YouTube

Got a YouTube account?

New: enable viewer-created translations and captions on your YouTube channel!

English subtitles

← True Artificial Intelligence will change everything | Juergen Schmidhuber | TEDxLakeComo

Get Embed Code
1 Language

Download

Showing Revision 7 created 10/25/2018 by Robert Clarke.

  1. [Music]

  2. When I was a boy, I wanted to
    maximise my impact on the world,
  3. and I was smart enough to realise
    that I am not very smart.
  4. And that I have to build a machine that
    learns to become much smarter than myself,
  5. such that it can solve all the problems
    that I cannot solve myself,
  6. and I can retire.
  7. And my first publication on
    that dates back 30 years: 1987.
  8. My diploma thesis, where I already try
    to solve the grand problem of AI,
  9. not only build a machine that learns a
    little bit here, learns a little bit there,
  10. but also learns to improve
    the learning algorithm itself.
  11. And the way it learns, the way
    it learns and so on recursively,
  12. without any limits except
    the limits of logics and physics.
  13. And, I'm still working
    on the same old thing,
  14. and I'm still pretty much
    saying the same thing,
  15. except, that now, more people
    are listening.
  16. Because, the learning algorithms that we
    have developed on the way to this goal,
  17. they are now on three thousand
    million smartphones.
  18. And all of you have them
    in your pockets.
  19. What you see here are the five most
    valuable companies of the Western world:
  20. Apple, Google, Facebook,
    Microsoft, and Amazon.
  21. And all of them are emphasising that AI,
    artificial intelligence,
  22. is central to what they are doing.
  23. And all of them,
    are using heavily,
  24. the deep learning methods that my team
    has developed since the early nineties,
  25. in Munich and in Switzerland.
  26. Especially something which is called:
    the long short-term memory.
  27. Has anybody in this room ever heard
    of the long short-term memory?
  28. Or the LSTM? Hands up,
    anybody ever heard of that?
  29. Okay. Has anybody never heard of
    the LSTM?
  30. I see we have a third group in this room:
    who didn't understand the question.
  31. The LSTM is a little bit like your brain:
  32. it's an artificial neural network
    which also has neurons,
  33. and in your brain, you've got about
    100 billion neurons.
  34. And each of them is connected to
    roughly 10,000 other neurons on average,
  35. Which means that you have got
    a million billion connections.
  36. And each of these connections has
    a strength which says,
  37. how much this neuron over here, influences
    that one over there at the next time step,
  38. And in the beginning all these connections
    are random and the system knows nothing,
  39. but then, through a smart learning algorithm,
    it learns from lots of examples,
  40. to translate the incoming data,
    such as video through the cameras,
  41. or audio through the microphones,
  42. or pain signals through the
    pain sensors.
  43. It learns to translate that
    into output actions,
  44. because some of these neurons are
    output neurons,
  45. that control speech muscles
    and finger muscles.
  46. And only through experience,
    it can learn to solve
  47. all kinds of interesting problems,
    such as driving a car
  48. or, do the speech recognition
    on your smartphone.
  49. Because, whenever you
    take out your smartphone,
  50. an Android phone, for example,
    and you speak to it,
  51. and you say: "Ok Google, show me
    the shortest way to Milano."
  52. Then it understands your speech,
  53. because there is a LSTM in there,
  54. which has learned to understand speech.
  55. Every 10 milliseconds,
    100 times a second,
  56. new inputs are coming
    from the microphone,
  57. and then translates it
  58. after thinking into letters which is then
    question to the search engine and it has
  59. long to do that by listening to lots of
    speech from women from me all kinds of
  60. people and that's how since
  61. 2015 Google speech recognition is
    now much better than it used to be
  62. the basic lsdm cell looks like that I
    don't have the time to explain that but at
  63. least I can list the names of
    the brilliant students in my lab who made
  64. that possible and what are
    the big companies doing with
  65. that well speech recognition is only
    one example if you are on Facebook is
  66. anybody on Facebook okay I
    use sometimes clicking at
  67. the translate button because somebody sent
    you something in a foreign language and
  68. then you can translate it is anybody
    doing that yeah whatever you do that you
  69. are waking up again
    a long short term memory and lsdm which
  70. has learned to translate text in
    one language into translated text and
  71. Facebook is doing that four billion times
    a day so every 50 every second
  72. 50,000 sentences are being translated by
    an LST am working for
  73. Facebook and another 50,000 in the
    second and another
  74. 50,000 and to see how much this thing
    is now permitting the modern world
  75. just note that almost 30 percent of
    the awesome computational power for
  76. interference and all these Google Data
  77. Centers all these data centers of Google
    are all over the world is used for LST on
  78. almost 30 percent if you have an
  79. Amazon echo you can ask a questions and it
    answers you and the voice that you hear
  80. it's not a recording it's
    an LS TM network which has learned from
  81. training examples to sound like
    a female voice if you have an iPhone and
  82. you're using
    the quick type it's trying to predict what
  83. you want to do next given
    all the previous context of what you did
  84. so far again that's an LS DM which
    has to do that so it's on
  85. a billion iPhones you are a large audience
    by my standards but when we started
  86. this work decades ago in the early
    90s only few people who were interested
  87. in that because computers were so slow and
    you couldn't do so much with it and I
  88. remember I gave a talk at
    a conference and there was just
  89. one single person in the audience
    a young lady I said young lady it's
  90. very embarrassing but apparently
    today I'm going to give this talk just to
  91. you and she said okay but please hurry I
    am the next speaker since then we
  92. have greatly profited from the fact
    that every five years computers again in
  93. ten times cheaper which is
    an old trend that has held since 1941 at
  94. least since this man Conrad Susan built
    the first working program control computer
  95. in Berlin and he could could do roughly
    one operation per second one and then
  96. ten years later for the same prize one
    could do 100 operations 30 years later
  97. 1 million operations were
    the same price and today after 75 years we
  98. can do a million billion times as much for
    the same price and the trend is not about
  99. to stop because the physical limits are
    much further out there rather soon and not
  100. so many years or decades we will for
    the first time have
  101. little computational devices that
    can compute as much as a human brain and
  102. this a trend doesn't break 50 years
    later there will be
  103. a little computational device for
    the same price that can compute as much as
  104. all 10 billion human brains taken
    together and there will not only be one of
  105. those devices but
    many many many everything
  106. is going to change already in
    2011 computers were fast enough such that
  107. our deep learning methods for
    the first time could achieve
  108. a superhuman pattern-recognition result and
    was the first superhuman result and
  109. the history of
    computer vision and back then computers
  110. were 20 times more expensive than today so
    today for the same price we can do
  111. 20 times as much and just a few five years
    ago five years ago when computers were
  112. 10 times more expensive than today we
    already could win for the first time
  113. medical imaging competitions what you see
    behind me is a slice through
  114. the female breast and the tissue that you
    see there has all kinds of
  115. cells and normally you need
    a trained doctor a trained the solid who
  116. is able to detect
    the dangerous cancer cells or
  117. pre-cancer cells now our stupid network
    knows nothing about cancer knows nothing
  118. about vision it knows nothing in
    the beginning but we can train it
  119. to imitate the human teacher
    the doctor and it became as good or better
  120. than the best competitors and
    very soon all of medical diagnosis
  121. is going to be superhuman and
    it's going to be mandatory because
  122. it's going to be so much better than
    the doctors after this all kinds of
  123. medical imaging startups
    were founded focusing just on this because
  124. it's so important we can also use lsdm
    to train robots one important thing I
  125. want to say is that we not only have
    systems that slavishly imitate what humans
  126. show them no we also have a eyes that set
    themselves their own goals and
  127. like little babies invent
    their own experiment to explore
  128. the world and to figure out what you
    can do in the world without a teacher and
  129. becoming more and
    more general problem solvers in
  130. the process by learning new skills on top
    of old skills and this is going to
  131. scale we call that artificial curiosity or
    a recent password is power plain
  132. learning to become a more and
    more general problems over by
  133. learning to invent like a scientist
    one new interesting goal after
  134. Nathan and and it's going to scale and I
    think in not so many years from now for
  135. the first time we are going to have
    an animal like
  136. AI you don't have that yet on the level of
    a little crowd which
  137. already can learn to use two worlds for
    example little monkey and once we have
  138. that it may take just a few decades to do
    the final step towards
  139. human level intelligence because
    technological evolution is about
  140. a million times a million times faster
    than biological evolution and
  141. biological evolution needed
    3.5 billion years to evolve a monkey
  142. a monkey from scratch but then just
    a few tens of millions of years
  143. afterwards to evolve
    human level intelligence we have
  144. a company which is called Mason's
    like birth in English
  145. Mason's but spelled in
    a different way which is trying to make
  146. this a reality and build
    the first true general and purpose AI at
  147. the moment almost all research in AI is
    very human centric and it's all about
  148. making human lives longer and
    healthier and easier and making humans
  149. more addicted to their smartphones but
    in the long run a eyes are going to
  150. especially the smart ones are going to set
    themselves their own goals and I have
  151. no doubt in my mind that they
    are going to become much smarter than we
  152. are and what are they going to do of
    course they are going to realize what we
  153. have realized a long time ago namely that
    most of the resources in
  154. the solar system or in general are not in
    our little biosphere they are out there in
  155. space and so of course they
    are going to emigrate and of course they
  156. are going to use trillions of
    self-replicating robot factories to expand
  157. in form of growing
  158. AI bubble which within a few
    hundred thousand years is going to cover
  159. the entire galaxy by senders and receivers
    such that a eyes can travel the way they
  160. are already traveling in my lab by radio
    from sender to receiver Wireless so what
  161. we are witnessing now is much more than
    just another Industrial Revolution this is
  162. something that transcends
    humankind and even life itself
  163. the last time something so important
    has happened was maybe 3.5 billion years
  164. ago when life was invented a new type of
    life is going to emerge from
  165. our little planet and
    it's going to colonize and transform
  166. the entire universe the universe is
    still young it's only 13.8 billion years
  167. old it's going to become much older than
    that many times more many times older
  168. than that so there's plenty of time
    to reach all of it or all of
  169. the visible parts totally within
    the limits of light speed and physics
  170. a new type of life is going to make
    the universe intelligent now of course we
  171. are not going to remain the crown of
    creation of course not but there is still
  172. beauty in seeing yourself as part of
    a grander process that leads the cosmos
  173. from low complexity towards
    higher complexity it's a privilege to live
  174. at a time where we can
    witness the beginnings of that and where
  175. we can contribute something to
    that thank you for your patience