제 2 강 생각하는기계 < 인공지능입문 > 강의노트 장병탁서울대학교컴퓨터공학부 & 인지과학 / 뇌과학협동과정 http://bi.snu.ac.kr/~btzhang/ Version: 20180312=> 20180313
목차 튜링테스트...... 3 중국어방논증........... 7 강인공지능과약인공지능..... 8 특이점....... 10 의식의문제와인공지능........ 13 인공지능과인간의삶..... 16 Reading Assignments........ 18 2018 Byoung- Tak Zhang, Seoul National University 2
튜링테스트 Alan Turing (1912-1954) Computing Machinery and Intelligence (1950) 2005, SNU Biointelligence Lab, http://bi.snu.ac.kr/ 3
생각하는 기계 Deep Blue Watson 1997 2011 AlphaGo 2016 4
제 2 강 튜링테스트와기능주의 q 기능주의 q 물리주의 q 계산주의 ( 마음이론 ) 2018 Byoung- Tak Zhang, Seoul National University 5
제 2 강 기계가생각할수있는가? q Turing (1950): Can Machines Think? 9 Objections 1. 신학적반론 (The Theological Objection) 2. 두려움반론 (The Heads in the Sand Objection) 3. 수학적반론 (The Mathematical Objection) 4. 의식성반론 (The Argument from Consciousness) 5. 무능력반론 (The Argument from Various Disabilities) 6. 창의성반론 (Lady Lovelace s Objection) 7. 신경시스템반론 (The Argument from Continuity in the Nervous System) 8. 행동의비형식화반론 (The Argument from Informality of Behavior) 9. 초감각지각반론 (The Argument from Extra- Sensory Perception) q Dreyfus (1972) Ø What Computers Can t Do: A Critique of Artificial Reason q Dreyfus (1992) Ø What Computers Still Can t Do: A Critique of Artificial Reason 2018 Byoung- Tak Zhang, Seoul National University 6
제 2 강 중국어방논증 q Searle (1980, 1984): 중국어방논증 (Chinese Room Argument) 전제 1. 뇌는마음을발생시킨다. 전제 2. 통사법은의미론에필수적인것이아니다. 전제 3. 컴퓨터프로그램은전적으로그것의형식적이고 통사론적구조에의해정의된다. 전제 4. 마음은의미론적내용을가지고있다. Ø 결론 1. 어떤컴퓨터프로그램도그스스로에의해마음의시스템과같은효과를주는데충분하지못하다. 프로그램은마음이아니다. Ø 결론 2. 마음을발생시키는뇌의기능이작동하는방식은다만컴퓨터프로그램을작동시키는능력에의한것만은아니다. Ø 결론 3. 마음을가진 ( 혹은마음을발생시키는 ) 모든것들은최소한뇌와동등한수준의인과력 (causal powers) 을가지고있어야만한다. 2018 Byoung- Tak Zhang, Seoul National University 7
제 2 강 강인공지능과약인공지능 q 좁은인공지능과넓은인공지능 Ø Narrow AI vs. broad AI, Specialists vs. generalists Ø Expert systems = Narrow AI q 강인공지능 vs. 약인공지능 Ø 강인공지능 : 인간의지능과동등한지능이기계로구현될수있다고보는입장 Ø 약인공지능 : 기계에지능을구현하는것이인간지능이해에도움이되고유용하다는입장 q 인간수준인공지능, 범용인공지능 Ø Human- level AI Ø AGI = Artificial General Intelligence Ø AGI vs. strong AI q 특이점 Ø 인공지능이인간의지능을능가하는시점 Ø Superhuman AI 2018 Byoung- Tak Zhang, Seoul National University 8
Technology AI with Deep Learning Narrow AI 1980 인공지능의수준 Cognitive AI Follows given goals and methods 1990 2010 2020 Superhuman AI Human-Level AI = AGI Works out own goals Agency Works out own methods, follows given goals 2030 Free Will Time 2050 Modified from Eliezer Yudkowsky & David Wood 9
특이점 (Singularity) Ray Kurzweil 10
지능폭발 (Good, 1965) 11
수퍼지능 12
제 2 강 의식의문제와인공지능 q 의식 (Consciousness) Ø Ø the state or quality of awareness, or, of being aware of an external object or something within oneself (Wikipedia) sentience, awareness, qualia, subjectivity, the ability to experience or to feel, wakefulness, having a sense of selfhood or soul, the fact that there is something "that it is like" to "have" or "be" it, and the executive control system of the mind. q 지향성 (Intention) q 자유의지 (Free Will) q 행위성 (Agency) q BDI (Belief, Desire, Intention) 2018 Byoung- Tak Zhang, Seoul National University 13
자유의지 (Free Will) 14
Libet 의실험 15
제 2 강 인공지능로봇 Humanoid Robot Nao (Aldebaran) Life- Like Robots (Hanson Robotics) Emotion Robot Pepper (SoftBank) Atlas (Boston Dynamics) (c) 2018 Biointelligence Lab, Seoul National University 16
인공지능과인간의삶 Her (2013) Robot & Frank (2012) Bicentennial Man (1999) Ex Machina (2015) 17
제 2 강 Reading (Watching) Assignments [Searle, 2015] Consciousness in Artificial Intelligence, Video (70 minutes), Talks at Google, 2015. Q: 인공지능이의식을가질수있는가? 이질문에대한 John Searle 의주장을설명하시오. [Kurzweil, 2017] Human- Level AI Is Just 12 Years Away (60 minutes). Q: 인공지능이인간의지능을능가하는특이점이올수있는가? 온다면언제오는가? 이질문에대한 Ray Kurzweil 의주장을설명하시오. 2018 Byoung- Tak Zhang, Seoul National University 18