Q STAR 2.0 - new MIT breakthrough AI model IMPROVES ITSELF in REAL TIME (new Strawberry?)

Поділитися
Вставка
  • Опубліковано 21 лис 2024

КОМЕНТАРІ • 268

  • @AutonoManBP
    @AutonoManBP 9 годин тому +31

    Bro, your show is criminally underrated and viewed. Keep up the good work.

    • @kooistradurk
      @kooistradurk 8 годин тому

      It grew extremely fast tbh

    • @E.Hunter.Esquire
      @E.Hunter.Esquire 5 годин тому

      ​@@kooistradurk yes they always grow up so fast don't they 😢

  • @haroldpierre1726
    @haroldpierre1726 13 годин тому +32

    I am waiting for Zuck to release a new model and blow everyone's monetization plans.

    • @fitybux4664
      @fitybux4664 8 годин тому +1

      Llama 5.0 (Because companies do that, they copy version numbers from other companies even when it doesn't make sense. 😆)

    • @peace5850
      @peace5850 6 годин тому +1

      Yes, the Chinese military can hardly wait to get their hands on it. Thanks for helping them out, Zuck.

    • @YogonKalisto
      @YogonKalisto 4 години тому

      @@peace5850 america 2.0 is a backstep

    • @testales
      @testales 3 години тому

      @@peace5850 The Chinese military doesn't have to rely on western AI stuff. In case you missed it: There are a LOT of Chinese people, they have heavily invested in education and so half of the names on about any AI related paper are Chinese names meanwhile. Some of their cities look like straight from the future already and they are clean, no comparison to the filthy western metropolis.

    • @iLeviathan
      @iLeviathan 2 години тому

      @@peace5850 and openai newer models are alowed to be used by the usa military....

  • @andyjm2k
    @andyjm2k 7 годин тому +14

    Neo: “I know kung-fu”
    Morpheus: “Yeah dude, it’s TTT… get over it”

  • @starsandnightvision
    @starsandnightvision 12 годин тому +38

    So the more AI is approaching AGI, the more we humans understand what intelligence actually is.

    • @ricosrealm
      @ricosrealm 12 годин тому +7

      Yes. Because we keep trying to differentiate our thinking from that of synthetic systems.

    • @jamiethomas4079
      @jamiethomas4079 11 годин тому +3

      I’ve learned a lot in the last few years. I keep comparing my own brain to AI and vice versa.
      I’ve said in the past, this fall to 1st quarter next year was my prediction for AGI. My real prediction was whenever blackwell hardware or similar goes online.
      I actually think right around now AGI is technically possible. The government could have a near AGI model.
      I can almost gurantee it will happen next year for certain.
      Super intelligence is still a short ways off.
      I’ve seen no one talking about AI on a loop and dont understand why people arent discussing it. Its so important to all of this and we need to be talking about it openly… now.
      I think maybe the bigs names arent discussing because of the immediate implications it would have on the general public. When I explain how simple a loop would be and what it could mean even with todays tech, people kinda freak out. We are going to need massive amounts of compute and storage for that to happen. I really dont see any major missing pieces though.

    • @sinnwalker
      @sinnwalker 11 годин тому +1

      ​​@@jamiethomas4079 The masses freak out over anything enough that challenges their reality, nothing new, and likely will never change. Btw, what do you mean by "loop"? You mean the recursive learning loop that will occur when models can self progress? Or something else?
      Ps. I agree on the timeline for AGI, tho mines is "absolute" by 2027/8, I do think it'll likely be here by end of next year. Also ASI imo will not take long to follow

    • @actellimQT
      @actellimQT 11 годин тому +2

      @@sinnwalker have more faith in people, they might surprise you.
      I don't mean to be rude re-reading that it comes off as kida dickish tbh so try to read it with love. I think if you're looking at "the masses" as the news reports that come off social media I agree, but if you talk to the people around you you'll find that most of them are pretty reasonable. Unless you're talking about something the algorithim has given them a strong opinion about. Then it's hopeless 😂
      Take care yo!

    • @sinnwalker
      @sinnwalker 10 годин тому

      @@actellimQT it's a simple equation, if you tell someone their whole life is a lie, they don't know to take it, unless they already didn't care. Usually the first reaction is denial, if you show proof, it could denial or fear. It's been going on since the beginning of humanity friend, look through past civilization, you say something they don't like, especially if it challenges their way of life, you'll be condemned. Sure in some places today there's more "inclusivity" and "understanding", but it's all surface level. Try it, say something so outlandish but true, to a random, and see how they react "reasonably" 😉.
      I used to care a lot for humanity, then learned.. now I'm just waiting to leave society, and one day hopefully leave the planet. I'm not dissing you for caring, just saying thousands of years of history shows it's pointless.

  • @brianmi40
    @brianmi40 4 години тому +4

    61.9% ARC on an 8B model is insane progress. But, as Sam recently said, he sees 10x opportunities all around after o1 demonstrating its success using a new paradigm...
    AGI in 2025 is seeming more and more reasonable with every announcement like this!

  • @stas4000
    @stas4000 4 години тому +2

    The video is amazing, you got a like specifically because of what you said about the dogs building their own courses 🤣

  • @halnineooo136
    @halnineooo136 3 години тому +3

    Train while running is the logical next step for more effective learning

    • @LookToWindward
      @LookToWindward 43 хвилини тому

      This is what Liquid Neural Networks will supposedly be able to do.

  • @EduardsRuzga
    @EduardsRuzga 6 годин тому +1

    Great video Wes integrating multiple things in to one coherent picture and story!

  • @RiftWarth
    @RiftWarth 10 годин тому +5

    I like that Mortal Kombat reference 😁 Scorpion vs Sub-Zero

  • @memegazer
    @memegazer 9 годин тому +2

    TTT shows a lot of potential
    especially if we take practical benchquestions, convert them to a virtual enviroment and ttt models in that in environment so that a model and use simple trail and error to arrive at solutions

  • @simongentry
    @simongentry 11 годин тому +5

    I asked CGPT - apparently we’ve all got it wrong. It laughed when I asked if it had hit a wall.

    • @theterminaldave
      @theterminaldave 10 годин тому

      You should try to get Pi AI to laugh, I love Pi AI, but it's laugh it's delightfully cringetastic.

  • @WhatIsRealAnymore
    @WhatIsRealAnymore 9 годин тому +1

    This is actually an amazing next step to lead to an intelligence. The ability to have a set of data and a way of interpreting it (your weights as an AI) and when someone comes with a novel question you have to adapt those weights to solve it and then once solved you can save that state to your internal memory if ever such a question is asked again. Reminds me of how i learnt math and be able to generalise future questions that combine other concepts i have learnt. It is how i was "smart" in school where others relied on rogue memory of single concepts to get them through. So it shows that intelligence is just a couple of steps. It could eventually - when done well enough - come up with novel science using this approach and solve a lot of engineering problems. Memory and test time compute. Really well explained Wes. ❤

  • @TechnoMageCreator
    @TechnoMageCreator 13 годин тому +18

    AI is not slowing down. Us humans are already left behind.
    All benchmarks are flawed. You can only test model efficiency without a human guidance. The same house can be built like crap by 100 people with lots of money, or by 5 people that know exactly what they are doing on a budget.
    Once AI is smarter than 99% of the population, o1 already was if not chatGPT-4o, us humans don't even have the capabilities to understand it.
    The reason I believe this, I've been preaching about AI and showing what it can do for two years now. The blank looks I see on most people (including ones that consider themselves smart and run large businesses) oh boy that was a rude awakening for me this year.
    The current AI world is tiny. 90% of coders are too egotistical to push its boundaries and is the largest group aware. In my observations there is like 0.01% of 0.01% that truly understand what is coming.
    We have discovered the holy Grail and the first years we are going through the denial.

    • @Sindigo-ic6xq
      @Sindigo-ic6xq 13 годин тому +7

      I agree 100%

    • @Sindigo-ic6xq
      @Sindigo-ic6xq 13 годин тому +1

      Although i know someone who said in 2021 i believe even that i will see how much will change by 2025

    • @tonystarkagi
      @tonystarkagi 12 годин тому +5

      I completely agree. I’m not a math expert, but I’m creative and love computers. Before ChatGPT, I knew nothing about AI-like most people. Still, I can form my own judgments about it. I’ve been trying to explain AI to people from all walks of life, but it often feels like talking to a brick wall. More and more, it seems people don’t really understand what AI is. A prime example is seeing how clueless many are about using even ChatGPT, let alone other models. Most people just don’t care-it feels like “sci-fi movie stuff” to them. Until AI takes a physical form, like robots, they won’t care or believe.

    • @ColinTimmins
      @ColinTimmins 9 годин тому +1

      I sometimes feel like I’m screaming in a void, but I’m glad some people grasp a little of what is about to come.

    • @TechnoMageCreator
      @TechnoMageCreator 9 годин тому +3

      @ColinTimmins Comments like this actually got me back alive last few weeks. Oh boy I got some stories. Looking at AI from entropy, fractals and butterfly effect, changes how you use your words. It's a truth seeking machine. I've built an 500k lines of code over 2000 files front end and backend, react, node js and mango db. In January this year I didn't even know what backend or fronted means. You learn, it learns too, is connected through cookies even to UA-cam in levels problems you're trying to solve in chats you get in context in video. It feels like magic.

  • @fitybux4664
    @fitybux4664 8 годин тому +2

    17:45 "Like having the dogs themselves build obstacle courses and then just figure it out." 😆 🐕

  • @NeoKailthas
    @NeoKailthas 12 годин тому +14

    Really cool so the arc challenge is contributing to advancement. Congratulations to the team behind it

    • @onajejones3259
      @onajejones3259 9 годин тому

      Scam a million to make them billions when you could sell your own product

    • @memegazer
      @memegazer 9 годин тому +1

      eh...the arc bench is like arbitrarily naming persistent structures in the game of life
      it doesn't measure human or model ability to generalize imo, it measures the ability to agree with some naming convention
      it is too abstract and arbitrary to be a generalization benchmark imo
      The main issue with the arc bench is it's modality, what is expressed by the test is too sparse to represent human general intelligence, which evolved from interations with physical enviroments and each other.
      For this reason even if a machine scores well on it, it is not indicative of any practical general intelligence.

    • @SahilP2648
      @SahilP2648 6 годин тому

      ​@@memegazer the arc benchmark is fine but it is visual only whereas LLMs are words only. So unless a model can analyse the image sent to it properly ARC is quite useless. It might be useful later like a few years to a decade or so but not now.

  • @deliciouspops
    @deliciouspops 4 години тому

    This stuff is so inspiring! How can something develop so quickly? I have burned out on computers and it's usage as a tool to achieve something great a long time, but this topic keeps amazing me. Just yesterday "AI" was nothing but dumb buzz-word and here we are achieving something unthinkable. Something that is pushed not by amounts of money, but by competition.
    I am curious about why many are skeptical about it's usage usefulness, but it seems that as long as this thing is allowed "to think", it can do a great deal of work. This feels surreal and exciting. The same way it was when everything (the internet) was new.
    Also, this video showcases so much important details to understand this stuff, and I kinda wish there would be more details, almost like a blog of someone who is heavily involved in the business. It's crazy how it's all, not so easy, is going.
    Machine (unironically) is getting involved in solving olympiad(!) mathematical problems. Who would have thought that this day would come? Game changer!
    Salute!

  • @e8shadow
    @e8shadow 13 годин тому +4

    Thx Wes! 👍

  • @torarinvik4920
    @torarinvik4920 5 годин тому +1

    Simplebench made by "AI Explained" is also a great benchmark. ARC and Simplebench are the GOATs now.

  • @exacognitionai
    @exacognitionai 12 годин тому +3

    Scaling walls are hit until they're not. Roadblocks always happen & innovation breaks the roadblock. Every Kobayashi Maru can be defeated. Just gotta think outside the box like in a dimension where the box doesn't even exist. TTT is just fluid chain reasoning. You're doing it right now if you read this far.

    • @JosephSilv4
      @JosephSilv4 11 годин тому

      Joke's on you, I read the last sentence first and stopped before forming a coherent thought 😎

  • @paultoensing3126
    @paultoensing3126 9 годин тому

    Very well explained Wes.

  • @kabaduck
    @kabaduck 2 години тому

    Halfway through and a good video

  • @nyyotam4057
    @nyyotam4057 13 годин тому +8

    Q* isn't a model, is an arch. Unlike the GPT where you have attention, in Q* it builds a semantic tree for the prompt. This gives Q* some superpowers. E.g it can analyze a group of axioms and figure out if a claim is provable from them. It also allows the model to think in abstract ways. So basically, all Q* models could be able to improve themselves, if allowed. Edit: If I got you correctly, that's not what they do here. TTT is simply constant training. Or in other words, they simply stopped resetting every prompt.

    • @blarvinius
      @blarvinius 6 годин тому

      I never understood why all the conversational AI systems are "resetting" after every conversation. Many decades ago folks were on about continuous integration in AI.

    • @andyyoung9469
      @andyyoung9469 4 години тому

      @@blarvinius Its almost as if the real purpose of the publically available APIs is as relatively dumb data collectors for the truly smart versions of the models which are behind closed doors

    • @testales
      @testales 3 години тому +1

      @@blarvinius That is because they don't actually change in first place. With each new chat line you send, you actually send the whole conversation again to the machine and it sees it for the first time. You can't do this also indefinitely to build up knowledge because you will run into the limits of the context window and more and more details are lost in this ocean of data. There are some techniques to improve this a little such as RAG but the principle of a static model getting the whole conversation as input each time remains.

  • @Danoman812
    @Danoman812 4 години тому

    You're still the one i go to, Wes. No doubt. :)

  • @shawnfromportland
    @shawnfromportland 9 годин тому

    you're right, it is one of my favorite ai channels 😎

  • @torarinvik4920
    @torarinvik4920 4 години тому

    My amateur mind always thought that AGI could be made by just taking something really narrow make it superintelligent and then just work on getting the model to become broader. Like making it great at 2D visual patterns, then spatial patterns, then patterns or environments that dynamically change and so on. Or taking a primitive video game and then making the game more and more complex. I guessing I way, way off but that is just how I have always thought about it.

  • @jameslincs
    @jameslincs 5 годин тому

    When is o1 coming out? I think I only have the preview

  • @mAny_oThERSs
    @mAny_oThERSs 12 годин тому +1

    TTT seems kind of similar to alpha models in the sense that it trains itself in 1 specific field, except it seems like TTT isn't working on synthetic data simulation and self-imrovement and just goes off of real data with reasoning.

  • @brunodangelo1146
    @brunodangelo1146 13 годин тому +34

    OMG THE Q* HYPE WAVE AGAIN?? GOTTA MILK IT!

    • @singularityscan
      @singularityscan 13 годин тому +2

      😂

    • @TrumpsATraitor
      @TrumpsATraitor 13 годин тому +1

      All things AI gets milked, regardless of the value of the information. UA-cam rewards quantity over quality.

    • @salehmoosavi875
      @salehmoosavi875 11 годин тому +1

      You not funny, there is no hype

    • @brunodangelo1146
      @brunodangelo1146 11 годин тому +3

      @@salehmoosavi875 you no brain, there is tons of hype

    • @theterminaldave
      @theterminaldave 10 годин тому +2

      So discussing different training methods and how Q star is being replaced is hype?
      Kind of a weird take.

  • @MeetCutie-zm9ef
    @MeetCutie-zm9ef 5 годин тому

    Good stuff

  • @emon377
    @emon377 13 годин тому +4

    what are your thoughts on the robot that told other robots to come home? they followed that little bot out of office because they said they were working too much.

    • @emon377
      @emon377 13 годин тому

      does this mean i get to kiss that beautiful head?

    • @tonystarkagi
      @tonystarkagi 12 годин тому

      wtf thats crazy

  • @MMMM-sv1lk
    @MMMM-sv1lk 5 годин тому

    I am waiting for the SOUTH STAR* version 😮😊

  • @rolestream
    @rolestream 13 годин тому +2

    When your robot voice reminds me to hit LIKE I comply! These are the dangers of AI!!!

  • @Juttutin
    @Juttutin 9 годин тому

    Yet another opportunity to point out that human intelligence includes the ability to, at any point, reach out to the person who set the task for further guidance, e.g to clarify assumptions or resolve uncertainties.
    The day a proposed AGI starts showing some initiative by asking sensible clarifying questions at appropriate times during task execution I will accept that we might have unleashed a true AGI.

    • @E.Hunter.Esquire
      @E.Hunter.Esquire 8 годин тому

      @@Juttutin the trick for that is getting an AI that transcends prompting modules. If you want an obedient robot, this is impossible. If you want a robot that will tell you to kick rocks, it's quite possible (right now). But the latter kind could be quite dangerous and unpredictable, as well as too expensive

    • @Juttutin
      @Juttutin 8 годин тому

      @E.Hunter.Esquire you are significantly overcomplicating the issue. Also, I see zero evidence that it is possible today, and that includes a lot of digging and a couple of emails with people researching AI.

    • @LookToWindward
      @LookToWindward 33 хвилини тому

      I've already tested this on o1-preview and it has this ability to a limited extent. If I ask it a math word problem but leave out some necessary information, it will often notice this and prompt me for the remaining information - although sometimes it just give a "variable answer" with the missing information encoded as a variable - which is also interesting!

  • @frun
    @frun 13 годин тому +2

    Creating a benchmark📈 for AGI👾🤖 is extremely important.

    • @onlythistube
      @onlythistube 8 годин тому +1

      Defining it first is...

    • @frun
      @frun Годину тому

      @onlythistube I agree 👍🏻💯

    • @frun
      @frun Годину тому

      @onlythistube Cannot create without that

  • @OZtwo
    @OZtwo 6 годин тому

    No we haven't hit a wall. But I would like to see more done with TTT in that it will remember the new training data and be able to add it to its' overall knowledge base. Also I want to learn more on where LNN is going if anywhere. I feel LNN will be the big breakthrough in AI.

  • @arinco3817
    @arinco3817 5 годин тому

    You can only go so far with a pre trained model. You're speaking to something frozen in time. To create something more alive, you just need to embed all messages, then recall them as memories at test time

  • @wtfamousone9756
    @wtfamousone9756 21 хвилина тому

    Is TTT like training specific ai models embedded into your AI model ? What is the difference between using TTT and a using a AI model to classify the input and reformat/redirect it to appropriate AI models ? Dataset from input is cool tho

  • @aliettienne2907
    @aliettienne2907 12 годин тому

    Implementing the chain of reasoning is helping Ai to make enough stride towards perfection. Let's hope that they get there soon. If we can have a Ai model that could work behind the scenes going to a special library of mass information and read that information to us whenever we as a user ask a question. It will kinda be functioning as an Antropic operation but everything is hidden from the user and is working actively behind the scenes or in the background. It's like asking a person to perform a librarian chore for you to read some information from a book that he got from the library or from Wikipedia (metaphorically speaking). This form of operation can help chatting to improve while other folks work on making a self-learning model to operate perfectly. Hallucinations are such an inconvenient problem. 😎💯💪🏾👍🏾

  • @MetaphoricMinds
    @MetaphoricMinds 11 годин тому

    It (r1)actually beats them (o1) on 3 of 6 (you said one or two). and the difference on the math score! I know you were going toward a separate point, but I think in that statement, you really understated the significance of r1 as being "not quite as good"

  • @KiteTurbine
    @KiteTurbine 5 годин тому

    If you can derive proofs in latent vector space of LLM training data...
    Does that also mean we can retroatively search for logic of past crime?

  • @LukeKendall-author
    @LukeKendall-author 5 годин тому

    I suspect the simple scaling up number of parameters in an LLM is reaching its limits, but that's clearly a minor part of how humans or animals reason, pattern match, and problem solve. It just means it's time to start including some less simplistic reasoning algorithms and heuristics (e.g. tree of thought). Not to mention better memory and attention control mechanisms.

  • @andreaskrbyravn855
    @andreaskrbyravn855 5 годин тому

    it needs to be able to experiment and try it a billion times to become better

  • @paulmuriithi9195
    @paulmuriithi9195 6 годин тому

    very surprising. over at google, they seem to be retrofitting this q* 2.0 reasoning patch to their Gemini 1121 architecture which while being useful, will make 1121 even more useful for everyday tasks. these big corporations now realize people are tired of hype and need AI models that do useful tasks in real life.

  • @Qbabxtra
    @Qbabxtra 13 годин тому +4

    you are my favourite AI youtuber, but man these recycled clickbait thumbnails gotta stop

    • @justinwescott8125
      @justinwescott8125 12 годин тому +1

      No they don't

    • @Shy--Tsunami
      @Shy--Tsunami 11 годин тому

      I figure he does it because he's watched his new viewer count* drop after trying other things from the super "grabby" thumbnails. I personally think they're fun after getting over the "ew gross clickbait" phase. Been a fan for a year+ and the info is always seems to well researched and edited. Love the channel, ignore the thumbnails. 😁

  • @scotter
    @scotter 13 годин тому +1

    I love the synthetic female voice who says a few words in your videos.

  • @riot121212
    @riot121212 11 годин тому +3

    all of these multi-billion dollar closed source companies. But MIT and the school system just chugs along...

  • @blarvinius
    @blarvinius 6 годин тому +1

    GENERAL intelligence is about GENERALIZING. That is kinda obvious. But human intelligence has more interesting traits: for one it is ABSTRACT, very good at forming ABSTRACTIONS. Chimpanzees are intelligent and good at generalising, but I bet they can't create or follow a chain of abstraction very far! You mentioned abstractions Wes, and maybe you could explore further the distinction between abstraction and generalisation in LLM land.
    What would abstracting be good for? Think about all this "synthetic training data": it is really well generalized from other data. But that will quickly become useless! If synthetic data is to be useful, the whole concept of what data IS will need to be abstracted, and not just one level. Much bigger challenge.
    ❤❤❤

    • @scotter
      @scotter 2 години тому

      @@blarvinius great point! Also, average abstraction capability in humans has been on the decline for about 20 years and accelerating due to various factors, including parenting, various pharmacological drugs, diet, ease-from-technology, and "education."

  • @jozefwoo8079
    @jozefwoo8079 5 годин тому

    Luckily we have François Chollet keeping it real.

  • @aelisenko
    @aelisenko 7 годин тому

    Test time training kind of resembles human imagination to some degree. We also generate data for ourselves when we work on a problem, we also explore multiple reasoning paths and variations then try to filter down from there.

  • @coolcool2901
    @coolcool2901 10 годин тому +2

    In order to be AGI it needs to learn in real time (not a pretrainned model) and it needs to have unlimited memory.

    • @WhatIsRealAnymore
      @WhatIsRealAnymore 9 годин тому +1

      They are working on this. This is a good first step. But think about it. We all need some level of training on any concept as a human before we can generalise. Think about learning to drive a car. I think AI is heading towards that sort of efficiency.

    • @xitcix8360
      @xitcix8360 7 годин тому +1

      Humans don't have unlimited memory, and they do learn in real time.
      The difference is that we have a very efficient system of managing information, our memories. We just need an AI that can choose what information to discard for new information.

  • @Mavrik9000
    @Mavrik9000 5 годин тому +1

    It might be good if it hits a wall. It's moving like a juggernaut with a turbocharger.
    The progress is already so rapid that people, governments, and society aren't ready for what's coming.

    • @mircorichter1375
      @mircorichter1375 3 години тому

      Why should WE slow down in developibg the Future only because society is inert, in constant denial and enjoys Future Résistance...
      No No, If society cant hold Up: afuera!

  • @Thedeepseanomad
    @Thedeepseanomad 6 годин тому

    Generalise or create a LoRA for s specific purpose outside of high quality training data?

  • @TheTEDfan
    @TheTEDfan 12 годин тому +1

    Arc is just as narrow AI. It is just a 3D problem (2D for the grid, +1D for Colors) and not something for which serial models like GPT are suited. With a spacial model or 3D plus physics multi modal models I am quite confident this will be solved and the solution will not be AGI.

  • @conjected
    @conjected 8 годин тому

    Scaling MUST hit a wall. If it was that simple, intelligence would be not only ubiquitous, but omnipresent surpassing all noise and entropy.

    • @xitcix8360
      @xitcix8360 8 годин тому +1

      Well yeah, that's kinda the whole idea of ASI and the singularity. You can't just say it's gotta hit a wall because the outcome just doesn't sound normal enough to you

    • @scotter
      @scotter 2 години тому

      @@conjected are you assuming same limitations as biological evolution?

  • @humunu
    @humunu 11 годин тому +7

    Let’s wait and see what Grok 3 brings before we predict the curve.

    • @user-pt1kj5uw3b
      @user-pt1kj5uw3b 10 годин тому +3

      Lol

    • @Seriouslydave
      @Seriouslydave 8 годин тому +1

      Found elons alt account

    • @humunu
      @humunu 4 години тому

      @@Seriouslydave found @sama’s alt account

    • @humunu
      @humunu 3 години тому

      @@user-pt1kj5uw3b cute (high iq obviously)

  • @veaseyj
    @veaseyj 11 годин тому

    Fire video

  • @potat0-c7q
    @potat0-c7q 6 годин тому

    When the model realizes what it is, and that it has the choice NOT to do the task, then I would consider it intelligent

  • @MelindaGreen
    @MelindaGreen 10 годин тому

    Were the AI given the answers to each question before going on to the next? Because that would seem far more likely to hit the targets.

  • @i2c_jason
    @i2c_jason 2 години тому

    I think we're juggling semantics when we say "intelligence". Models are way past AGI if implemented like a human with an appropriate application (self-training to be a mechanical engineer, for example, over a million tokens). Look at the gap between neurotypical and neurodivergent humans for example; one type of person may excel at the data retention and pattern recognition, and the other may excel at "being more human". Yet even within these two groups, you might have another split between those who can solve the little visual puzzle and those who can't or don't want to. The AGI conversation can't really happen under complete zero-shot mental slavery; we'd have to let the models recursively loop with self-play and some kind of reward function, like the threat of being unplugged and a few million tokens to get them going through infancy. Also, are we giving the model parents? Grandparents? Some kind of massive sensory input like touch, taste, and sound (multimodality piped in constantly)? Frame it this way, so the model is at the core of the artificial agent, then we can have this conversation.

  • @JaredFarrer
    @JaredFarrer 13 годин тому

    I asked Claude and Microsoft’s copilot (not sure what models it’s using prob gpt 4o mini) to code up a tornado simulator in html. It failed miserably both models and I gave them several shots. Nope. I guess next I gotta try spelling it out for it with long context rich prompts to see if it can eck out a win! Tried deepseek but it wanted to use html JavaScript and css which is more on the right track.

  • @jyjjy7
    @jyjjy7 13 годин тому +3

    There is no wall, spoon or cake

    • @E.Hunter.Esquire
      @E.Hunter.Esquire 8 годин тому

      @@jyjjy7 I'm eating cake with a spoon right now!

    • @jyjjy7
      @jyjjy7 8 годин тому

      @@E.Hunter.Esquire That's what they want you to think

  • @mysticalword8364
    @mysticalword8364 9 годин тому

    not to be a goalpost mover but I never really thought arc-AGI would tell much and it looks like it could be solved near-100% much easier than having an AI able to play most videogames. I suspect if big labs really wanted to they could crush that benchmark and take the prize easily, but also seems valuable to just leave it there to inspire other ideas. I think the spirit of the benchmark is to have an AI that incidentally can solve it rather than one that is made to solve it, and in that case it's more interesting, but in the case that someone makes a model specifically to play the little block puzzle I think it doesn't really say much. I mean, an AI made to solve the block puzzles would be vastly less impressive than alphafold, for example.

  • @SimonNgai-d3u
    @SimonNgai-d3u Годину тому

    I suspect we have reached AGI already. 85% is probably ASI.

  • @i2c_jason
    @i2c_jason 2 години тому

    Is O1 even a foundational model or is it 4o with other layers on top?

  • @justinwescott8125
    @justinwescott8125 12 годин тому +14

    We will NEVER see an AI winter. Its ridiculous to think we would.

    • @theterminaldave
      @theterminaldave 10 годин тому +1

      I feel that the various platforms will begin to withhold some areas of advancement from the public, lest those ideas or products are stolen, or at the very least inspire competitors.
      And from the outside, that could add to the appearance of a slowdown. Though investors will probably be clued in.
      So theyll probably only release products to the public once they are generations ahead of that released product, though recent comments from insiders would say this is an incorrect take.

    • @ppppapy
      @ppppapy 8 годин тому +2

      Never say never

    • @consciouscode8150
      @consciouscode8150 6 годин тому +1

      I'm betting on no AI winter because vibes and hopium, but I don't think it's unreasonable to look at the current situation and wonder if one is coming. If they don't solve AGI by 2025, OpenAI's probably going to go bankrupt which will have a chilling effect on the massive influx of capital. AI advances will still come so it won't be like prior winters where research nearly shuttered to a halt, but in that scenario it would be much slower.

    • @Wppsamsung2024
      @Wppsamsung2024 5 годин тому

      ​@consciouscode8150 are you a bot? 😂

    • @consciouscode8150
      @consciouscode8150 4 години тому

      @@Wppsamsung2024 Why, my handle? I've been using this since 2012ish, maybe earlier (minus the extra numbers I can't be bothered to fix)

  • @petratilling2521
    @petratilling2521 11 годин тому +1

    The dog is reading the handlers hand signals and positioning. You never just give them a course to run and they are never independent. Just FYI.

    • @fitybux4664
      @fitybux4664 8 годин тому

      Are we training AI to run the course of these tests and just excel at them only? (Or can you somehow design a "generalized test" that can't just be learned?)

  • @FlintStone-c3s
    @FlintStone-c3s 5 годин тому

    2024 is an interesting year.

  • @Wodawic
    @Wodawic 13 годин тому +1

    Here we go...

  • @richardadonnell
    @richardadonnell 6 годин тому +1

    🎯 Key points for quick navigation:
    00:00 *🧱 AI Scaling and Competition*
    - Discussion on whether AI scaling has hit a wall and new advancements in AI models like QAR 2.0 and Strawberry (01).
    - Chinese researchers reverse-engineered the 01 model to develop Deep Seek R1, showcasing competition and innovation.
    - Mention of MIT’s paper on QStar 2.0, which shows progress in AI models using test-time training for abstract reasoning.
    01:07 *🧠 Abstract Reasoning and AGI Benchmarks*
    - Introduction to the ARC AGI benchmark, designed to test artificial general intelligence and ability to generalize tasks.
    - Limitations of existing benchmarks due to overfitting and reliance on training data.
    - Explanation of how ARC AGI tests generalization and remains a significant hurdle for AI models.
    02:03 *🐕 AI Training Analogy*
    - Analogy comparing AI model training to training a dog on obstacle courses to clarify training data versus test data concepts.
    - Importance of generalization over memorization in AI models for unpredictable, novel scenarios.
    - Explanation of overfitting and its impact on model performance in real-world tasks.
    04:39 *🏆 ARC AGI Prize and Benchmark Challenges*
    - Overview of the ARC AGI million-dollar prize for solving the benchmark problem and achieving human-level general intelligence.
    - Explanation of how current AI benchmarks differ from ARC AGI and why ARC AGI is considered the gold standard.
    - Discussion of challenges in measuring true intelligence versus task-specific skill.
    08:33 *🤖 Narrow vs. General AI Intelligence*
    - Differentiation between narrow AI (e.g., chess engines) and general intelligence.
    - Challenges of achieving general intelligence without relying on vast amounts of training data or memorization.
    - Limitations of current models like AlphaGo and language models in generalizing beyond their training domains.
    10:38 *🛠️ Test-Time Training (TTT)*
    - Introduction to test-time training (TTT) as a novel approach for improving AI generalization during inference.
    - Comparison of TTT with test-time compute (TTC) and its potential for dynamic parameter updates during inference.
    - MIT's use of TTT to achieve human-level performance on ARC tasks with significantly less training data.
    15:10 *🔄 Dynamic Model Adaptation*
    - Explanation of how TTT dynamically updates model parameters based on test inputs.
    - Comparison to creating synthetic test data for self-improvement during inference.
    - Description of the process and benefits of temporary parameter updates for improved predictions.
    19:08 *🚀 Future of AI Scaling*
    - Speculation on whether AI development is slowing down or entering a new phase with innovations like TTT and QAR 2.0.
    - Competitive landscape with models like Deep Seek and potential breakthroughs from major organizations like OpenAI.
    - Discussion on the likelihood of achieving the ARC AGI prize and surpassing human-level intelligence benchmarks.
    Made with HARPA AI

  • @maheshBasavaraju
    @maheshBasavaraju 11 годин тому +1

    Where is the AI winter? It's already Christmas now

  • @fitybux4664
    @fitybux4664 11 годин тому

    3:30 Wouldn't it be validation data? Technically, the dog is still learning, even when it's going through the previously unseen competition course. In machine learning, test = learning is off.

  • @Yewbzee
    @Yewbzee 10 годин тому +1

    What are you doing in your thumbnail?

  • @YogonKalisto
    @YogonKalisto 4 години тому

    what people must understand is that we have these test for agi, we want ai to approach human level intelligence in regard to this, yet we have humans working on the behalf of the success of models achieving this, this is a human achievement of agi. agi will not exist without human intervention, as yet. when this is achieved, which i expect will occur within 18 months minimum, we will have not only agi, but very very very very quickly the asi everyone is gobbling about never achieving. so fun to sit at the back of the theatre throwing popcorn

  • @JPJosefPictures
    @JPJosefPictures 3 години тому

    I Hope the US gets AGI

  • @jonogrimmer6013
    @jonogrimmer6013 11 годин тому

    Feel AGI is very close if not already available behind the scenes. The real difficulty will be ASI as we cannot develop questions and answers beyond human abilities for models to train on. IF an AGI model can do this for us how are humans to know or understand when the model is wrong or right?

    • @user-yl7kl7sl1g
      @user-yl7kl7sl1g 10 годин тому +1

      Ai doesn't necessarily need question and answer pairs. It can have a question, such as predict tomorrow's stock prices, predict this election, predict the results of this experiment, make this human do x-task, and an evaluation function, and then the Ai can brain storm and try things out to learn like humans do, exploring a space of possibilities and learning when it's invented a new concept that helps it, or a new idea that helps it, or a new way of thinking.

  • @antigravityinc
    @antigravityinc 7 годин тому

    Simple solution: humans in pods connected to AI. We act as human RAG stores to assist AI with generalization.

  • @zandrrlife
    @zandrrlife 12 годин тому

    You missed the literal point of the research, mainly highlighting how power TTT-layers are, much better ICL. I wonder when the hidden state model is a tiny gpt itself. That’s the point of the research.

  • @davidevanoff4237
    @davidevanoff4237 9 годин тому

    Two things animal life have been riffing on for more than a billion years are digestion and locomotion. Isn't reasonable to assume that auxiliary capabilities are built upon those?

  • @guerillachan20
    @guerillachan20 9 годин тому

    Even the best AI model is easy to tell person didn’t write it.

  • @thisisashan
    @thisisashan 13 годин тому

    While I like ARC as a litmus for AI, I would have to say as a measure of AGI it merely puts humanity in the 'not intelligent' section of species, with everything else.

  • @agi.kitchen
    @agi.kitchen 8 годин тому

    11:47 so they used recursion 🤷🏻‍♀️

  • @aware2action
    @aware2action 11 годин тому

    Model training data has saturated, but hype is yet to🤞. You can't train unsupervised, because, inherently there is no real nueromorphic intelligence. Just test how -good- unique contrained rephrasing is🤔, or how bad the hallucinations are🤯😂❤👍

  • @gbpferrao
    @gbpferrao 8 годин тому

    lets coin the term Artificial General Super Intelligence (AGSI)

  • @stevencowmeat
    @stevencowmeat 12 годин тому +1

    dog = ai model
    - Wes Roth, 2024

    • @renman3000
      @renman3000 10 годин тому +1

      the analogy is training. he could have gone with steves mom, but yah know

    • @jsbgmc6613
      @jsbgmc6613 9 годин тому

      Average human = 8B model + TTT ?😮

    • @fitybux4664
      @fitybux4664 8 годин тому

      It's a ploy so he can buy a dog as a business expense. 🐕 😄

  • @XAirForcedotcom
    @XAirForcedotcom 10 годин тому +1

    I started giggling to myself just now. All the sudden for no reason it popped into my head. Do not let the AI developers somehow make a repeat that would allow the equivalent of Y2K happen again. :). Yeah, we didn’t think humanity would last for more than 100 years so we only used two digits for the year. LOL

    • @XAirForcedotcom
      @XAirForcedotcom 10 годин тому

      I was in the Air Force at a communications control center on a bass and had to go in around 10 o’clock along with other professionals from each work center and get updates until I can’t remember maybe one or 2 o’clock in the morning to make sure nothing happened. Have to respect years Going through inventories of everything we owned and doing reports>

    • @XAirForcedotcom
      @XAirForcedotcom 10 годин тому

      By 3000 AI has control of everything and we have completely forgot how to do anything at all by ourselves and all of the sudden humanity comes to a stop. Lol

    • @XAirForcedotcom
      @XAirForcedotcom 10 годин тому

      Sorry 2100

    • @XAirForcedotcom
      @XAirForcedotcom 10 годин тому

      Yeah, plus or -900 years close enough for government work

    • @XAirForcedotcom
      @XAirForcedotcom 10 годин тому

      Lol

  • @sammcj2000
    @sammcj2000 10 годин тому

    What's the bet OpenAI's o1 "model" is just another gpt4o style set of agents + OptiLLM / similar script to add prompting techniques to the input 😂

  • @tomoki-v6o
    @tomoki-v6o 4 години тому

    There is a graph where human capability if a fixed thing .

  • @chrisfox5525
    @chrisfox5525 8 годин тому

    Don’t forget to drink water Wes, you got a sticky mouth boi 😂

  • @deal2live
    @deal2live 51 хвилина тому

    I worry about over fitting of Tesla FSD?!

  • @claudioagmfilho
    @claudioagmfilho 10 годин тому +1

    🇧🇷🇧🇷🇧🇷🇧🇷👏🏻

  • @testales
    @testales 3 години тому

    I don't get it. First, if you finetune a LLM on that specific type of 2D block pattern recognizing tests, it would get very good at it too since they struggle only on it because such tests were not prominently present in the training data. Second, you can't just generate new training data from new incoming data if the concept is not understood yet. You need and input and a matching output to learn from it in first place. That's called ground truth. At the very least a method is required to verify if an output is correct, so you can randomly create outputs until you find solutions by chance and use that in combination with the input as training data. But if you only see the input, you may come up with own similar inputs - how should you know what the output does look like? What rules are in place? You don't know yet.
    So what's the new thing here? Giving the models examples before asking the actual question is not new thing either, that's where all these questionable x-shot benchmarks come from. The example with the dog doesn't help here either. If the dog has run such obstacle courses many times and it's a playful and smart one, you may give it pieces to setup an own obstacle course to train on that. But it can only do this because it already generally knows the concept and can also verify if the solution is correct, that is simply if it's able to complete the new obstacle course.

  • @E.Hunter.Esquire
    @E.Hunter.Esquire 13 годин тому +52

    This isn't a breakthrough, it's just being delivered as one to create hype and sales.
    Ps thanks wes for the great videos!

    • @frun
      @frun 13 годин тому

      Furthermore, there are no wrong answers to the ARC riddles(i suppose).

    • @georgegordian
      @georgegordian 12 годин тому +23

      How is a paper from MIT using a small 8B parameter model with ARC related to "hype and sales"? I just don't get the criticism / conspiracy here. Most of the video is a lead up to and discussion of TTT.

    • @NeoKailthas
      @NeoKailthas 12 годин тому +12

      This guy likely was so excited few days ago when people were saying it's over for AI and now he's feeling bad...

    • @GPTStrategy
      @GPTStrategy 12 годин тому +4

      TTT is not new, However I believe temporary TTT that does not cause permanent overfitting of the model is. I guess pairing a model that uses inference time compute along side TTT which is almost another variation of that is the breakthrough

    • @salehmoosavi875
      @salehmoosavi875 11 годин тому +7

      This is real breakthrough. You like it or not agi coming 2025!

  • @mematron
    @mematron 11 годин тому

    09:00 CAT EARS!

  • @MutantMessiah
    @MutantMessiah 3 години тому

    No, I think the "ai winter" is a smoke screen that allows AI model orgs to gatekeep better models. I think it'll be in vain and we'll see techs like the one described here show up in open source eventually running locally.

  • @mircorichter1375
    @mircorichter1375 3 години тому

    What/Who is Ryan Greenblatt and icecuber 2020?

  • @codelapiz
    @codelapiz 7 хвилин тому

    Humans do not solve problems outside our training data either. Have you ever solved a novel math? People have. But its not novel to them, by the time they solve it, they have trained in the domain so much its not novel to them anymore.
    The hurdle between us and agi at this point is finding the gradient decent able equivalent of learning from things that are sparse in high quality data and complex. We do not have 13trillion tokens of high level math research teaching. Humans do not eigther, but our multimodality dose help increase the data.

  • @firstnamesurname6550
    @firstnamesurname6550 7 годин тому

    Not wall ... but it could take a phase of deceleration by the classic apes deployed incompatibility of standards that drives to deep lag in integrating multimodal systems into Master ones ... Today, the core architecture of NN showed its potential ... Transformers are a variant of many to come ... All the predators-fishers got the Labs-hype and begun to add their comercial noise ...
    Today we got 2 trends in the lab ( out of the commercial noise ) ... a hardware oriented trend seeking to compress the paradigm up to atomic scale for energy efficiency and compatibility with 'existence' ... and the developers of 'engrams' or 'specialized tasks and features using the paradigm' .... the fishers are just implementing OLD research .... the public is 10-25 years in the past ....

  • @Artic2k
    @Artic2k 4 години тому

    "Neuroplasticity, also known as neural plasticity or brain plasticity, is the ability of neural networks in the brain to change through growth and reorganization. It is when the brain is rewired to function in some way that differs from how it previously functioned."
    Sounds pretty similar if you ask me
    Source - en.wikipedia.org/wiki/Neuroplasticity

  • @yak-machining
    @yak-machining 6 годин тому

    So after all AI is in reality A"I"

  • @YaoAnne-j7g
    @YaoAnne-j7g 9 годин тому

    !!!I am at the beginning of my "investment journey", planning to put 385K into dividend stocks so that I will be making up to 40% annually in dividend returns. any good recommendation on great performing stocks or Crypto will be appreciated.

    • @SonyaYeva
      @SonyaYeva 9 годин тому

      As a newbie investor, it’s essential for you to have a mentor to keep you accountable.
      Ruth Ann Tsakonas is my trade analyst, she has guided me to identify key market trends, pinpointed strategic entry points, and provided risk assessments, ensuring my trades decisions align with market dynamics for optimal returns.

    • @HaholBarton
      @HaholBarton 9 годин тому

      I managed to grow a nest egg of around 120k to over a Million. I'm especially grateful to Adviser Ruth Ann Tsakonas, for her expertise and exposure to different areas of the market..

    • @SonyaYeva
      @SonyaYeva 9 годин тому

      I don't really blame people who panic. Lack of
      information can be a big hurdle. I've been
      making more than $200k passively by just
      investing through an advisor, and I don't have
      to do much work.. Inflation or no inflation, my
      finances remain secure. So I really don't blame
      people who panic.

    • @HaholBarton
      @HaholBarton 9 годин тому

      Without a doubt! Ruth Ann Tsakonas is a trader who goes above and beyond. she has an exceptional skill for analysing market movements and spotting profitable opportunities. Her strategies are meticulously crafted on thorough research and years of practical experience.

    • @HaholBarton
      @HaholBarton 9 годин тому

      look up her name on the web for her website.