April 25, Competing and cooperating with AI. Pantelis P. Analytis. Human behavior in Chess. Competing with AI. Cooperative machines?

Transcription

1 April 25, / 47

2 / 47

3 The case of chess 3 / 47

4 chess The first stage was the orientation phase, in which the subject assessed the situation determined a very general idea of what to do next. The second stage, the exploration phase was manifested by looking at some branches of the game tree. The third stage, or investigation phase resulted in the subject choosing a probable best move. Finally, in the fourth stage, the proof phase, saw the subject confirming with him/herself that the results of the investigation were valid. 4 / 47

5 chess Masters search through about the same number of possibilities as weaker players-perhaps even fewer, almost certainly not more-but they are very good at coming up with the right moves for further consideration, whereas weaker players spend considerable time analyzing the consequences of bad moves. De Groot found an intriguing difference between masters weaker players in his short-term memory experiments. Masters showed a remarkable ability to reconstruct a chess position almost perfectly after viewing it for only 5 sec. 5 / 47

6 Pattern recognition in chess experts (Chase Simon, 1978) 6 / 47

7 Pattern recognition in chess experts (Chase Simon, 1978) 7 / 47

8 Artificial intelligence timeline 8 / 47

9 Search strategies possible boards possible games since sequence matters. 9 / 47

10 Search strategies 10 / 47

11 Samuel s checkers player 11 / 47

12 Bernstein s chess player 12 / 47

13 Tesauro s TD-gammon 13 / 47

14 Kasparov vs. Deep Blue 14 / 47

15 IBM s Watson 15 / 47

16 winners in Jeopardy 16 / 47

17 Watson s progress 17 / 47

18 Evidence accumulation in humans 18 / 47

19 Performance In Atari games 19 / 47

20 Performance In Atari games 20 / 47

21 AlphaGo 21 / 47

22 AlphaGo 22 / 47

23 AlphaGo 23 / 47

24 Texas holdem Libratus 24 / 47

25 The graph coloring problem 25 / 47

26 The graph coloring problem (Kearns et al. 2006) 26 / 47

27 The graph coloring problem (Kearns et al. 2006) 27 / 47

28 With a little help from the bots (Shirado Christakis, 2017) 28 / 47

29 With a little help from the bots (Shirado Christakis, 2017) 29 / 47

30 Cooperation with machines, (Crall et al. 2017) 30 / 47

31 Cooperation with machines (Crall et al. 2017) 31 / 47

32 Cooperation with machines (Crall et al. 2017) 32 / 47

33 33 / 47

34 Elisa: the first chatbot Developed by Joseph Weizenbaum in 1966 to imitate the behavior of a Rogerian psychotherapist. 34 / 47

35 Parry: bot with schizophrenia the only serious interesting attempt by any program designer to win even a severely modified Turing has been Kenneth Colby. He had genuine psychiatrists interview PARRY. He did not suggest that they might be talking or typing to a computer; rather he made up some plausible story about why they were communicating with a real live patient via teletype. Then he took the PARRY transcript, inserted it into a group of teletype transcripts gave them to another group of experts?more psychiatrists? said, One of these was a conversation with a computer. Can you figure out which one it was? They couldn t. in Daniel Dennett 35 / 47

36 Captcha the inverse Turing 36 / 47

37 The trolley problem 37 / 47

38 The trolley problem: the fat guy 38 / 47

39 Variations of the problem 39 / 47

40 The autonomous car dilemma 40 / 47

41 The autonomous car dilemma Study one had 182 participants, while study two / 47

42 The autonomous car dilemma Study three had 259 participants, study four 267, study five 367 study six / 47

43 The position based model 43 / 47

44 The position based model 44 / 47

45 The cascade model 45 / 47

46 Dynamic Bayesian network model 46 / 47