Kendrew Lecture 2021 pt1 - Using AI to accelerate scientific discovery - Demis Hassabis

Поделиться
HTML-код
  • Опубликовано: 20 дек 2024

Комментарии •

  • @halneufmille
    @halneufmille 3 года назад +4

    AI research we often hear about mostly seems to be language related, computer graphics or recommending stuff to buy. Not saying it's not useful, but seeing Demis makes leaps and bounds in such fundamental and important applications is a breath of fresh air.

  • @quaidcarlobulloch9300
    @quaidcarlobulloch9300 2 года назад +1

    At 15:32 there’s that beautiful dynamic slide. I wonder if there’s any loops to previous plays representing maybe more of a gradient (not gradient decent mind you) instead of a tree..?

  • @sumankalyan70
    @sumankalyan70 3 года назад

    Fabulous talk

  • @alexmorehead6723
    @alexmorehead6723 3 года назад +5

    Fantastic talk!

    • @quaidcarlobulloch9300
      @quaidcarlobulloch9300 2 года назад +1

      @Alex you have all things AI in your home picture but no videos? If you’re watching this I’m curious what else you’ve found in the area

    • @alexmorehead6723
      @alexmorehead6723 2 года назад +1

      @@quaidcarlobulloch9300 haha, great catch my friend. That's just a preview of a future project I aim to bring to light soon

    • @quaidcarlobulloch9300
      @quaidcarlobulloch9300 2 года назад

      @@alexmorehead6723 Very cool! Are you coming at it from the hardware or software perspective?

    • @alexmorehead6723
      @alexmorehead6723 2 года назад

      @@quaidcarlobulloch9300 Most likely AI from the software perspective, but I'm not opposed to digging into the hardware of it all (I'm studying Computer Science at the PhD level, with an open mind towards hardware)

    • @quaidcarlobulloch9300
      @quaidcarlobulloch9300 2 года назад

      @@alexmorehead6723 Nice! Broad computer science with a focus on AI or broad AI with a focus on computer science? I’m considering getting a PhD in eecs for parallel processing that mimics the brain more closely, and software is definitely an interest!

  • @wubowm
    @wubowm 3 года назад

    This is truly a great talk!

  • @do_lee_di
    @do_lee_di 3 года назад

    Inspiring!

  • @ditomaximal
    @ditomaximal 3 года назад

    very inspiring work

  • @switchpointcokr
    @switchpointcokr 3 года назад

    중요하면서도 유머가 넘치는 훌륭한 강의!

  • @josephniepce7887
    @josephniepce7887 3 года назад

    I was here in November 2021 ;)

  • @TheGallowtree
    @TheGallowtree 2 года назад

    "Chess and Go, and one other game which I've forgotten..."
    How could you forget Starcraft?!

  • @quaidcarlobulloch9300
    @quaidcarlobulloch9300 2 года назад +1

    If the AI is based on a scientist then it should also be able to modulate its thinking time.

    • @quaidcarlobulloch9300
      @quaidcarlobulloch9300 2 года назад

      Furthermore/through temp outputs which could even indicate that it was thinking such as “hold on I need more time to think.” That temp output example could even be a preexisting output from which the AI learns that it gets more thinking time. But I guess that’s an observation it would be making in and of itself spending time on, ah recursion! But maybe it would learn that only once and use something like memory to avoid repetitive thoughts!

    • @romanisaev6616
      @romanisaev6616 Год назад

      uhoh

  • @joeneighbor
    @joeneighbor 3 года назад

    Next comes "Dr. Eldon Tyrell", fully genetically engineered organics :-P