Introduction to GANs, NIPS 2016 | Ian Goodfellow, OpenAI

แชร์
ฝัง
  • เผยแพร่เมื่อ 23 ส.ค. 2017
  • NIPS 2016 Workshop on Adversarial Training
    www.iangoodfellow.com/slides/2...
    arxiv.org/abs/1701.00160
    Introduction to Generative Adversarial Networks

ความคิดเห็น • 63

  • @IKGchiller
    @IKGchiller 3 ปีที่แล้ว +127

    They made us read goodfellows papers in every single AI class and never bothered to mention that he's literally just our age

    • @matthewchunk3689
      @matthewchunk3689 3 ปีที่แล้ว +15

      it's like comparing zion williamson to your average college player: it will just hurt students' feelings

    • @techgamer1333
      @techgamer1333 ปีที่แล้ว +1

      Sad , But they are mentioning some lectures on youtube though🤣🤣😂😂 check his lectures on youtube😅😅

  • @joecincotta5805
    @joecincotta5805 ปีที่แล้ว +8

    His recall in the Q&A was pure gold. I want to channel that during PhD defence

  • @wireghost897
    @wireghost897 ปีที่แล้ว +9

    The Q&A section basically shows you how Ian thinks and learns stuff. He has absolute clarity of the maths behinds not just his work but the related work that he has studies. It is as he he did not just understand the maths, but he pretty much speaks maths like english.

    • @Gabcikovo
      @Gabcikovo ปีที่แล้ว

      Thank God

    • @Gabcikovo
      @Gabcikovo ปีที่แล้ว

      Copy, paste & upgrade.. Thank you.. 🇸🇰

  • @francoisdavid8487
    @francoisdavid8487 3 ปีที่แล้ว

    Super clear explanation! Awesome presentation

  • @sabrinad1922
    @sabrinad1922 5 ปีที่แล้ว +40

    From 26:52, I really liked his last comment about how to teach yourself anything. Really inspiring.

  • @PixelSenseiAvi
    @PixelSenseiAvi 5 ปีที่แล้ว +5

    The man in the house!!!!

  • @sureshothwarang
    @sureshothwarang 5 ปีที่แล้ว +5

    This dude is awesome

  • @mermich
    @mermich 7 หลายเดือนก่อน

    can I say that Ian Goodfellow is the GOAT in modern computer science? T.T

  • @mayukhmalidas
    @mayukhmalidas 3 ปีที่แล้ว +1

    Great session

  • @bright1402
    @bright1402 4 ปีที่แล้ว +36

    I think in the Q&A session, every sentence that Ian said could be a paper 😂

  • @dosomething3
    @dosomething3 6 ปีที่แล้ว +76

    “Let somebody else ask questions” - Crickets. .

  • @TusharKale9
    @TusharKale9 3 ปีที่แล้ว +2

    GANs very well explained. Great examples too. Thank you very much. Tushar

  • @hzg4013
    @hzg4013 3 ปีที่แล้ว +1

    what a legend!!!

  • @jackcuervo2262
    @jackcuervo2262 2 ปีที่แล้ว +1

    Wow Ian is awesome!

  • @arielgamboavazquez4250
    @arielgamboavazquez4250 5 ปีที่แล้ว +12

    Ianm you have always been a good fellow :v

  • @vikaspoddar9456
    @vikaspoddar9456 ปีที่แล้ว

    The Man, The Myth one n only Ian Goodfellow

  • @saikrishnapujari
    @saikrishnapujari 3 ปีที่แล้ว +4

    As this guy is now with Apple, any influence on their Memoji and Animoji or their Camera improvement using GAN ?

  • @lorforlinux
    @lorforlinux 5 ปีที่แล้ว +16

    26:50 - 28:10 is the best part

    • @zrmsraggot
      @zrmsraggot 5 ปีที่แล้ว

      Nah you are wrong but you feel that way because he tells you how you should learn, and that's obviously what you re/were trying to do at the time you watched the video.

  • @zheqingzhang6136
    @zheqingzhang6136 4 ปีที่แล้ว

    Abstract Expressionism 19:23

  • @igorcherepanov4765
    @igorcherepanov4765 6 ปีที่แล้ว +3

    do already exist any approaches to solve the problem with counting?

  • @alexbarnadas
    @alexbarnadas 3 ปีที่แล้ว

    Where was it?

  • @supreetkurdekar5726
    @supreetkurdekar5726 4 ปีที่แล้ว +3

    Ian Goodfellow is Jim Halpert!!!

  • @bingeltube
    @bingeltube 5 ปีที่แล้ว +1

    Recommendable

  • @generichuman_
    @generichuman_ 2 ปีที่แล้ว +3

    19:36 one way to fix this, is to not use the Chernobyl data set.

  • @quickpresent8987
    @quickpresent8987 ปีที่แล้ว

    how could this video "lan Goodfellow" are having less views than Siraj Raval video - * -

  • @MPXVM
    @MPXVM 5 ปีที่แล้ว +1

    9:14 - schrödinger's cat logic or Quantum state :) 50% - 50%

  • @Gabcikovo
    @Gabcikovo ปีที่แล้ว

    2:48 both players are neural networks

    • @Gabcikovo
      @Gabcikovo ปีที่แล้ว

      5:16

    • @Gabcikovo
      @Gabcikovo ปีที่แล้ว

      5:26 the goal of the generator is to fool the discriminator

    • @Gabcikovo
      @Gabcikovo ปีที่แล้ว +1

      5:30 eventually the generator is forced to produce data as true as possible

    • @Gabcikovo
      @Gabcikovo ปีที่แล้ว +1

      Uhrik and Putin cried bitter tears

    • @Gabcikovo
      @Gabcikovo ปีที่แล้ว

      3:10 one of the players is trained :) to do the best as possible on the worst possible input

  • @vanglequy7844
    @vanglequy7844 5 ปีที่แล้ว +15

    Here is my summary: So basically what he is trying to say is that let D tries to defeat G and vice versa, until the battle becomes so challenging for both sides. And then ML users can benefits from the skills of the two (D and G) to make cool things like fake images, fake videos, and fake news that are indistinguishable. And in many cases, these fake things are useful. Don't ask me what happens under the hood. Need to read his paper and some more background info. Brilliant work I must say, but not so good "introduction to GANs" for broad audience.

    • @phuccoiinkorea3341
      @phuccoiinkorea3341 5 ปีที่แล้ว

      cha hieu gi ong a =))

    • @redryan20000
      @redryan20000 5 ปีที่แล้ว

      Sounds pretty bad tbh

    • @generationai2211
      @generationai2211 4 ปีที่แล้ว +1

      Hi, this comment partially inspired a recent video I made, I agree that this video is hard for the “general public” to understand.

    • @agiisahebbnnwithnoobjectiv228
      @agiisahebbnnwithnoobjectiv228 3 ปีที่แล้ว

      The objective function of animal brain and hence AGI is to maximize impact. You heard it first from me. gj

  • @robelle1357
    @robelle1357 3 ปีที่แล้ว +1

    Everything was great except for the QA session. The questioner was rude.

    • @jackq2331
      @jackq2331 ปีที่แล้ว

      Not really,his question help us think.

  • @catsaresocute650
    @catsaresocute650 2 ปีที่แล้ว

    The last part is bs, the right price doesn't exist. It's always a proposal of worth and it's a proposal that can be accepted/rejected. That's what makes it work. They are all proposals of the value of something and every single marked participend is judge. That's the fucking point. NOT some ideal price.

  • @tzywg3399
    @tzywg3399 3 ปีที่แล้ว +2

    Down voters were scared by the fallout cow.

  • @wiiiiktor
    @wiiiiktor 5 ปีที่แล้ว +7

    i like Ian, but it's hard to follow his mundane voice tone ...

    • @stonetech1000
      @stonetech1000 5 ปีที่แล้ว +12

      I think his voice is just fine

    • @p.z.8355
      @p.z.8355 4 ปีที่แล้ว +1

      hes a robot ;)

    • @prathyusha5393
      @prathyusha5393 4 ปีที่แล้ว

      I had similar difficulty for quite some time ... But, I guess it's focus.. rather divided focus ..
      With a lot of practice... I am able to focus on the content at it's root, and a little more continually now ..
      Edit: Not talking about Ian here... Was speaking in general..
      Love listening to Ian s lectures

  • @kicksessentials7
    @kicksessentials7 3 ปีที่แล้ว

    The narrow wren analytically share because tortellini similarly smile upon a telling domain. unkempt, charming veil

  • @kulturfreund6631
    @kulturfreund6631 5 ปีที่แล้ว +3

    This is something for really intelligent people, not for me.

    • @BiancaAguglia
      @BiancaAguglia 4 ปีที่แล้ว +17

      This is hard for me to understand also, but only because I haven't studied enough math and ML to understand Ian. I'm sure you are the same: you're smart enough to understand it, but you haven't put in the work and time to learn ML. There's nothing stopping you from learning it, if you want. 😊

    • @DeViLTh0rn
      @DeViLTh0rn 4 ปีที่แล้ว

      @@BiancaAguglia i like your positive comment, do you have twitter?

    • @JousefM
      @JousefM 4 ปีที่แล้ว +4

      @@BiancaAguglia It's definitely not easy to get into the field if you really want to understand it and there are people grasping the concepts faster and more easily than you or I do but that should not demotivate us and keep working on the skill every day :)

    • @BiancaAguglia
      @BiancaAguglia 4 ปีที่แล้ว +1

      @@JousefM True. 😊

    • @shubhamchandra9258
      @shubhamchandra9258 3 ปีที่แล้ว +1

      @@JousefM You should be concerned about your own development only. You're good as long as you're learning and improving. Competition is an unnecessary burden.

  • @WaterMelonHead9922
    @WaterMelonHead9922 3 ปีที่แล้ว +1

    Instead of speaking in disjointed mathematical jargon, he could have went more into practical applications and how to test/troubleshoot the models by showing actual code. Most people won't get anything out of this

    • @manUNITED986
      @manUNITED986 ปีที่แล้ว +7

      That is absolutely nonsensical, if you understood what he was saying you can go out of your way to find architectures that define what he speaks of. Just say that you're lazy and move on.