Why The Next AI Breakthroughs Will Be In Reasoning, Not Scaling

Поделиться
HTML-код
  • Опубликовано: 14 ноя 2024

Комментарии • 88

  • @chapterme
    @chapterme День назад +12

    Chapters (Powered by ChapterMe) -
    0:00 Intro
    1:15 The intelligence age
    4:18 YC o1 hackathon
    12:09 4 orders of magnitude
    14:42 The architecture of o1
    21:52 Getting that final 10-15% of accuracy
    32:06 The companies/ideas that should pivot because of o1
    34:44 Outro

  • @winddude9
    @winddude9 23 часа назад +83

    are they not allowed to talk about claude or something?

    • @brucebain7340
      @brucebain7340 19 часов назад +7

      They mentioned claude several times in the last episode genius

    • @vvmm3712
      @vvmm3712 19 часов назад +9

      ​@@brucebain7340Yeah they did but I think they are quite biased towards OpenAI. I have used premium versions of Gemini, OpenAI and Claude. Claude is still neck to neck, if not ahead of OpenAI. Also YC needs to be more objective in their POV. I think it happens because most of them talk to founders and read articles etc. rather than using the models extensively themselves to build stuff. Makes a lot of difference in how one perceives technology.

    • @brucebain7340
      @brucebain7340 19 часов назад

      Of course they are biased. They're human.
      Whoever expects them to be impartial will be disappointed.

    • @vvmm3712
      @vvmm3712 18 часов назад

      ​@@brucebain7340The bias is not about what to order for dinner. The bias is that company 'A' is going to unlock AGI, Company 'A's models are going to be orders of magnitude better because Mr. Sam told us so. This bias can lead to misjudgements on YC's part which can impact its own investments which run into millions. I mean won't it be sad that someone can simply walk into your office and skew your own vision of something.

    • @vvmm3712
      @vvmm3712 18 часов назад +3

      ​@@brucebain7340 The bias is not about what to order for dinner. The bias is also not from a colleague of mine who thinks that Apple built the Apple Intelligence(AI)!
      The bias is from a group of people running one of the topmost accelerators in the world.
      The bias is that company 'A' is going to unlock AGI, Company 'A's models are going to be orders of magnitude better because Mr. S told us so. This bias can lead to misjudgements on YC's part which can impact its own investment which runs into millions.

  • @counterfeit25
    @counterfeit25 20 часов назад +6

    4:30 Diode Computers is doing PCB (printed circuit board) design, not chip design (like NVIDIA)

  • @IntegrandoIA
    @IntegrandoIA Час назад +1

    Timing is a little off here, just as plateau conversations and supporting data keep coming up

  • @ZevUhuru
    @ZevUhuru 15 часов назад +6

    LOL. What is Gary's obsession with 'Raw Dogging', he's said this on multiple videos. Bro, stay safe out there 😂

  • @GNARGNARHEAD
    @GNARGNARHEAD 18 часов назад +4

    those hackathon results were really impressive 🤯

  • @princee9385
    @princee9385 19 часов назад +7

    Given all these developments, I need a remote job. PhD since 2022.

  • @yadniksable
    @yadniksable 4 часа назад +1

    Advice to everyone building a product. Ai is for scaling your solution your insight in a particular domain. So work on insight building and use ai model to scale that solution to solve the problem at a deeper level for target user.

  • @Manwith6secondmemory
    @Manwith6secondmemory 14 часов назад +4

    Bro omg, are we hitting a wall or not

  • @gpshangari
    @gpshangari 23 часа назад +12

    The people in chat against OpenAi have no idea what's going on. It's too late to stop AI, it's not a fad and it is not going away. There is no new fad. The only thing that kept humanity together was intelligence, now things are going to accelerate. But the benefits will only reach a few of course.

    • @realbillnye
      @realbillnye 15 часов назад +1

      It's more so that LLM cannot truly reason. The reasoning from o1 is chain of thought using the same LLM backend. This will likely not scale to AGI. I'm here for the ride, let's see

    • @drichards4426
      @drichards4426 12 часов назад +1

      Who’s saying it’s a fad ?

    • @MichaelBabich
      @MichaelBabich 56 минут назад

      LOL. Nobody needs to "stop AI" since there is NO AI at all. There is no intelligence in LLMs and in what OpenAI does. Yes, it's a nice tool that can process a lot of data and structure or recombine it one or another way but it has many many limitations that's impossible to overcome since they require totally different technology. It's not even close to anything that could be called AI.

  • @alonroth11
    @alonroth11 День назад +3

    Can you show the info you are looking at....

  • @varunvummadi350
    @varunvummadi350 4 часа назад

    Thanks a lot for shout out and adapting us Garry 😂

  • @Daniely-z2k
    @Daniely-z2k 4 часа назад

    I love the girl. she's so authentic. When she speaks, it feels like you're right there with her, having a conversation as if you're sitting together in a cozy coffee shop, where everyone feels like friends.

  • @Arcticwhir
    @Arcticwhir 15 часов назад +1

    very inspiring and exciting podcast to look forward to the future.

  • @NilsWestgardh
    @NilsWestgardh 22 часа назад +3

    The AI bot beating Dota 2 pros at The International put OpenAI on the map for me.

  • @jamesulan1
    @jamesulan1 4 часа назад

    Yeah, AI for customer support is solving something like 20-35% of most customers' support tickets. Tickets are usually the easier ones that come up most frequently.

    • @othmane-mezian
      @othmane-mezian 2 часа назад

      This could be more than 35%.. most common issues are often 70% to 80%

  • @vivek-singh-se
    @vivek-singh-se 22 часа назад +7

    Please share link to Sam Altman's essay referred early in the video.

    • @aaithubarla
      @aaithubarla 12 часов назад

      Bruh, Google it.

    • @jean-phil
      @jean-phil 8 часов назад

      google The Intelligence Age

    • @aaithubarla
      @aaithubarla 6 часов назад

      can you not google it? Is it that hard?

  • @Analyse_US
    @Analyse_US 20 часов назад +11

    Feels like alot of coolaid being chugged back here. I am personally becoming more skeptical by the day. Microsoft probably has invested the most money and time in this space and their AI driven products are under whelming. I am still wanting and waiting to see a compelling AI LLM driven productivity tool. All I am seeing is demos that include a high degree of deception. For example, the first Gemini duck demo, Devin Upwork demo, Tesla robots with human controllers etc. I am at the stage where I need to actually have access to the product, to believe any claims.

    • @vvmm3712
      @vvmm3712 19 часов назад +4

      Kinda agree..

    • @vdimension6300
      @vdimension6300 15 часов назад +3

      Same

    • @jean-phil
      @jean-phil 8 часов назад

      hmm there are already products using AI .. just look at google already generating 25% of their code by AI, think of all the AI features in Adobe Lightroom and other Adobe products, AI features in phones, self driving cars, AI finding 0 day hack .. I even used co-pilot this week at work to generate some reports .. and this is just the beginning

    • @winnerswritethestory3370
      @winnerswritethestory3370 7 часов назад

      you must be living under a rock then

    • @Analyse_US
      @Analyse_US 7 часов назад

      ​@@winnerswritethestory3370Name a successful commercial LLM product that isn't a base model.

  • @elliptictree
    @elliptictree 22 часа назад +4

    Interesting

  • @Dom-zy1qy
    @Dom-zy1qy 8 часов назад

    You want AGI? Heres how:
    1. Make an LLM good enough to create and implement highly accurate environments for use in RL. This should work with any arbitrary task.
    2. Train a good policy
    3. Profit?

  • @jonathanedwardgibson
    @jonathanedwardgibson День назад +2

    The mind expands, not stacks.

  • @netsurfer256
    @netsurfer256 20 часов назад +1

    lbh current models and APIs already can handle the scaling cases of use

  • @parvbhullar
    @parvbhullar 21 час назад +1

    Garry is obsessed with evals 😊

  • @sudheerkumarme
    @sudheerkumarme 23 часа назад +5

    I would like to see more Indian startups funded by YC. Please consider opening an YC India to invest in Indian startups.

    • @CardboardBoxed
      @CardboardBoxed 20 часов назад +2

      Indian startups and have some of the worst ROI. They’re usually not internationally trusted so many VCs won’t fund them. You’re better off looking for domestic investors.

  • @michaelocean4788
    @michaelocean4788 22 часа назад +3

    I get confused when Gary uses 'eval' to refer to testing, as it means something different in Python. Terms like 'metrics' or 'benchmarks' are more common in the LLM context and feel more precise.

    • @Dom-zy1qy
      @Dom-zy1qy 8 часов назад

      I see "eval" more commonly used than "metrics".

  • @soyhenryxyz
    @soyhenryxyz 19 часов назад +16

    "If an LLM task is hallucinating, it’s likely doing too much. Break it down into steps"

    • @pin65371
      @pin65371 13 часов назад

      The mistake people make when using these tools is they are not specific enough. Treat is like a human or team of humans and your results improve.

    • @ccash3290
      @ccash3290 13 часов назад +2

      ​@@pin65371 Treat it like a child; Adults are smart and can understand ambiguity and ask for clarity.

    • @cmw3737
      @cmw3737 4 часа назад

      It usually means that there are missing parameters that it has to guess. Instead it should understand ambiguity and be more interactive and say "I don't know" or "it depends" and ask for what it needs to answer the request.

  • @hongyihuang3560
    @hongyihuang3560 13 часов назад

    I do not believe that AI will be able to do chip design better than humans. PCBs maybe, chips no. It took the brightest people around the globe (literally any good chips you see today touches at least decades of research from Japan, Europe, US IP). There are so many specialties in chip design including testing, production ready, simulation, RF, documentation, security, compiler that’s beyond the capabilities of a closed loop LLM.
    If LLMs PCB design capability transfers to SoCs, China would have already made and beat Apple & NVIDIA.
    The first half is absolutely make believe, solve physics, nuclear power, climate… seriously? AI will solve societal issues? I find it hard to believe that next gen data centers for AI will need nuclear power.
    The second half I agree more on the real progress of AI: do tests, create a moat by building agents and accumulate proprietary data.
    Speculation is dangerous, I hope people can think for themselves.

  • @JS-mj9en
    @JS-mj9en 21 час назад +1

    AI should eliminate the need for customer support, so an AI customer support solution seems destined to fail

  • @qet-lab
    @qet-lab 3 часа назад

    We keep getting 101 from open AI & thats the problem.

  • @tonypeng8792
    @tonypeng8792 22 часа назад

    Amazing!

  • @no-wai
    @no-wai День назад +3

    I hope you guys are not blinded by Openai

  • @chychywoohoo
    @chychywoohoo 6 минут назад

    "thousands of days" maybe just say years

  • @vladimirbosinceanu5778
    @vladimirbosinceanu5778 6 часов назад

    more biomimicry --> intelligence on tap

  • @Mayeverycreaturefindhappiness
    @Mayeverycreaturefindhappiness 19 часов назад

    didn't Orion disappoint? I don't think we can just assume it will keep scaling. I am excited for the o series.

  • @DakshGuptaCuriosium
    @DakshGuptaCuriosium День назад

    shoutout atopile

  • @thehappydaysapp
    @thehappydaysapp 23 часа назад +4

    All this advancement is amazing but what I do not understand is, how is most of this advancement actually helping humanity? How is it helping the majority of humanity and not the one percent of investors?

    • @scienceinc.9453
      @scienceinc.9453 22 часа назад

      There's no law in physics to ensure that

    • @sprobertson
      @sprobertson 22 часа назад +1

      That's not what companies are for

    • @williamliu796
      @williamliu796 18 часов назад +1

      you can now talk to one of these LLMs and learn almost anything / ask questions about anything. Every kid with access to the internet can has a personal tutor for every subject for free or $20/month. I’d say humanity is being helped.

    • @zipytshorts
      @zipytshorts 17 часов назад +1

      you know when companies are more efficient that means you have cheaper products/services or any other goods

    • @AlexWilkinsonYYC
      @AlexWilkinsonYYC 16 часов назад

      Capitalism ... is for capitalists. 😉 It's in the name.

  • @AlexWilkinsonYYC
    @AlexWilkinsonYYC 16 часов назад

    4 people agreeing on everything is boring. Get a homeless guy in there or something.
    I can't wait for them to replace all the customer service agente with AI because ill make an AI call center that calls all their customer service centers and tries to manipulate them.
    Adoption isnt going to be super quick because having a surface area for attack thay massive is a huge liability.

  • @KamilHussainShaikh
    @KamilHussainShaikh 2 часа назад

    If AI reasons then people will start believing in God sooner.

  • @jvijayavallabh5869
    @jvijayavallabh5869 День назад +5

    First

  • @AntonioLaPlaca
    @AntonioLaPlaca 19 часов назад

    lol

  • @Төлеби
    @Төлеби 17 часов назад +1

    Diana Hu beautiful

  • @rustamkhujarustamov2929
    @rustamkhujarustamov2929 День назад

    third

  • @Drackomass
    @Drackomass День назад

    Second

  • @XX-pl9wp
    @XX-pl9wp 22 часа назад

    You should listen to the Marvin Minsky conversations from 80s AI Winter. Please stop talking about AGI until then. You're clueless!!!

  • @seanlive6975
    @seanlive6975 19 часов назад +1

    Still not seeing these tools achieve anything that humans are not capable of. Some efficiency gains maybe and useful as a learning tool. I think it's a limitation of the stochastic parrot from training data approach, it's never going to be creative and bring new innovations. That will need a new approach entirely.

    • @AlexWilkinsonYYC
      @AlexWilkinsonYYC 16 часов назад

      No way. I can't put bugs all over my code base nearly as fast as Cursor can.

    • @Carthodon
      @Carthodon 6 часов назад

      @@AlexWilkinsonYYC You surprised a lol out of me.

  • @man4hire
    @man4hire День назад

    Fifth

  • @rkara2
    @rkara2 20 часов назад

    AGI means you have solved the horizontal scalability problem. Which means having the ability to access any companies database public or private.
    So can one of you geniuses explain how exactly that is going to happen??
    My guess is that all of you are being duped by Altman because he has to say s**t like that to keep his investors satisfied and feeling good about themselves lol 😂