Artificial	
  Intelligence,	
  Watson	
  and	
  the	
  final	
  checkmate	
  
The	
  joining	
  of	
  genetics	
  to	
  bits,	
  the	
  essential	
  merger	
  

March	
  03,	
  2011,	
  11:14	
  AM	
  	
  

By	
  Ricardo	
  Murer	
  
B.S.	
  in	
  Computer	
  Science	
  (USP)	
  and	
  Master's	
  Degree	
  in	
  Communications	
  (USP).	
  
Specialist	
  in	
  digital	
  strategy	
  and	
  new	
  technologies.	
  
Follow@rdmurer	
  

Artificial	
  Intelligence	
  (AI)	
  is	
  a	
  field	
  of	
  research	
  that	
  is	
  both	
  fascinating	
  and	
  mysterious,	
  
influenced	
  by	
  science	
  fiction	
  to	
  such	
  extent	
  that	
  many	
  of	
  us	
  find	
  it	
  impossible	
  to	
  
differentiate	
  what	
  is	
  real	
  and	
  what	
  is	
  the	
  fruit	
  of	
  the	
  imagination	
  of	
  Hollywood	
  
writers	
  and	
  directors.	
  	
  

The	
  subject	
  recently	
  returned	
  to	
  the	
  media	
  when	
  IBM's	
  computer	
  "Watson”,	
  a	
  "Deep	
  
Question	
  &Answer"	
  (DeepQA)	
  machine,	
  participated	
  in	
  the	
  American	
  TV	
  show	
  
Jeopardy!,	
  beating	
  two	
  of	
  its	
  human	
  competitors.	
  The	
  show	
  is	
  exquisite	
  by	
  
submitting	
  the	
  answers	
  to	
  participants	
  who	
  must	
  then	
  find	
  the	
  corresponding	
  
question.	
  	
  

Watson	
  is	
  in	
  fact	
  a	
  combination	
  of	
  90	
  servers,	
  16	
  Terabytes	
  of	
  memory,	
  and	
  a	
  
processing	
  capacity	
  of	
  180,000	
  Gigabytes	
  per	
  second!	
  The	
  most	
  beautiful	
  and	
  purest	
  
parallel	
  brute	
  force	
  of	
  computing	
  today.	
  This	
  is	
  not	
  the	
  first	
  time	
  that	
  a	
  "deep"	
  beats	
  
a	
  human	
  being	
  in	
  a	
  competition.	
  In	
  1997,	
  "Deep	
  Blue"	
  (you	
  must	
  remember),	
  an	
  IBM	
  
chess	
  computer,	
  beat	
  the	
  then	
  champion	
  Garry	
  Kasparov.	
  Are	
  we	
  experiencing	
  the	
  
onset	
  of	
  our	
  decline,	
  our	
  defeat	
  to	
  machines	
  smarter	
  than	
  we	
  are?	
  

Fortunately,	
  not	
  yet.	
  In	
  both	
  cases,	
  we	
  are	
  referring	
  to	
  specialized	
  machines,	
  
computers	
  able	
  to	
  perform	
  efficiently,	
  probabilistically	
  and	
  extremely	
  rapidly,	
  in	
  the	
  
analysis	
  and	
  weighing	
  of	
  millions	
  of	
  possibilities,	
  and	
  through	
  a	
  combination	
  of	
  
algorithms,	
  make	
  the	
  least	
  unfavorable	
  decision.	
  	
  

While	
  then	
  scientists	
  had	
  initially	
  intended	
  to	
  reproduce	
  the	
  laws	
  of	
  thought	
  and	
  
create	
  a	
  machine	
  that	
  emulated	
  the	
  human	
  being,	
  the	
  approach	
  today	
  is	
  different.	
  
That	
  is	
  because	
  the	
  objective	
  of	
  creating	
  a	
  machine	
  that	
  has	
  the	
  capacity	
  to	
  learn,	
  
contemplate,	
  think	
  and	
  rationalize	
  has	
  proven	
  to	
  be	
  well	
  beyond	
  our	
  capacity.	
  	
  

The	
  complexity	
  of	
  such	
  a	
  task	
  stems	
  from	
  the	
  fact	
  that	
  intelligence	
  cannot	
  be	
  defined	
  
and	
  understood	
  solely	
  by	
  one	
  field	
  of	
  knowledge	
  but	
  by	
  the	
  sum	
  of	
  all	
  fields	
  such	
  as	
  
biology,	
  philosophy,	
  psychology	
  and	
  mathematics.	
  Among	
  the	
  new	
  approaches,	
  
connectionism	
  attempts	
  to	
  draw	
  computer	
  processes	
  close	
  to	
  the	
  human	
  brain	
  
through	
  "neural	
  networks",	
  while	
  the	
  symbolism	
  is	
  behind	
  specialist	
  systems.	
  	
  

AI	
  today	
  has	
  concentrated	
  many	
  of	
  its	
  efforts	
  into	
  simulating	
  human	
  cognitive	
  
capacities	
  though	
  simulation	
  and	
  understanding	
  of	
  natural	
  language,	
  recognition	
  of	
  
patterns	
  and	
  learning.	
  	
  


	
                                                                1	
  
However,	
  we	
  are	
  still	
  far	
  from	
  creating	
  the	
  perfect	
  simulation	
  of	
  the	
  human	
  brain	
  
and	
  eventually	
  something	
  that	
  can	
  be	
  deemed	
  the	
  result	
  of	
  intelligence,	
  even	
  if	
  by	
  
something	
  of	
  a	
  rudimentary	
  intelligence.	
  Perhaps	
  it	
  is	
  because	
  the	
  computer,	
  the	
  
one	
  that	
  counts,	
  or	
  the	
  ordinateur,	
  the	
  one	
  that	
  sorts	
  to	
  order	
  and	
  classifies,	
  remain,	
  
in	
  spite	
  all	
  techno-­‐scientific	
  advancements,	
  a	
  spectacular	
  calculating	
  and	
  counting	
  
machine,	
  and	
  nothing	
  more.	
  	
  

Perhaps	
  because	
  genetics	
  is	
  yet	
  to	
  join	
  with	
  bits,	
  the	
  essential	
  merger	
  able	
  to	
  bring	
  
light	
  to	
  the	
  first	
  thinking	
  android,	
  which	
  will	
  from	
  generation	
  to	
  generation	
  create	
  
new	
  machines	
  able	
  to	
  evolve	
  from	
  simple	
  toolmakers	
  to	
  builders	
  of	
  complex	
  
societies.	
  With	
  the	
  passing	
  of	
  time,	
  they	
  will	
  be	
  capable	
  of	
  simulating	
  not	
  just	
  
intelligence,	
  but	
  also	
  other	
  human	
  capacities	
  such	
  as	
  aesthetic	
  feeling,	
  moral	
  
judgment,	
  friendship	
  and	
  emotions,	
  converting	
  what	
  was	
  once	
  artificial	
  into	
  natural.	
  
The	
  final	
  checkmate.	
  

	
  




	
                                                               2	
  

Artificial intelligence, watson and the final checkmate

  • 1.
    Artificial  Intelligence,  Watson  and  the  final  checkmate   The  joining  of  genetics  to  bits,  the  essential  merger   March  03,  2011,  11:14  AM     By  Ricardo  Murer   B.S.  in  Computer  Science  (USP)  and  Master's  Degree  in  Communications  (USP).   Specialist  in  digital  strategy  and  new  technologies.   Follow@rdmurer   Artificial  Intelligence  (AI)  is  a  field  of  research  that  is  both  fascinating  and  mysterious,   influenced  by  science  fiction  to  such  extent  that  many  of  us  find  it  impossible  to   differentiate  what  is  real  and  what  is  the  fruit  of  the  imagination  of  Hollywood   writers  and  directors.     The  subject  recently  returned  to  the  media  when  IBM's  computer  "Watson”,  a  "Deep   Question  &Answer"  (DeepQA)  machine,  participated  in  the  American  TV  show   Jeopardy!,  beating  two  of  its  human  competitors.  The  show  is  exquisite  by   submitting  the  answers  to  participants  who  must  then  find  the  corresponding   question.     Watson  is  in  fact  a  combination  of  90  servers,  16  Terabytes  of  memory,  and  a   processing  capacity  of  180,000  Gigabytes  per  second!  The  most  beautiful  and  purest   parallel  brute  force  of  computing  today.  This  is  not  the  first  time  that  a  "deep"  beats   a  human  being  in  a  competition.  In  1997,  "Deep  Blue"  (you  must  remember),  an  IBM   chess  computer,  beat  the  then  champion  Garry  Kasparov.  Are  we  experiencing  the   onset  of  our  decline,  our  defeat  to  machines  smarter  than  we  are?   Fortunately,  not  yet.  In  both  cases,  we  are  referring  to  specialized  machines,   computers  able  to  perform  efficiently,  probabilistically  and  extremely  rapidly,  in  the   analysis  and  weighing  of  millions  of  possibilities,  and  through  a  combination  of   algorithms,  make  the  least  unfavorable  decision.     While  then  scientists  had  initially  intended  to  reproduce  the  laws  of  thought  and   create  a  machine  that  emulated  the  human  being,  the  approach  today  is  different.   That  is  because  the  objective  of  creating  a  machine  that  has  the  capacity  to  learn,   contemplate,  think  and  rationalize  has  proven  to  be  well  beyond  our  capacity.     The  complexity  of  such  a  task  stems  from  the  fact  that  intelligence  cannot  be  defined   and  understood  solely  by  one  field  of  knowledge  but  by  the  sum  of  all  fields  such  as   biology,  philosophy,  psychology  and  mathematics.  Among  the  new  approaches,   connectionism  attempts  to  draw  computer  processes  close  to  the  human  brain   through  "neural  networks",  while  the  symbolism  is  behind  specialist  systems.     AI  today  has  concentrated  many  of  its  efforts  into  simulating  human  cognitive   capacities  though  simulation  and  understanding  of  natural  language,  recognition  of   patterns  and  learning.       1  
  • 2.
    However,  we  are  still  far  from  creating  the  perfect  simulation  of  the  human  brain   and  eventually  something  that  can  be  deemed  the  result  of  intelligence,  even  if  by   something  of  a  rudimentary  intelligence.  Perhaps  it  is  because  the  computer,  the   one  that  counts,  or  the  ordinateur,  the  one  that  sorts  to  order  and  classifies,  remain,   in  spite  all  techno-­‐scientific  advancements,  a  spectacular  calculating  and  counting   machine,  and  nothing  more.     Perhaps  because  genetics  is  yet  to  join  with  bits,  the  essential  merger  able  to  bring   light  to  the  first  thinking  android,  which  will  from  generation  to  generation  create   new  machines  able  to  evolve  from  simple  toolmakers  to  builders  of  complex   societies.  With  the  passing  of  time,  they  will  be  capable  of  simulating  not  just   intelligence,  but  also  other  human  capacities  such  as  aesthetic  feeling,  moral   judgment,  friendship  and  emotions,  converting  what  was  once  artificial  into  natural.   The  final  checkmate.       2