SlideShare a Scribd company logo
1 of 56
Download to read offline
Code	
  Biology	
  and	
  (the	
  future	
  of)	
  	
  
Ar5ficial	
  Intelligence	
  
	
  
Joachim	
  De	
  Beule	
  
Recent	
  advances	
  in	
  AI	
  
	
   	
   	
   	
   	
   	
  Deep	
  learning	
  
A	
  dark	
  future	
  
	
  	
  	
  	
  	
  	
  	
  Superintelligences	
  more	
  dangerous	
  	
  	
  	
  	
  
	
  	
  	
  	
  	
  	
  	
  than	
  nukes	
  
A	
  brighter	
  future	
  
	
   	
   	
  	
  	
  	
  	
  Collec5ve	
  intelligence	
  
 “A	
  revolu*on	
  in	
  ar*ficial	
  intelligence	
  is	
  currently	
  
sweeping	
  through	
  computer	
  science.	
  The	
  technique	
  is	
  
called	
  deep	
  learning	
  and	
  it’s	
  affec*ng	
  everything	
  from	
  
facial	
  and	
  voice	
  to	
  fashion	
  and	
  economics.”	
  
“In	
  some	
  sense	
  deep	
  learning	
  is	
  what	
  happened	
  when	
  machine	
  learning	
  hit	
  big	
  data”	
  
“Two	
  kinds	
  of	
  data:	
  raw	
  data	
  (pictures,	
  music,	
  …)	
  and	
  symbolic	
  data	
  (text)”	
  
“With	
  deep	
  learning,	
  we	
  can	
  bridge	
  the	
  gap	
  between	
  the	
  physical	
  world	
  and	
  the	
  
	
   	
   	
   	
   	
   	
   	
   	
   	
   	
   	
   	
   	
   	
   	
  world	
  of	
  compu5ng”	
  	
  	
  
	
   	
   	
   	
   	
   	
  	
  
	
   	
   	
   	
   	
   	
   	
   	
  -­‐-­‐	
  Adam	
  Berenzweig,	
  founding	
  CTO	
  of	
  Clarifai	
  
Ref:	
  Deep	
  Learning:	
  Intelligence	
  from	
  Big	
  Data,	
  Tue	
  Sep	
  16,	
  2014,	
  Stanford	
  Graduate	
  School	
  of	
  Business	
  
	
  
Neural	
  Networks	
  of	
  the	
  80’s	
  
	
  
 
What’s	
  New?	
  
	
  
ü  Big	
  Data	
  
	
  
•  The	
  internet	
  &	
  Social	
  Media	
  
•  Metadata:	
  tags,	
  transla5ons,	
  …	
  
•  Mechanical	
  Turk	
  
Ref:	
  Deep	
  Learning:	
  Intelligence	
  from	
  Big	
  Data,	
  Tue	
  Sep	
  16,	
  2014,	
  Stanford	
  Graduate	
  School	
  of	
  Business	
  
 
What’s	
  New?	
  
	
  
ü  Big	
  Data	
  
ü  Scale	
  
	
  
•  80’s:	
  1-­‐10M	
  (106)	
  neurons/synap5c	
  connec5ons	
  
•  Google	
  Brain:	
  1B	
  (109)	
  	
  
(10M	
  video’s,	
  16k	
  computers,	
  3	
  days)	
  
•  	
  Adult:	
  100T	
  (1014)	
  	
  
•  	
  Infant:	
  1Q	
  (1015)	
  	
  
	
  
Ref:	
  Deep	
  Learning:	
  Intelligence	
  from	
  Big	
  Data,	
  Tue	
  Sep	
  16,	
  2014,	
  Stanford	
  Graduate	
  School	
  of	
  Business	
  
 
What’s	
  New?	
  
	
  
ü  Big	
  Data	
  
ü  Scale	
  
ü  Algorithmic	
  advances	
  
	
  
•  Successive	
  layers	
  of	
  learning/representa5on	
  	
  	
  
•  Unsupervised	
  pre-­‐training	
  	
  
	
  à	
  Structure	
  NN	
  (feature	
  detectors)	
  
•  Then	
  supervised	
  back-­‐prop	
  
	
  à	
  classify/predict	
  labeled	
  data	
  
Ref:	
  Deep	
  Learning:	
  Intelligence	
  from	
  Big	
  Data,	
  Tue	
  Sep	
  16,	
  2014,	
  Stanford	
  Graduate	
  School	
  of	
  Business	
  
 
What’s	
  New?	
  
	
  
ü  Big	
  Data	
  
ü  Scale	
  
ü  Algorithmic	
  advances	
  
	
  
We	
  have	
  been	
  able	
  to	
  reduce	
  the	
  word	
  error	
  rate	
  for	
  speech	
  by	
  over	
  30%	
  compared	
  to	
  
previous	
  methods.	
  This	
  means	
  that	
  rather	
  than	
  having	
  one	
  word	
  in	
  4	
  or	
  5	
  incorrect,	
  now	
  the	
  
error	
  rate	
  is	
  one	
  word	
  in	
  7	
  or	
  8.	
  While	
  s5ll	
  far	
  from	
  perfect,	
  this	
  is	
  the	
  most	
  drama5c	
  change	
  
in	
  accuracy	
  since	
  the	
  introduc5on	
  of	
  hidden	
  Markov	
  modeling	
  in	
  1979,	
  and	
  as	
  we	
  add	
  more	
  
data	
  to	
  the	
  training	
  we	
  believe	
  that	
  we	
  will	
  get	
  even	
  becer	
  results.	
  
November	
  18,	
  2014	
  
Asked	
   whether	
   two	
   unfamiliar	
  
photos	
   of	
   faces	
   show	
   the	
   same	
  
person,	
  a	
  human	
  being	
  will	
  get	
  it	
  
right	
   97.53	
   percent	
   of	
   the	
   5me.	
  
New	
   sodware	
   developed	
   by	
  
researchers	
  at	
  Facebook	
  can	
  score	
  
97.25	
   percent	
   on	
   the	
   same	
  
challenge,	
  regardless	
  of	
  varia5ons	
  
in	
  ligh5ng	
  or	
  whether	
  the	
  person	
  
in	
  the	
  picture	
  is	
  directly	
  facing	
  the	
  
camera.	
  
Feb	
  26,	
  2015	
  
•  Isotherm	
  is	
  to	
  temperature	
  as	
  isobar	
  is	
  to?	
  (i)	
  atmosphere,	
  (ii)	
  wind,	
  (iii)	
  pressure,	
  (iv)	
  la*tude,	
  (v)	
  
current.	
  
	
  
•  Iden*fy	
  two	
  words	
  (one	
  from	
  each	
  set	
  of	
  brackets)	
  that	
  form	
  a	
  connec*on	
  (analogy)	
  when	
  paired	
  
with	
  the	
  words	
  in	
  capitals:	
  CHAPTER	
  (book,	
  verse,	
  read),	
  ACT	
  (stage,	
  audience,	
  play).	
  
	
  
•  Which	
  is	
  the	
  odd	
  one	
  out?	
  (i)	
  calm,	
  (ii)	
  quiet,	
  (iii)	
  relaxed,	
  (iv)	
  serene,	
  (v)	
  unruffled.	
  
	
  
•  	
  Which	
  word	
  is	
  closest	
  to	
  IRRATIONAL?	
  (i)	
  intransigent,	
  (ii)	
  irredeemable,	
  (iii)	
  unsafe,	
  (iv)	
  lost,	
  (v)	
  
nonsensical.	
  
	
  
•  Which	
  word	
  is	
  most	
  opposite	
  to	
  MUSICAL?	
  (i)	
  discordant,	
  (ii)	
  loud,	
  (iii)	
  lyrical,	
  (iv)	
  verbal,	
  (v)	
  
euphonious.	
  
Ref:	
  arxiv.org/abs/1505.07909	
  :	
  Solving	
  Verbal	
  Comprehension	
  Ques5ons	
  in	
  IQ	
  Test	
  by	
  Knowledge-­‐	
  Powered	
  Word	
  Embedding	
  
The	
  future?	
  
	
  
The	
  future?	
  
	
  
“I	
  am	
  in	
  the	
  camp	
  that	
  is	
  concerned	
  about	
  super	
  intelligence.	
  First	
  the	
  machines	
  will	
  do	
  a	
  lot	
  of	
  jobs	
  for	
  
us	
  and	
  not	
  be	
  super	
  intelligent.	
  That	
  should	
  be	
  posi*ve	
  if	
  we	
  manage	
  it	
  well.	
  A	
  few	
  decades	
  a[er	
  that,	
  
though,	
  the	
  intelligence	
  is	
  strong	
  enough	
  to	
  be	
  a	
  concern.	
  I	
  agree	
  with	
  Elon	
  Musk	
  and	
  some	
  others	
  on	
  
this	
  and	
  don't	
  understand	
  why	
  some	
  people	
  are	
  not	
  concerned.”	
  
Stephen	
  Hawking	
  (hcp://www.bbc.com/news/technology-­‐30290540)	
  
	
  	
  
	
  "The	
  development	
  of	
  full	
  ar*ficial	
  intelligence	
  could	
  spell	
  the	
  end	
  of	
  the	
  human	
  race	
  […]	
  
	
  	
  	
  	
  	
  	
  It	
  would	
  take	
  off	
  on	
  its	
  own,	
  and	
  re-­‐design	
  itself	
  at	
  an	
  ever	
  increasing	
  rate	
  […]	
  
	
  	
  	
  	
  	
  	
  Humans,	
  who	
  are	
  limited	
  by	
  slow	
  biological	
  evolu*on,	
  couldn't	
  compete,	
  and	
  would	
  be	
  	
  	
  	
  	
  	
  
	
  	
  	
  	
  	
  	
  superseded.”	
  
•  Oren	
  Etzioni	
  (Computer	
  science,	
  Univ.	
  Washington,	
  CEO	
  of	
  the	
  Allen	
  Ins5t.	
  for	
  Ar5ficial	
  Intelligence):	
  
	
  
	
  “The	
  popular	
  dystopian	
  vision	
  of	
  AI	
  is	
  wrong	
  for	
  one	
  simple	
  reason:	
  it	
  equates	
  intelligence	
  with	
  
autonomy.	
  That	
  is,	
  it	
  assumes	
  a	
  smart	
  computer	
  will	
  create	
  its	
  own	
  goals,	
  and	
  have	
  its	
  own	
  will,	
  and	
  will	
  
use	
  its	
  faster	
  processing	
  abili*es	
  and	
  deep	
  databases	
  to	
  beat	
  humans	
  at	
  their	
  own	
  game.	
  It	
  assumes	
  
that	
  with	
  intelligence	
  comes	
  free	
  will,	
  but	
  I	
  believe	
  those	
  two	
  things	
  are	
  en*rely	
  different”	
  
	
  
•  Michael	
  Licman	
  (AI,	
  Brown	
  Univ.,	
  former	
  program	
  chair	
  for	
  the	
  Ass.	
  of	
  the	
  Advancmnt	
  of	
  AI):	
  
	
  “There	
  are	
  indeed	
  concerns	
  about	
  the	
  near-­‐term	
  future	
  of	
  AI	
  —	
  algorithmic	
  traders	
  crashing	
  the	
  
economy,	
  or	
  sensi*ve	
  power	
  grids	
  overreac*ng	
  to	
  fluctua*ons	
  and	
  shucng	
  down	
  electricity	
  for	
  large	
  
swaths	
  of	
  the	
  popula*on.	
  [...]	
  These	
  worries	
  should	
  play	
  a	
  central	
  role	
  in	
  the	
  development	
  and	
  
deployment	
  of	
  new	
  ideas.	
  But	
  dread	
  predic*ons	
  of	
  computers	
  suddenly	
  waking	
  up	
  and	
  turning	
  on	
  us	
  are	
  
simply	
  not	
  realis*c.”	
  
	
  
•  Yann	
  LeCun	
  (Facebook’s	
  director	
  of	
  research,	
  one	
  of	
  the	
  world’s	
  top	
  experts	
  in	
  deep	
  learning):	
  
	
  “Some	
  people	
  have	
  asked	
  what	
  would	
  prevent	
  a	
  hypothe*cal	
  super-­‐intelligent	
  autonomous	
  
benevolent	
  A.I.	
  to	
  “reprogram”	
  itself	
  and	
  remove	
  its	
  built-­‐in	
  safeguards	
  against	
  gecng	
  rid	
  of	
  humans.	
  
Most	
  of	
  these	
  people	
  are	
  not	
  themselves	
  A.I.	
  researchers,	
  or	
  even	
  computer	
  scien*sts.”	
  
	
  
•  Andrew	
  Ng	
  (founded	
  Google’s	
  Google	
  Brain	
  project,	
  now	
  Chief	
  Scien5st	
  at	
  Baidu):	
  
	
  “Computers	
  are	
  becoming	
  more	
  intelligent	
  and	
  that’s	
  useful	
  as	
  in	
  self-­‐driving	
  cars	
  or	
  speech	
  
recogni*on	
  systems	
  or	
  search	
  engines.	
  That’s	
  intelligence,”	
  he	
  said.	
  “But	
  sen*ence	
  and	
  consciousness	
  is	
  
not	
  something	
  that	
  most	
  of	
  the	
  people	
  I	
  talk	
  to	
  think	
  we’re	
  on	
  the	
  path	
  to.”	
  
Assump5on:	
  Deeper	
  level	
  neurons	
  are	
  more	
  “abstract”	
  
However,	
  what	
  was	
  discovered:	
  
-­‐  A	
  single	
  neuron's	
  feature	
  is	
  no	
  more	
  interpretable	
  as	
  a	
  
meaningful	
  feature	
  than	
  a	
  random	
  set	
  of	
  neurons.	
  	
  
-­‐  NN’s	
  do	
  not	
  "unscramble"	
  the	
  data	
  by	
  mapping	
  features	
  
to	
  individual	
  neurons	
  in	
  say	
  the	
  final	
  layer.	
  The	
  
informa5on	
  that	
  the	
  network	
  extracts	
  is	
  just	
  as	
  much	
  
distributed	
  across	
  all	
  of	
  the	
  neurons	
  as	
  it	
  is	
  localized	
  in	
  a	
  
single	
  neuron.	
  	
  
-­‐  Furthermore,	
  Every	
  deep	
  neural	
  network	
  has	
  "blind	
  
spots"	
  in	
  the	
  sense	
  that	
  there	
  are	
  inputs	
  that	
  are	
  very	
  
close	
  to	
  correctly	
  classified	
  examples	
  that	
  are	
  
misclassified.	
  
The	
  Symbol	
  Grounding	
  Problem	
  
010000110101010
011110101010100
110100101010100
1011010101111…
Jpeg	
  
coding	
  
01000001
01000011
01010100
ASCII	
  
coding	
  
CAT	
  
Deep	
  NN	
  
Harnad,	
  S.	
  (1990)	
  
The	
  Symbol	
  Grounding	
  Problem	
  
010000110101010
011110101010100
110100101010100
1011010101111…
Jpeg	
  
coding	
  
01000001
01000011
01010100
ASCII	
  
coding	
  
CAT	
  
Human	
  
coding	
  
Human	
  
coding	
  
Deep	
  NN	
  
Human	
  
Qualifica5on	
  or	
  Semiosis	
  
Harnad,	
  S.	
  (1990)	
  
The	
  Symbol	
  Grounding	
  Problem	
  
•  Categories	
  (signs	
  and	
  meanings)	
  are	
  ar5facts	
  
•  The	
  rela5on	
  between	
  them	
  is	
  arbitrary	
  
•  They	
  are	
  realized	
  by	
  agents	
  performing	
  semiosis	
  	
  
Diagram	
  of	
  Self-­‐regula5on	
  
The	
  future?	
  
	
  
The	
  future?	
  
	
  
“Collec*ve	
  intelligence	
  is	
  the	
  opposite	
  of	
  ar*ficial	
  intelligence”	
  
Ø  Outer	
  world	
  onto	
  inner	
  world	
  	
  
	
  (human	
  neuronal	
  coding)	
  
Ø 	
  	
  Inner	
  worlds	
  onto	
  each	
  other	
  	
  
	
   	
  (collec5ve	
  intelligence)	
  	
  
Ø 	
  	
  Collec5ve	
  intelligence	
  onto	
  inner	
  	
  	
  
	
  world	
  	
  
•  Semiosis	
  (life)	
  =	
  self-­‐regula5on	
  (produc5on	
  and	
  consump5on	
  of	
  variety,	
  closure)	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
   Self-­‐regulatory	
  system	
  (Agent)	
  
•  Semiosis	
  (life)	
  =	
  self-­‐regula5on	
  (produc5on	
  and	
  consump5on	
  of	
  variety,	
  closure)	
  
•  Tool	
  usage	
  (supplementa5on	
  of	
  variety)	
  
•  Semiosis	
  (life)	
  =	
  self-­‐regula5on	
  (produc5on	
  and	
  consump5on	
  of	
  variety,	
  closure)	
  
•  Tool	
  usage	
  (supplementa5on	
  of	
  variety)	
  
•  Extension	
  and	
  specializa5on	
  (constraints)	
  
	
  “Now,	
  as	
  the	
  Internet	
  revolu*on	
  unfolds,	
  we	
  are	
  seeing	
  not	
  merely	
  an	
  extension	
  of	
  
mind	
  but	
  a	
  unity	
  of	
  mind	
  and	
  machine,	
  two	
  networks	
  coming	
  together	
  as	
  one.”	
  
	
  
[Deepstuff,	
  May	
  25,	
  2015]	
  
•  Semiosis	
  (life)	
  =	
  self-­‐regula5on	
  (produc5on	
  and	
  consump5on	
  of	
  variety,	
  closure)	
  
•  Tool	
  usage	
  (supplementa5on	
  of	
  variety)	
  
•  Extension	
  and	
  specializa5on	
  (constraints)	
  
•  Coordina5on	
  (conven5onal	
  codes)	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
Agent	
  1	
  
Agent	
  2	
  
•  Semiosis	
  (life)	
  =	
  self-­‐regula5on	
  (produc5on	
  and	
  consump5on	
  of	
  variety,	
  closure)	
  
•  Tool	
  usage	
  (supplementa5on	
  of	
  variety)	
  
•  Extension	
  and	
  specializa5on	
  (constraints)	
  
•  Coordina5on	
  (conven5onal	
  codes)	
  
	
  	
  
à	
  Metasystem	
  or	
  “Major	
  Transi5on”	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
Agent	
  1	
  
Agent	
  2	
  
Meta	
  
agent	
  
“In	
  a	
  sense,	
  deep	
  learning	
  is	
  what	
  happened	
  when	
  machine	
  learning	
  hit	
  big	
  data”	
  
“Two	
  kinds	
  of	
  data:	
  raw	
  data	
  (pictures,	
  music,	
  …)	
  and	
  symbolic	
  data	
  (text)”	
  
“With	
  deep	
  learning,	
  we	
  can	
  bridge	
  the	
  gap	
  between	
  the	
  physical	
  world	
  and	
  the	
  
	
  world	
  of	
  compu5ng”	
  
	
  	
  	
  
	
   	
   	
   	
   	
   	
  	
  
	
   	
   	
   	
   	
   	
   	
   	
  -­‐-­‐	
  Adam	
  Berenzweig,	
  founding	
  CTO	
  of	
  Clarifai	
  
The	
  Next	
  Major	
  Transi5on?	
  
	
  
	
  
	
  
	
  
Symbolic	
  
	
  
Collec5ve	
  intelligence	
  
(deep	
  learning)	
  
Physical	
  
	
  
Collec5ve	
  ac5ng	
  
(da5ng,	
  vo5ng,	
  …)	
  
Informa5on	
  seeking	
  
ac5ng	
  
Tagging	
  and	
  training	
  
Replica*on	
  always	
  involves	
  coding!	
  
3,	
  15	
  or	
  33	
  numbers?	
  
The	
  Symbol	
  Grounding	
  Problem	
  Harnad,	
  S.	
  (1990)	
  
ü  	
  A	
  robot	
  may	
  not	
  injure	
  a	
  human	
  being	
  or,	
  through	
  
inac5on,	
  allow	
  a	
  human	
  being	
  to	
  come	
  to	
  harm.	
  
ü  A	
  robot	
  must	
  obey	
  the	
  orders	
  given	
  to	
  it	
  by	
  human	
  
beings,	
  except	
  where	
  such	
  orders	
  would	
  conflict	
  with	
  the	
  
First	
  Law.	
  
ü  A	
  robot	
  must	
  protect	
  its	
  own	
  existence	
  as	
  long	
  as	
  such	
  
protec5on	
  does	
  not	
  conflict	
  with	
  the	
  First	
  or	
  Second	
  
Laws.	
  
MIT	
  Technology	
  review,	
  Robert	
  D.	
  Hof,	
  April	
  23,	
  2014	
  
Professor	
  Geoff	
  Hinton,	
  who	
  was	
  hired	
  by	
  Google	
  two	
  years	
  ago	
  to	
  help	
  develop	
  intelligent	
  opera5ng	
  
systems,	
  said	
  that	
  the	
  company	
  is	
  on	
  the	
  brink	
  of	
  developing	
  algorithms	
  with	
  the	
  capacity	
  for	
  logic,	
  
natural	
  conversa5on	
  and	
  even	
  flirta5on.	
  
	
  
“Basically,	
  they’ll	
  have	
  common	
  sense”	
  
	
  
“Thought	
  vectors,	
  Hinton	
  explained,	
  work	
  at	
  a	
  higher	
  level	
  by	
  extrac5ng	
  something	
  closer	
  to	
  actual	
  
meaning”	
  
Code Biology and the Rise of Deep Learning AI
Code Biology and the Rise of Deep Learning AI

More Related Content

What's hot

Case study on deep learning
Case study on deep learningCase study on deep learning
Case study on deep learningHarshitBarde
 
Superintelligence: how afraid should we be?
Superintelligence: how afraid should we be?Superintelligence: how afraid should we be?
Superintelligence: how afraid should we be?David Wood
 
Inauguration Function - Ohio Center of Excellence in Knowledge-Enabled Comput...
Inauguration Function - Ohio Center of Excellence in Knowledge-Enabled Comput...Inauguration Function - Ohio Center of Excellence in Knowledge-Enabled Comput...
Inauguration Function - Ohio Center of Excellence in Knowledge-Enabled Comput...Artificial Intelligence Institute at UofSC
 
What will the world be like 50 years 20 (1)
What will the world be like 50 years 20 (1)What will the world be like 50 years 20 (1)
What will the world be like 50 years 20 (1)nitut1
 
Waiting for Exascale
Waiting for ExascaleWaiting for Exascale
Waiting for ExascaleGary Johnson
 
What’s Next in computing & the role of cloud FPGAs
What’s Next in computing & the role of cloud FPGAsWhat’s Next in computing & the role of cloud FPGAs
What’s Next in computing & the role of cloud FPGAsDionysios Diamantopoulos
 
Kim Solez Technology, the Future of Medicine, and the Bridge between Transpla...
Kim Solez Technology, the Future of Medicine, and the Bridge between Transpla...Kim Solez Technology, the Future of Medicine, and the Bridge between Transpla...
Kim Solez Technology, the Future of Medicine, and the Bridge between Transpla...Kim Solez ,
 
[TRANSCRIPT] Do we have a right to freedom of thought?
 [TRANSCRIPT] Do we have a right to freedom of thought?  [TRANSCRIPT] Do we have a right to freedom of thought?
[TRANSCRIPT] Do we have a right to freedom of thought? Jim Stroud
 
Artificial Intelligence is back, Deep Learning Networks and Quantum possibili...
Artificial Intelligence is back, Deep Learning Networks and Quantum possibili...Artificial Intelligence is back, Deep Learning Networks and Quantum possibili...
Artificial Intelligence is back, Deep Learning Networks and Quantum possibili...John Mathon
 
WHERE IS THE BIG DATA INDUSTRY GOING? from Structure:Data 2013
WHERE IS THE BIG DATA INDUSTRY GOING? from Structure:Data 2013WHERE IS THE BIG DATA INDUSTRY GOING? from Structure:Data 2013
WHERE IS THE BIG DATA INDUSTRY GOING? from Structure:Data 2013Gigaom
 
Machine creativity TED Talk 2.0
Machine creativity TED Talk 2.0Machine creativity TED Talk 2.0
Machine creativity TED Talk 2.0Cameron Aaron
 
Basic questions about artificial intelligence
Basic questions about artificial intelligenceBasic questions about artificial intelligence
Basic questions about artificial intelligenceAqib Memon
 

What's hot (20)

Case study on deep learning
Case study on deep learningCase study on deep learning
Case study on deep learning
 
superintelligence
superintelligencesuperintelligence
superintelligence
 
Superintelligence: how afraid should we be?
Superintelligence: how afraid should we be?Superintelligence: how afraid should we be?
Superintelligence: how afraid should we be?
 
Is digital technology re-wiring your brain?
Is digital technology re-wiring your brain?Is digital technology re-wiring your brain?
Is digital technology re-wiring your brain?
 
Inauguration Function - Ohio Center of Excellence in Knowledge-Enabled Comput...
Inauguration Function - Ohio Center of Excellence in Knowledge-Enabled Comput...Inauguration Function - Ohio Center of Excellence in Knowledge-Enabled Comput...
Inauguration Function - Ohio Center of Excellence in Knowledge-Enabled Comput...
 
What will the world be like 50 years 20 (1)
What will the world be like 50 years 20 (1)What will the world be like 50 years 20 (1)
What will the world be like 50 years 20 (1)
 
Cognitive computing
Cognitive computingCognitive computing
Cognitive computing
 
Waiting for Exascale
Waiting for ExascaleWaiting for Exascale
Waiting for Exascale
 
What’s Next in computing & the role of cloud FPGAs
What’s Next in computing & the role of cloud FPGAsWhat’s Next in computing & the role of cloud FPGAs
What’s Next in computing & the role of cloud FPGAs
 
Kim Solez Technology, the Future of Medicine, and the Bridge between Transpla...
Kim Solez Technology, the Future of Medicine, and the Bridge between Transpla...Kim Solez Technology, the Future of Medicine, and the Bridge between Transpla...
Kim Solez Technology, the Future of Medicine, and the Bridge between Transpla...
 
Computer Vision++: Where Do We Go from Here?
Computer Vision++: Where Do We Go from Here?Computer Vision++: Where Do We Go from Here?
Computer Vision++: Where Do We Go from Here?
 
Big Data Better Life
Big Data Better LifeBig Data Better Life
Big Data Better Life
 
[TRANSCRIPT] Do we have a right to freedom of thought?
 [TRANSCRIPT] Do we have a right to freedom of thought?  [TRANSCRIPT] Do we have a right to freedom of thought?
[TRANSCRIPT] Do we have a right to freedom of thought?
 
Artificial Intelligence is back, Deep Learning Networks and Quantum possibili...
Artificial Intelligence is back, Deep Learning Networks and Quantum possibili...Artificial Intelligence is back, Deep Learning Networks and Quantum possibili...
Artificial Intelligence is back, Deep Learning Networks and Quantum possibili...
 
Cloud Superintelligence
Cloud SuperintelligenceCloud Superintelligence
Cloud Superintelligence
 
Seminar
SeminarSeminar
Seminar
 
Artificial intel
Artificial intelArtificial intel
Artificial intel
 
WHERE IS THE BIG DATA INDUSTRY GOING? from Structure:Data 2013
WHERE IS THE BIG DATA INDUSTRY GOING? from Structure:Data 2013WHERE IS THE BIG DATA INDUSTRY GOING? from Structure:Data 2013
WHERE IS THE BIG DATA INDUSTRY GOING? from Structure:Data 2013
 
Machine creativity TED Talk 2.0
Machine creativity TED Talk 2.0Machine creativity TED Talk 2.0
Machine creativity TED Talk 2.0
 
Basic questions about artificial intelligence
Basic questions about artificial intelligenceBasic questions about artificial intelligence
Basic questions about artificial intelligence
 

Similar to Code Biology and the Rise of Deep Learning AI

Machine Learning, AI and the Brain
Machine Learning, AI and the Brain Machine Learning, AI and the Brain
Machine Learning, AI and the Brain TechExeter
 
AI and the Researcher: ChatGPT and DALL-E in Scholarly Writing and Publishing
AI and the Researcher: ChatGPT and DALL-E in Scholarly Writing and PublishingAI and the Researcher: ChatGPT and DALL-E in Scholarly Writing and Publishing
AI and the Researcher: ChatGPT and DALL-E in Scholarly Writing and PublishingErin Owens
 
Introduction to Artificial Intelligences
Introduction to Artificial IntelligencesIntroduction to Artificial Intelligences
Introduction to Artificial IntelligencesMeenakshi Paul
 
computer science engineering spe ialized in artificial Intelligence
computer science engineering spe ialized in artificial Intelligencecomputer science engineering spe ialized in artificial Intelligence
computer science engineering spe ialized in artificial IntelligenceKhanKhaja1
 
AI – Risks, Opportunities and Ethical Issues.pdf
AI – Risks, Opportunities and Ethical Issues.pdfAI – Risks, Opportunities and Ethical Issues.pdf
AI – Risks, Opportunities and Ethical Issues.pdfAdam Ford
 
What really is Artificial Intelligence about?
What really is Artificial Intelligence about? What really is Artificial Intelligence about?
What really is Artificial Intelligence about? Harmony Kwawu
 
Artificial Intelligence: Existential Threat or Our Best Hope for the Future?
Artificial Intelligence: Existential Threat or Our Best Hope for the Future?Artificial Intelligence: Existential Threat or Our Best Hope for the Future?
Artificial Intelligence: Existential Threat or Our Best Hope for the Future?James Hendler
 
Artificial Intelligence and Intuition
Artificial  Intelligence  and  IntuitionArtificial  Intelligence  and  Intuition
Artificial Intelligence and IntuitionViktor Dörfler
 
Technologies Demystified: Artificial Intelligence
Technologies Demystified: Artificial IntelligenceTechnologies Demystified: Artificial Intelligence
Technologies Demystified: Artificial IntelligencePioneers.io
 
AI – Risks, Opportunities and Ethical Issues April 2023.pdf
AI – Risks, Opportunities and Ethical Issues April 2023.pdfAI – Risks, Opportunities and Ethical Issues April 2023.pdf
AI – Risks, Opportunities and Ethical Issues April 2023.pdfAdam Ford
 
Sp14 cs188 lecture 1 - introduction
Sp14 cs188 lecture 1  - introductionSp14 cs188 lecture 1  - introduction
Sp14 cs188 lecture 1 - introductionAmer Noureddin
 
Introduction to Artificial intelligence and ML
Introduction to Artificial intelligence and MLIntroduction to Artificial intelligence and ML
Introduction to Artificial intelligence and MLbansalpra7
 
SP14 CS188 Lecture 1 -- Introduction.pptx
SP14 CS188 Lecture 1 -- Introduction.pptxSP14 CS188 Lecture 1 -- Introduction.pptx
SP14 CS188 Lecture 1 -- Introduction.pptxssuser851498
 
How artificial intelligence can help you today
How artificial intelligence can help you todayHow artificial intelligence can help you today
How artificial intelligence can help you todayHenrik de Gyor
 

Similar to Code Biology and the Rise of Deep Learning AI (20)

Machine Learning, AI and the Brain
Machine Learning, AI and the Brain Machine Learning, AI and the Brain
Machine Learning, AI and the Brain
 
Ai titech-virach-20191026
Ai titech-virach-20191026Ai titech-virach-20191026
Ai titech-virach-20191026
 
AI and the Researcher: ChatGPT and DALL-E in Scholarly Writing and Publishing
AI and the Researcher: ChatGPT and DALL-E in Scholarly Writing and PublishingAI and the Researcher: ChatGPT and DALL-E in Scholarly Writing and Publishing
AI and the Researcher: ChatGPT and DALL-E in Scholarly Writing and Publishing
 
When AI becomes a data-driven machine, and digital is everywhere!
When AI becomes a data-driven machine, and digital is everywhere!When AI becomes a data-driven machine, and digital is everywhere!
When AI becomes a data-driven machine, and digital is everywhere!
 
ai.ppt
ai.pptai.ppt
ai.ppt
 
ai.ppt
ai.pptai.ppt
ai.ppt
 
Introduction to Artificial Intelligences
Introduction to Artificial IntelligencesIntroduction to Artificial Intelligences
Introduction to Artificial Intelligences
 
ai.ppt
ai.pptai.ppt
ai.ppt
 
computer science engineering spe ialized in artificial Intelligence
computer science engineering spe ialized in artificial Intelligencecomputer science engineering spe ialized in artificial Intelligence
computer science engineering spe ialized in artificial Intelligence
 
ai.ppt
ai.pptai.ppt
ai.ppt
 
AI – Risks, Opportunities and Ethical Issues.pdf
AI – Risks, Opportunities and Ethical Issues.pdfAI – Risks, Opportunities and Ethical Issues.pdf
AI – Risks, Opportunities and Ethical Issues.pdf
 
What really is Artificial Intelligence about?
What really is Artificial Intelligence about? What really is Artificial Intelligence about?
What really is Artificial Intelligence about?
 
Artificial Intelligence: Existential Threat or Our Best Hope for the Future?
Artificial Intelligence: Existential Threat or Our Best Hope for the Future?Artificial Intelligence: Existential Threat or Our Best Hope for the Future?
Artificial Intelligence: Existential Threat or Our Best Hope for the Future?
 
Artificial Intelligence and Intuition
Artificial  Intelligence  and  IntuitionArtificial  Intelligence  and  Intuition
Artificial Intelligence and Intuition
 
Technologies Demystified: Artificial Intelligence
Technologies Demystified: Artificial IntelligenceTechnologies Demystified: Artificial Intelligence
Technologies Demystified: Artificial Intelligence
 
AI – Risks, Opportunities and Ethical Issues April 2023.pdf
AI – Risks, Opportunities and Ethical Issues April 2023.pdfAI – Risks, Opportunities and Ethical Issues April 2023.pdf
AI – Risks, Opportunities and Ethical Issues April 2023.pdf
 
Sp14 cs188 lecture 1 - introduction
Sp14 cs188 lecture 1  - introductionSp14 cs188 lecture 1  - introduction
Sp14 cs188 lecture 1 - introduction
 
Introduction to Artificial intelligence and ML
Introduction to Artificial intelligence and MLIntroduction to Artificial intelligence and ML
Introduction to Artificial intelligence and ML
 
SP14 CS188 Lecture 1 -- Introduction.pptx
SP14 CS188 Lecture 1 -- Introduction.pptxSP14 CS188 Lecture 1 -- Introduction.pptx
SP14 CS188 Lecture 1 -- Introduction.pptx
 
How artificial intelligence can help you today
How artificial intelligence can help you todayHow artificial intelligence can help you today
How artificial intelligence can help you today
 

Recently uploaded

Training state-of-the-art general text embedding
Training state-of-the-art general text embeddingTraining state-of-the-art general text embedding
Training state-of-the-art general text embeddingZilliz
 
DevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platformsDevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platformsSergiu Bodiu
 
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Mark Simos
 
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024BookNet Canada
 
Take control of your SAP testing with UiPath Test Suite
Take control of your SAP testing with UiPath Test SuiteTake control of your SAP testing with UiPath Test Suite
Take control of your SAP testing with UiPath Test SuiteDianaGray10
 
Dev Dives: Streamline document processing with UiPath Studio Web
Dev Dives: Streamline document processing with UiPath Studio WebDev Dives: Streamline document processing with UiPath Studio Web
Dev Dives: Streamline document processing with UiPath Studio WebUiPathCommunity
 
SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024Lorenzo Miniero
 
Time Series Foundation Models - current state and future directions
Time Series Foundation Models - current state and future directionsTime Series Foundation Models - current state and future directions
Time Series Foundation Models - current state and future directionsNathaniel Shimoni
 
The Fit for Passkeys for Employee and Consumer Sign-ins: FIDO Paris Seminar.pptx
The Fit for Passkeys for Employee and Consumer Sign-ins: FIDO Paris Seminar.pptxThe Fit for Passkeys for Employee and Consumer Sign-ins: FIDO Paris Seminar.pptx
The Fit for Passkeys for Employee and Consumer Sign-ins: FIDO Paris Seminar.pptxLoriGlavin3
 
Rise of the Machines: Known As Drones...
Rise of the Machines: Known As Drones...Rise of the Machines: Known As Drones...
Rise of the Machines: Known As Drones...Rick Flair
 
Moving Beyond Passwords: FIDO Paris Seminar.pdf
Moving Beyond Passwords: FIDO Paris Seminar.pdfMoving Beyond Passwords: FIDO Paris Seminar.pdf
Moving Beyond Passwords: FIDO Paris Seminar.pdfLoriGlavin3
 
DevoxxFR 2024 Reproducible Builds with Apache Maven
DevoxxFR 2024 Reproducible Builds with Apache MavenDevoxxFR 2024 Reproducible Builds with Apache Maven
DevoxxFR 2024 Reproducible Builds with Apache MavenHervé Boutemy
 
Unraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdfUnraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdfAlex Barbosa Coqueiro
 
TeamStation AI System Report LATAM IT Salaries 2024
TeamStation AI System Report LATAM IT Salaries 2024TeamStation AI System Report LATAM IT Salaries 2024
TeamStation AI System Report LATAM IT Salaries 2024Lonnie McRorey
 
Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 3652toLead Limited
 
unit 4 immunoblotting technique complete.pptx
unit 4 immunoblotting technique complete.pptxunit 4 immunoblotting technique complete.pptx
unit 4 immunoblotting technique complete.pptxBkGupta21
 
Transcript: New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024Transcript: New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024BookNet Canada
 
What is Artificial Intelligence?????????
What is Artificial Intelligence?????????What is Artificial Intelligence?????????
What is Artificial Intelligence?????????blackmambaettijean
 
Advanced Computer Architecture – An Introduction
Advanced Computer Architecture – An IntroductionAdvanced Computer Architecture – An Introduction
Advanced Computer Architecture – An IntroductionDilum Bandara
 
What is DBT - The Ultimate Data Build Tool.pdf
What is DBT - The Ultimate Data Build Tool.pdfWhat is DBT - The Ultimate Data Build Tool.pdf
What is DBT - The Ultimate Data Build Tool.pdfMounikaPolabathina
 

Recently uploaded (20)

Training state-of-the-art general text embedding
Training state-of-the-art general text embeddingTraining state-of-the-art general text embedding
Training state-of-the-art general text embedding
 
DevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platformsDevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platforms
 
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
 
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
 
Take control of your SAP testing with UiPath Test Suite
Take control of your SAP testing with UiPath Test SuiteTake control of your SAP testing with UiPath Test Suite
Take control of your SAP testing with UiPath Test Suite
 
Dev Dives: Streamline document processing with UiPath Studio Web
Dev Dives: Streamline document processing with UiPath Studio WebDev Dives: Streamline document processing with UiPath Studio Web
Dev Dives: Streamline document processing with UiPath Studio Web
 
SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024
 
Time Series Foundation Models - current state and future directions
Time Series Foundation Models - current state and future directionsTime Series Foundation Models - current state and future directions
Time Series Foundation Models - current state and future directions
 
The Fit for Passkeys for Employee and Consumer Sign-ins: FIDO Paris Seminar.pptx
The Fit for Passkeys for Employee and Consumer Sign-ins: FIDO Paris Seminar.pptxThe Fit for Passkeys for Employee and Consumer Sign-ins: FIDO Paris Seminar.pptx
The Fit for Passkeys for Employee and Consumer Sign-ins: FIDO Paris Seminar.pptx
 
Rise of the Machines: Known As Drones...
Rise of the Machines: Known As Drones...Rise of the Machines: Known As Drones...
Rise of the Machines: Known As Drones...
 
Moving Beyond Passwords: FIDO Paris Seminar.pdf
Moving Beyond Passwords: FIDO Paris Seminar.pdfMoving Beyond Passwords: FIDO Paris Seminar.pdf
Moving Beyond Passwords: FIDO Paris Seminar.pdf
 
DevoxxFR 2024 Reproducible Builds with Apache Maven
DevoxxFR 2024 Reproducible Builds with Apache MavenDevoxxFR 2024 Reproducible Builds with Apache Maven
DevoxxFR 2024 Reproducible Builds with Apache Maven
 
Unraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdfUnraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdf
 
TeamStation AI System Report LATAM IT Salaries 2024
TeamStation AI System Report LATAM IT Salaries 2024TeamStation AI System Report LATAM IT Salaries 2024
TeamStation AI System Report LATAM IT Salaries 2024
 
Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365
 
unit 4 immunoblotting technique complete.pptx
unit 4 immunoblotting technique complete.pptxunit 4 immunoblotting technique complete.pptx
unit 4 immunoblotting technique complete.pptx
 
Transcript: New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024Transcript: New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
 
What is Artificial Intelligence?????????
What is Artificial Intelligence?????????What is Artificial Intelligence?????????
What is Artificial Intelligence?????????
 
Advanced Computer Architecture – An Introduction
Advanced Computer Architecture – An IntroductionAdvanced Computer Architecture – An Introduction
Advanced Computer Architecture – An Introduction
 
What is DBT - The Ultimate Data Build Tool.pdf
What is DBT - The Ultimate Data Build Tool.pdfWhat is DBT - The Ultimate Data Build Tool.pdf
What is DBT - The Ultimate Data Build Tool.pdf
 

Code Biology and the Rise of Deep Learning AI

  • 1. Code  Biology  and  (the  future  of)     Ar5ficial  Intelligence     Joachim  De  Beule  
  • 2. Recent  advances  in  AI              Deep  learning   A  dark  future                Superintelligences  more  dangerous                        than  nukes   A  brighter  future                Collec5ve  intelligence  
  • 3.  “A  revolu*on  in  ar*ficial  intelligence  is  currently   sweeping  through  computer  science.  The  technique  is   called  deep  learning  and  it’s  affec*ng  everything  from   facial  and  voice  to  fashion  and  economics.”  
  • 4. “In  some  sense  deep  learning  is  what  happened  when  machine  learning  hit  big  data”   “Two  kinds  of  data:  raw  data  (pictures,  music,  …)  and  symbolic  data  (text)”   “With  deep  learning,  we  can  bridge  the  gap  between  the  physical  world  and  the                                world  of  compu5ng”                                    -­‐-­‐  Adam  Berenzweig,  founding  CTO  of  Clarifai  
  • 5. Ref:  Deep  Learning:  Intelligence  from  Big  Data,  Tue  Sep  16,  2014,  Stanford  Graduate  School  of  Business     Neural  Networks  of  the  80’s    
  • 6.   What’s  New?     ü  Big  Data     •  The  internet  &  Social  Media   •  Metadata:  tags,  transla5ons,  …   •  Mechanical  Turk   Ref:  Deep  Learning:  Intelligence  from  Big  Data,  Tue  Sep  16,  2014,  Stanford  Graduate  School  of  Business  
  • 7.   What’s  New?     ü  Big  Data   ü  Scale     •  80’s:  1-­‐10M  (106)  neurons/synap5c  connec5ons   •  Google  Brain:  1B  (109)     (10M  video’s,  16k  computers,  3  days)   •   Adult:  100T  (1014)     •   Infant:  1Q  (1015)       Ref:  Deep  Learning:  Intelligence  from  Big  Data,  Tue  Sep  16,  2014,  Stanford  Graduate  School  of  Business  
  • 8.   What’s  New?     ü  Big  Data   ü  Scale   ü  Algorithmic  advances     •  Successive  layers  of  learning/representa5on       •  Unsupervised  pre-­‐training      à  Structure  NN  (feature  detectors)   •  Then  supervised  back-­‐prop    à  classify/predict  labeled  data   Ref:  Deep  Learning:  Intelligence  from  Big  Data,  Tue  Sep  16,  2014,  Stanford  Graduate  School  of  Business  
  • 9.   What’s  New?     ü  Big  Data   ü  Scale   ü  Algorithmic  advances    
  • 10.
  • 11.
  • 12.
  • 13. We  have  been  able  to  reduce  the  word  error  rate  for  speech  by  over  30%  compared  to   previous  methods.  This  means  that  rather  than  having  one  word  in  4  or  5  incorrect,  now  the   error  rate  is  one  word  in  7  or  8.  While  s5ll  far  from  perfect,  this  is  the  most  drama5c  change   in  accuracy  since  the  introduc5on  of  hidden  Markov  modeling  in  1979,  and  as  we  add  more   data  to  the  training  we  believe  that  we  will  get  even  becer  results.  
  • 14.
  • 16. Asked   whether   two   unfamiliar   photos   of   faces   show   the   same   person,  a  human  being  will  get  it   right   97.53   percent   of   the   5me.   New   sodware   developed   by   researchers  at  Facebook  can  score   97.25   percent   on   the   same   challenge,  regardless  of  varia5ons   in  ligh5ng  or  whether  the  person   in  the  picture  is  directly  facing  the   camera.  
  • 18. •  Isotherm  is  to  temperature  as  isobar  is  to?  (i)  atmosphere,  (ii)  wind,  (iii)  pressure,  (iv)  la*tude,  (v)   current.     •  Iden*fy  two  words  (one  from  each  set  of  brackets)  that  form  a  connec*on  (analogy)  when  paired   with  the  words  in  capitals:  CHAPTER  (book,  verse,  read),  ACT  (stage,  audience,  play).     •  Which  is  the  odd  one  out?  (i)  calm,  (ii)  quiet,  (iii)  relaxed,  (iv)  serene,  (v)  unruffled.     •   Which  word  is  closest  to  IRRATIONAL?  (i)  intransigent,  (ii)  irredeemable,  (iii)  unsafe,  (iv)  lost,  (v)   nonsensical.     •  Which  word  is  most  opposite  to  MUSICAL?  (i)  discordant,  (ii)  loud,  (iii)  lyrical,  (iv)  verbal,  (v)   euphonious.   Ref:  arxiv.org/abs/1505.07909  :  Solving  Verbal  Comprehension  Ques5ons  in  IQ  Test  by  Knowledge-­‐  Powered  Word  Embedding  
  • 19.
  • 22. “I  am  in  the  camp  that  is  concerned  about  super  intelligence.  First  the  machines  will  do  a  lot  of  jobs  for   us  and  not  be  super  intelligent.  That  should  be  posi*ve  if  we  manage  it  well.  A  few  decades  a[er  that,   though,  the  intelligence  is  strong  enough  to  be  a  concern.  I  agree  with  Elon  Musk  and  some  others  on   this  and  don't  understand  why  some  people  are  not  concerned.”   Stephen  Hawking  (hcp://www.bbc.com/news/technology-­‐30290540)        "The  development  of  full  ar*ficial  intelligence  could  spell  the  end  of  the  human  race  […]              It  would  take  off  on  its  own,  and  re-­‐design  itself  at  an  ever  increasing  rate  […]              Humans,  who  are  limited  by  slow  biological  evolu*on,  couldn't  compete,  and  would  be                        superseded.”  
  • 23.
  • 24. •  Oren  Etzioni  (Computer  science,  Univ.  Washington,  CEO  of  the  Allen  Ins5t.  for  Ar5ficial  Intelligence):      “The  popular  dystopian  vision  of  AI  is  wrong  for  one  simple  reason:  it  equates  intelligence  with   autonomy.  That  is,  it  assumes  a  smart  computer  will  create  its  own  goals,  and  have  its  own  will,  and  will   use  its  faster  processing  abili*es  and  deep  databases  to  beat  humans  at  their  own  game.  It  assumes   that  with  intelligence  comes  free  will,  but  I  believe  those  two  things  are  en*rely  different”     •  Michael  Licman  (AI,  Brown  Univ.,  former  program  chair  for  the  Ass.  of  the  Advancmnt  of  AI):    “There  are  indeed  concerns  about  the  near-­‐term  future  of  AI  —  algorithmic  traders  crashing  the   economy,  or  sensi*ve  power  grids  overreac*ng  to  fluctua*ons  and  shucng  down  electricity  for  large   swaths  of  the  popula*on.  [...]  These  worries  should  play  a  central  role  in  the  development  and   deployment  of  new  ideas.  But  dread  predic*ons  of  computers  suddenly  waking  up  and  turning  on  us  are   simply  not  realis*c.”     •  Yann  LeCun  (Facebook’s  director  of  research,  one  of  the  world’s  top  experts  in  deep  learning):    “Some  people  have  asked  what  would  prevent  a  hypothe*cal  super-­‐intelligent  autonomous   benevolent  A.I.  to  “reprogram”  itself  and  remove  its  built-­‐in  safeguards  against  gecng  rid  of  humans.   Most  of  these  people  are  not  themselves  A.I.  researchers,  or  even  computer  scien*sts.”     •  Andrew  Ng  (founded  Google’s  Google  Brain  project,  now  Chief  Scien5st  at  Baidu):    “Computers  are  becoming  more  intelligent  and  that’s  useful  as  in  self-­‐driving  cars  or  speech   recogni*on  systems  or  search  engines.  That’s  intelligence,”  he  said.  “But  sen*ence  and  consciousness  is   not  something  that  most  of  the  people  I  talk  to  think  we’re  on  the  path  to.”  
  • 25. Assump5on:  Deeper  level  neurons  are  more  “abstract”   However,  what  was  discovered:   -­‐  A  single  neuron's  feature  is  no  more  interpretable  as  a   meaningful  feature  than  a  random  set  of  neurons.     -­‐  NN’s  do  not  "unscramble"  the  data  by  mapping  features   to  individual  neurons  in  say  the  final  layer.  The   informa5on  that  the  network  extracts  is  just  as  much   distributed  across  all  of  the  neurons  as  it  is  localized  in  a   single  neuron.     -­‐  Furthermore,  Every  deep  neural  network  has  "blind   spots"  in  the  sense  that  there  are  inputs  that  are  very   close  to  correctly  classified  examples  that  are   misclassified.  
  • 26.
  • 27.
  • 28.
  • 29.
  • 30. The  Symbol  Grounding  Problem   010000110101010 011110101010100 110100101010100 1011010101111… Jpeg   coding   01000001 01000011 01010100 ASCII   coding   CAT   Deep  NN   Harnad,  S.  (1990)  
  • 31. The  Symbol  Grounding  Problem   010000110101010 011110101010100 110100101010100 1011010101111… Jpeg   coding   01000001 01000011 01010100 ASCII   coding   CAT   Human   coding   Human   coding   Deep  NN   Human   Qualifica5on  or  Semiosis   Harnad,  S.  (1990)  
  • 32. The  Symbol  Grounding  Problem   •  Categories  (signs  and  meanings)  are  ar5facts   •  The  rela5on  between  them  is  arbitrary   •  They  are  realized  by  agents  performing  semiosis     Diagram  of  Self-­‐regula5on  
  • 34. The  future?     “Collec*ve  intelligence  is  the  opposite  of  ar*ficial  intelligence”  
  • 35.
  • 36. Ø  Outer  world  onto  inner  world      (human  neuronal  coding)   Ø     Inner  worlds  onto  each  other        (collec5ve  intelligence)     Ø     Collec5ve  intelligence  onto  inner        world    
  • 37. •  Semiosis  (life)  =  self-­‐regula5on  (produc5on  and  consump5on  of  variety,  closure)                               Self-­‐regulatory  system  (Agent)  
  • 38. •  Semiosis  (life)  =  self-­‐regula5on  (produc5on  and  consump5on  of  variety,  closure)   •  Tool  usage  (supplementa5on  of  variety)  
  • 39. •  Semiosis  (life)  =  self-­‐regula5on  (produc5on  and  consump5on  of  variety,  closure)   •  Tool  usage  (supplementa5on  of  variety)   •  Extension  and  specializa5on  (constraints)    “Now,  as  the  Internet  revolu*on  unfolds,  we  are  seeing  not  merely  an  extension  of   mind  but  a  unity  of  mind  and  machine,  two  networks  coming  together  as  one.”     [Deepstuff,  May  25,  2015]  
  • 40. •  Semiosis  (life)  =  self-­‐regula5on  (produc5on  and  consump5on  of  variety,  closure)   •  Tool  usage  (supplementa5on  of  variety)   •  Extension  and  specializa5on  (constraints)   •  Coordina5on  (conven5onal  codes)                         Agent  1   Agent  2  
  • 41. •  Semiosis  (life)  =  self-­‐regula5on  (produc5on  and  consump5on  of  variety,  closure)   •  Tool  usage  (supplementa5on  of  variety)   •  Extension  and  specializa5on  (constraints)   •  Coordina5on  (conven5onal  codes)       à  Metasystem  or  “Major  Transi5on”                       Agent  1   Agent  2   Meta   agent  
  • 42. “In  a  sense,  deep  learning  is  what  happened  when  machine  learning  hit  big  data”   “Two  kinds  of  data:  raw  data  (pictures,  music,  …)  and  symbolic  data  (text)”   “With  deep  learning,  we  can  bridge  the  gap  between  the  physical  world  and  the    world  of  compu5ng”                                      -­‐-­‐  Adam  Berenzweig,  founding  CTO  of  Clarifai  
  • 43. The  Next  Major  Transi5on?           Symbolic     Collec5ve  intelligence   (deep  learning)   Physical     Collec5ve  ac5ng   (da5ng,  vo5ng,  …)   Informa5on  seeking   ac5ng   Tagging  and  training  
  • 44.
  • 45. Replica*on  always  involves  coding!   3,  15  or  33  numbers?   The  Symbol  Grounding  Problem  Harnad,  S.  (1990)  
  • 46.
  • 47.
  • 48. ü   A  robot  may  not  injure  a  human  being  or,  through   inac5on,  allow  a  human  being  to  come  to  harm.   ü  A  robot  must  obey  the  orders  given  to  it  by  human   beings,  except  where  such  orders  would  conflict  with  the   First  Law.   ü  A  robot  must  protect  its  own  existence  as  long  as  such   protec5on  does  not  conflict  with  the  First  or  Second   Laws.  
  • 49.
  • 50.
  • 51.
  • 52. MIT  Technology  review,  Robert  D.  Hof,  April  23,  2014  
  • 53.
  • 54. Professor  Geoff  Hinton,  who  was  hired  by  Google  two  years  ago  to  help  develop  intelligent  opera5ng   systems,  said  that  the  company  is  on  the  brink  of  developing  algorithms  with  the  capacity  for  logic,   natural  conversa5on  and  even  flirta5on.     “Basically,  they’ll  have  common  sense”     “Thought  vectors,  Hinton  explained,  work  at  a  higher  level  by  extrac5ng  something  closer  to  actual   meaning”