model if you don't need the reasoning edge

    • @matteofalduto766
      @matteofalduto766 2 месяца назад +7

      Still faster than me an probably you

    • @spin4team4096
      @spin4team4096 2 месяца назад +5

      It's slow because it's very new technology. AI in the past never actually "thought" or "reasoned". It was just predicting the answer based on its data, so if you asked a question that was never asked before, it may not get it right.
      GPT-4o did have some minor capabilities with this, as it could "work" and code an algorithm to solve certain types of problems.
      But GPT-o1 is actually reasoning and not just predicting. Which makes the capabilities even larger. Yet it's a very new technology that has never yet been seen before, so of course it's going to be slow. But you can expect it to get faster in the future, of course ;)

    • @ricosrealm
      @ricosrealm 2 месяца назад +3

      It's slow because it is generating 100's of different ideas in the background before choosing the right one. Faster than a human can write all of that reasoning up into a report.

    • @f.libaax7408
      @f.libaax7408 2 месяца назад

      @@spin4team4096 It is closed source and overhyped. Every ML model is based on statistics and not reasoning.

  • @PeterTroutman
    @PeterTroutman 2 месяца назад

    what a low quality video. Why does every tech video look like its being done off cuff by some random surfer bum

  • @xitize
    @xitize 2 месяца назад +2

    This is not reasoning, llm are never made to reasoning. Its scraping data, labeling into model, and showing where the problem already solved into sets of words.

    • @citizen3000
      @citizen3000 Месяц назад

      Thanks for your deep insights

  • @johnvanderschuit
    @johnvanderschuit Месяц назад

    Still no advanced voice mode for 4o lmao