Jay Alammar on LLMs, RAG, and AI Engineering

Поделиться
HTML-код
  • Опубликовано: 29 окт 2024

Комментарии • 28

  • @snarkyboojum
    @snarkyboojum 2 месяца назад +7

    Great to see a summary like this of how a real practitioner in the field is solving real problems with real companies. We do the same in my role and my teams. Not as sexy as the latest AGI hype or LLM research, but where the bulk of the “journeyman” GenAI work is for the next several years at least! Thanks for sharing Jay’s insights with us.

  • @DeniseMilaTeresa
    @DeniseMilaTeresa Месяц назад +26

    aistructuralreview AI fixes this. s, RAG, and AI Engineering.

  • @Darkon10199
    @Darkon10199 2 месяца назад +1

    Love this guy. Great discussion as usual, unlucky it was so "short"

  • @manslaughterinc.9135
    @manslaughterinc.9135 2 месяца назад +1

    Roughly around 46:00 regarding layered levels of complexity in a paper. I can't emphasize enough how many times I come back to specific papers and reread them, then walk away knowing more than the first time I read them because they sent me down a path that caused me to learn more, then 3 months later, I'm coming back and finding new gems, 3 months after that, doing the same thing. Every AI researcher should be taking notes.

  • @nasrawi11
    @nasrawi11 2 месяца назад +2

    Jay is a wonderful knowledgable person. Makes things simple . Thanks you.

  • @toadlguy
    @toadlguy 2 месяца назад +2

    Most amazing, Tim, how you seem to be right on top of what I am currently investigating 😄. I just started reading Jay's book on O'Reilly (pre-release) and you come out with this terrific interview, so I can get a more nuanced view of the author. Excited to read his latest, having gotten my earlier understanding of how transformers work from his 2017 paper. Thanks once again, Tim.

  • @CodexPermutatio
    @CodexPermutatio 2 месяца назад +2

    Top quality content as always. My favorite AI channel.

  • @muhannadobeidat
    @muhannadobeidat 2 месяца назад +1

    Excellent video, touches on many different topics, has great practical advice, pointers to education material, etc… Jay is talented and doing this purely out of passion for technology and innovation.

  • @michaelstewart529
    @michaelstewart529 2 месяца назад

    He certainly said a lot of words.

  • @saulyarhi675
    @saulyarhi675 2 месяца назад

    Thank you so much for all these knowledge so well explained 🎉❤

  • @arowindahouse
    @arowindahouse 2 месяца назад

    Thank you, Tim. This video brought me some illusion regarding the future uses of LLMs; sometimes is easy to think they're just stupid and useless

  • @paulk6900
    @paulk6900 2 месяца назад +5

    Can you please focus again on more complex topics that are like u previously? I still appreciate ur content but somehow it gets quite repetitive and deviates from the technical side to more and more soft topics

  • @mahmoudelaskare4982
    @mahmoudelaskare4982 2 месяца назад

    great talk

  • @tacticalgaryvrgamer8913
    @tacticalgaryvrgamer8913 2 месяца назад +2

    So basic is the material that while I don't think anything in here is wrong I didn't learn anything either

  • @lexer_
    @lexer_ 2 месяца назад +4

    I usually like these conversations because they allow you to meet the person instead of their sales mask. But in this one he never really took his mask off at all. I just listened to an hour of just marketing and I am honestly pretty disappointed.

  • @LuigiSimoncini
    @LuigiSimoncini 2 месяца назад +15

    Sobering content. AGI's nowhere to be seen, Cohere building a path for soft landing from the heights of the hypeness, real business cases still missing

    • @xthesayuri5756
      @xthesayuri5756 2 месяца назад +4

      We haven't even seen GPT-5, Claude-4 or Llama-4 and you are already declaring defeat based on one single generation of LLMs 😂💀

    • @Tommy31416
      @Tommy31416 2 месяца назад +2

      @LuigiSimoncini agree. Tech bros think a 60% success rate of an application with an LLM in it is great. In the corporate world that is a harbinger of death strategically, procedurally, commercially and/or legally. The idea of stable applications, enterprise grade i.e. 99.99% stability, that contains an LLM has always been a laughable concept. Tech bros gonna tech bro.

    • @Manwith6secondmemory
      @Manwith6secondmemory 2 месяца назад

      @@Tommy31416 AGI or not, current SOTA models in a much more computationally efficient and scalable/integrated format is enough to change society as a whole let alone automate large parts of most people’s jobs. Its already changing society (people becoming emotionally attached romantic bots for example)
      Also, humans operate at 99% accuracy. Are you joking? Humans are error prone to the highest degree. Everywhere, every system is riddled with inefficiencies and constant failures. Failure is fundamental to our condition we get better from it (sometimes we dont tho)
      We are also have very resource intensive logic engines that need an 8 hour reboot every 12 hours or they breakdown

    • @francisco444
      @francisco444 2 месяца назад +1

      There are so many use cases already. Have you not seen the waves of layoffs?

    • @pedro336
      @pedro336 2 месяца назад

      @@francisco444 Layoffs are mostly related to the end of low interest rates. This results in less people being hired, more layoffs etc. And this results in even more layoffs because there is now more talent on the market, for potentially lower salaries. So companies can purge their low performers and replace them with better/cheaper ones.

  • @bwowekeith4472
    @bwowekeith4472 2 месяца назад +1

    Second

  • @AliMoeeny
    @AliMoeeny 2 месяца назад

    zeroth 😏

  • @renjithravindran5018
    @renjithravindran5018 2 месяца назад +1

    First!😂

  • @KafkaUnderTheTree
    @KafkaUnderTheTree 2 месяца назад

    Third :D