Artificial Intelligence and Expertise: the Two Faces of the Same - - PowerPoint PPT Presentation

▶

Feb 04, 2023 326 likes •1.32k views

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin Matthieu Vergne Nara Institute of Science and

SLIDE 1

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin

Matthieu Vergne

Nara Institute of Science and Technology

HMCL Workshop – February 4th, 2017

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 1/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 2

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Outline

1 How is (Artificial) Intelligence Defined? 2 Debates Still Ongoing 3 A Strong Parallel with Expertise 4 Impact on AI 5 Conclusion

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 2/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 3

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of Psychology

Gottfredson [1997] got 52 signatories for a definition of intelligence: Intelligence is a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 3/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 4

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of Psychology

Gottfredson [1997] got 52 signatories for a definition of intelligence: Intelligence is a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience. Nice, but...

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 3/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 5

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of Psychology

Gottfredson [1997] got 52 signatories for a definition of intelligence: Intelligence is a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience. Nice, but... Open ended

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 3/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 6

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of Psychology

Gottfredson [1997] got 52 signatories for a definition of intelligence: Intelligence is a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience. Nice, but... Open ended Human-centred

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 3/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 7

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of Psychology

Gottfredson [1997] got 52 signatories for a definition of intelligence: Intelligence is a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience. Nice, but... Open ended Human-centred Not representative of all the perspectives in the field: psychometric, physiological, social, etc.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 3/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 8

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of Psychology

Gottfredson [1997] got 52 signatories for a definition of intelligence: Intelligence is a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience. Nice, but... Open ended Human-centred Not representative of all the perspectives in the field: psychometric, physiological, social, etc. No broadly agreed, precise definition yet. (Mackintosh [2011], Urbina [2011], Willis et al. [2011], Davidson and Kemp [2011], Franklin [2014])

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 3/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 9

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions:

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 10

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 11

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 12

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions. Few common features:

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 13

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions. Few common features:

interaction with the environment

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 14

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions. Few common features:

interaction with the environment ability to satisfy goals

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 15

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions. Few common features:

interaction with the environment ability to satisfy goals ability to adapt oneself

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 16

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions. Few common features:

interaction with the environment ability to satisfy goals ability to adapt oneself

Result: Intelligence measures an agent’s ability to achieve goals in a wide range of environments.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 17

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions. Few common features:

interaction with the environment ability to satisfy goals ability to adapt oneself

Result: Intelligence measures an agent’s ability to achieve goals in a wide range of environments. Machine Intelligence Research Institute (Muehlhauser [2013])

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 18

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions. Few common features:

interaction with the environment ability to satisfy goals ability to adapt oneself

Result: Intelligence measures an agent’s ability to achieve goals in a wide range of environments. Machine Intelligence Research Institute (Muehlhauser [2013]) Intelligence = efficient cross-domain optimization

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 19

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions. Few common features:

interaction with the environment ability to satisfy goals ability to adapt oneself

Result: Intelligence measures an agent’s ability to achieve goals in a wide range of environments. Machine Intelligence Research Institute (Muehlhauser [2013]) Intelligence = efficient cross-domain optimization Not necessarily wide set of domains

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 20

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

In the field of AI

Legg and Hutter [2007a] gathered many definitions: 70 definitions from dictionaries, encyclopedias, grouped statements, individual psychologists and AI researchers. From 1 aspect to open ended definitions. Few common features:

interaction with the environment ability to satisfy goals ability to adapt oneself

Result: Intelligence measures an agent’s ability to achieve goals in a wide range of environments. Machine Intelligence Research Institute (Muehlhauser [2013]) Intelligence = efficient cross-domain optimization Not necessarily wide set of domains Not only goal fulfilment, efficiency too

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 4/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 21

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Still a matter of debates

No clear view of when/where it will end

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 5/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 22

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Still a matter of debates

No clear view of when/where it will end AI is about reproducing intelligence

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 5/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 23

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Still a matter of debates

No clear view of when/where it will end AI is about reproducing intelligence So Psychology cannot be left aside

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 5/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 24

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Still a matter of debates

No clear view of when/where it will end AI is about reproducing intelligence So Psychology cannot be left aside But various perspectives with disagreements

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 5/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 25

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Still a matter of debates

No clear view of when/where it will end AI is about reproducing intelligence So Psychology cannot be left aside But various perspectives with disagreements What if we are just looking at the wrong place?

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 5/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 26

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Still a matter of debates

No clear view of when/where it will end AI is about reproducing intelligence So Psychology cannot be left aside But various perspectives with disagreements What if we are just looking at the wrong place? Intelligence is about performance

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 5/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 27

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Still a matter of debates

No clear view of when/where it will end AI is about reproducing intelligence So Psychology cannot be left aside But various perspectives with disagreements What if we are just looking at the wrong place? Intelligence is about performance But expertise too!

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 5/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 28

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Still a matter of debates

No clear view of when/where it will end AI is about reproducing intelligence So Psychology cannot be left aside But various perspectives with disagreements What if we are just looking at the wrong place? Intelligence is about performance But expertise too! And it has already a well agreed definition! (Ackerman [2011], Ericsson [2006])

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 5/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 29

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience.

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 30

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions:

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 31

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc.

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 32

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc. special knowledge: abstract knowledge, complex ideas, etc.

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 33

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc. special knowledge: abstract knowledge, complex ideas, etc. training/experience: learning, adaptiveness

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 34

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc. special knowledge: abstract knowledge, complex ideas, etc. training/experience: learning, adaptiveness Measures too (Ericsson [2006]):

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 35

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc. special knowledge: abstract knowledge, complex ideas, etc. training/experience: learning, adaptiveness Measures too (Ericsson [2006]): Reproducibly superior performance

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 36

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc. special knowledge: abstract knowledge, complex ideas, etc. training/experience: learning, adaptiveness Measures too (Ericsson [2006]): Reproducibly superior performance → IQ tests, AI in games

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 37

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc. special knowledge: abstract knowledge, complex ideas, etc. training/experience: learning, adaptiveness Measures too (Ericsson [2006]): Reproducibly superior performance → IQ tests, AI in games Lengthy, domain-related experience

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 38

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc. special knowledge: abstract knowledge, complex ideas, etc. training/experience: learning, adaptiveness Measures too (Ericsson [2006]): Reproducibly superior performance → IQ tests, AI in games Lengthy, domain-related experience → animal/AI survival

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 39

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc. special knowledge: abstract knowledge, complex ideas, etc. training/experience: learning, adaptiveness Measures too (Ericsson [2006]): Reproducibly superior performance → IQ tests, AI in games Lengthy, domain-related experience → animal/AI survival Social assessments

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 40

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Expertise vs. Intelligence

Definition of expert1: having, involving, or displaying special skill or knowledge derived from training or experience. Similar to intelligence definitions: special skill: reason, plan, solve problems, etc. special knowledge: abstract knowledge, complex ideas, etc. training/experience: learning, adaptiveness Measures too (Ericsson [2006]): Reproducibly superior performance → IQ tests, AI in games Lengthy, domain-related experience → animal/AI survival Social assessments → social theories, Turing test

1http://www.merriam-webster.com/dictionary/expert Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 6/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 41

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

From Expertise to Intelligence

Just a bit of refinement:

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 7/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 42

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

From Expertise to Intelligence

Just a bit of refinement: Expertise = special skills/knowledge, but intelligence too

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 7/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 43

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

From Expertise to Intelligence

Just a bit of refinement: Expertise = special skills/knowledge, but intelligence too Expertise is about a particular domain (domain-relevant skills/knowledge)

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 7/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 44

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

From Expertise to Intelligence

Just a bit of refinement: Expertise = special skills/knowledge, but intelligence too Expertise is about a particular domain (domain-relevant skills/knowledge) Intelligence is about genericity (domain-generic skills/knowledge)

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 7/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 45

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

From Expertise to Intelligence

Just a bit of refinement: Expertise = special skills/knowledge, but intelligence too Expertise is about a particular domain (domain-relevant skills/knowledge) Intelligence is about genericity (domain-generic skills/knowledge) Both are complementary aspects of high performance.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 7/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 46

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance:

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 47

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 48

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 49

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Hunting Generic Specific Relevant Hunt pheasants Irrelevant

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 50

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Hunting Generic Specific Relevant Hunt pheasants Irrelevant Hunt chicks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 51

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Domain-specificity (inverse of genericity): Hunting Generic Specific Relevant Hunt pheasants Irrelevant Hunt chicks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 52

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Domain-specificity (inverse of genericity): Reserved for the domain Hunting Generic Specific Relevant Hunt pheasants Irrelevant Hunt chicks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 53

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Domain-specificity (inverse of genericity): Reserved for the domain Using this skill/knowledge ⇒ Perform in the domain Hunting Generic Specific Relevant Hunt pheasants Irrelevant Hunt chicks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 54

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Domain-specificity (inverse of genericity): Reserved for the domain Using this skill/knowledge ⇒ Perform in the domain Hunting Generic Specific Relevant Search for patterns Hunt pheasants Irrelevant Hunt chicks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 55

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Domain-specificity (inverse of genericity): Reserved for the domain Using this skill/knowledge ⇒ Perform in the domain Hunting Generic Specific Relevant Search for patterns Hunt pheasants Irrelevant Imagine a story Hunt chicks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 56

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Domain-specificity (inverse of genericity): Reserved for the domain Using this skill/knowledge ⇒ Perform in the domain Hunting Generic Specific Expertise Search for patterns Hunt pheasants Irrelevant Imagine a story Hunt chicks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 57

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Domain-specificity (inverse of genericity): Reserved for the domain Using this skill/knowledge ⇒ Perform in the domain Hunting Intelligence Specific Relevant Search for patterns Hunt pheasants Irrelevant Imagine a story Hunt chicks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 58

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

About Relevance and Genericity

Domain-relevance: Important or significant for the domain Perform in the domain ⇒ Use this skill/knowledge Domain-specificity (inverse of genericity): Reserved for the domain Using this skill/knowledge ⇒ Perform in the domain Most generic: skills on/knowledge about the agent itself. Hunting Generic Specific Relevant Search for patterns Hunt pheasants Irrelevant Imagine a story Hunt chicks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 8/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 59

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . skill or knowledge derived from training or experience.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 60

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . skill or knowledge derived from training or experience. Generalise human-centred terms:

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 61

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . skill or knowledge derived from training or experience. Generalise human-centred terms: Knowledge

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 62

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . skill or data derived from training or experience. Generalise human-centred terms: Knowledge → Data

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 63

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . skill or data derived from training or experience. Generalise human-centred terms: Knowledge → Data Skill

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 64

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . process or data derived from training or experience. Generalise human-centred terms: Knowledge → Data Skill → Process

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 65

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . process or data derived from training or experience. Generalise human-centred terms: Knowledge → Data Skill → Process Experience

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 66

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . process or data obtained through training or generation. Generalise human-centred terms: Knowledge → Data Skill → Process Experience → Data/process generation

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 67

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . process or data obtained through training or generation. Generalise human-centred terms: Knowledge → Data Skill → Process Experience → Data/process generation Training

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 68

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . process or data obtained through transfer or generation. Generalise human-centred terms: Knowledge → Data Skill → Process Experience → Data/process generation Training → Data/process transfer

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 69

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Generalisation for non-Human Performance

Definition: having, involving, or displaying domain-. . . process or data obtained through transfer or generation. Generalise human-centred terms: Knowledge → Data Skill → Process Experience → Data/process generation Training → Data/process transfer

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 9/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 70

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP):

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 71

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP): Artificial Expertise (AE) = domain-relevant performance

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 72

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP): Artificial Expertise (AE) = domain-relevant performance Artificial Intelligence (AI’) = domain-generic performance

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 73

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP): Artificial Expertise (AE) = domain-relevant performance Artificial Intelligence (AI’) = domain-generic performance Many if not most of systems fall in the umbrella of AE:

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 74

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP): Artificial Expertise (AE) = domain-relevant performance Artificial Intelligence (AI’) = domain-generic performance Many if not most of systems fall in the umbrella of AE: Expert systems in medicine and banks

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 75

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP): Artificial Expertise (AE) = domain-relevant performance Artificial Intelligence (AI’) = domain-generic performance Many if not most of systems fall in the umbrella of AE: Expert systems in medicine and banks Recommendation systems in e-commerce

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 76

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP): Artificial Expertise (AE) = domain-relevant performance Artificial Intelligence (AI’) = domain-generic performance Many if not most of systems fall in the umbrella of AE: Expert systems in medicine and banks Recommendation systems in e-commerce AI in games

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 77

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP): Artificial Expertise (AE) = domain-relevant performance Artificial Intelligence (AI’) = domain-generic performance Many if not most of systems fall in the umbrella of AE: Expert systems in medicine and banks Recommendation systems in e-commerce AI in games Automatic translation

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 78

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP): Artificial Expertise (AE) = domain-relevant performance Artificial Intelligence (AI’) = domain-generic performance Many if not most of systems fall in the umbrella of AE: Expert systems in medicine and banks Recommendation systems in e-commerce AI in games Automatic translation etc.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 79

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Don’t say AI, but AP!

Two aspects of Artificial Performance (AP): Artificial Expertise (AE) = domain-relevant performance Artificial Intelligence (AI’) = domain-generic performance Many if not most of systems fall in the umbrella of AE: Expert systems in medicine and banks Recommendation systems in e-commerce AI in games Automatic translation etc. Works corresponding to AI’ would mainly relate to AGI (e.g. brain emulation).

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 10/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 80

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Summary

How to define intelligence?

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 11/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 81

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Summary

How to define intelligence? Converging definitions in AI

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 11/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 82

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Summary

How to define intelligence? Converging definitions in AI Main inspiration from humans lacks convergence

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 11/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 83

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Summary

How to define intelligence? Converging definitions in AI Main inspiration from humans lacks convergence Greater support when looking at expert performance

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 11/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 84

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Summary

How to define intelligence? Converging definitions in AI Main inspiration from humans lacks convergence Greater support when looking at expert performance Definitions based on expertise leads to definitions of intelligence close to AI definitions

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 11/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 85

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Summary

How to define intelligence? Converging definitions in AI Main inspiration from humans lacks convergence Greater support when looking at expert performance Definitions based on expertise leads to definitions of intelligence close to AI definitions Looking at AI as AP (AE+AI’) provides a non polemical view

f the field and helps to identify future works

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 11/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 86

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Summary

How to define intelligence? Converging definitions in AI Main inspiration from humans lacks convergence Greater support when looking at expert performance Definitions based on expertise leads to definitions of intelligence close to AI definitions Looking at AI as AP (AE+AI’) provides a non polemical view

f the field and helps to identify future works

What we did not speak about:

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 11/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 87

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Summary

How to define intelligence? Converging definitions in AI Main inspiration from humans lacks convergence Greater support when looking at expert performance Definitions based on expertise leads to definitions of intelligence close to AI definitions Looking at AI as AP (AE+AI’) provides a non polemical view

f the field and helps to identify future works

What we did not speak about: Intelligence as expertise acquisition

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 11/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 88

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Summary

How to define intelligence? Converging definitions in AI Main inspiration from humans lacks convergence Greater support when looking at expert performance Definitions based on expertise leads to definitions of intelligence close to AI definitions Looking at AI as AP (AE+AI’) provides a non polemical view

f the field and helps to identify future works

What we did not speak about: Intelligence as expertise acquisition Technological singularity vs. expertise evidence

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 11/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 89

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Future Work

Several aspects should be developed:

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 12/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 90

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Future Work

Several aspects should be developed: Formalisation (ongoing)

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 12/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 91

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Future Work

Several aspects should be developed: Formalisation (ongoing) Extension of existing systems (e.g. add domain-generic)

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 12/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 92

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Future Work

Several aspects should be developed: Formalisation (ongoing) Extension of existing systems (e.g. add domain-generic) Analysis of recursive self-improvement

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 12/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 93

Intelligence Definitions Debates Still Ongoing Parallel with Expertise Impact on AI Conclusion

Thanks for your attention. Questions?

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 13/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 94

References

P. L. Ackerman. Intelligence and Expertise. In R. J. Sternberg and S. B. Kaufman,

editors, The Cambridge Handbook of Intelligence, pages 847–860. Cambridge University Press, Cambridge, 2011. ISBN 978-0-511-97724-4. URL http://www.cambridge.org/us/academic/subjects/psychology/cognition/ cambridge-handbook-intelligence.

J. E. Davidson and I. A. Kemp. Contemporary Models of Intelligence. In R. J.

Sternberg and S. B. Kaufman, editors, The Cambridge Handbook of Intelligence, pages 58–82. Cambridge University Press, Cambridge, 2011. ISBN 978-0-511-97724-4. URL http://www.cambridge.org/us/academic/subjects/ psychology/cognition/cambridge-handbook-intelligence.

K. A. Ericsson. An Introduction to Cambridge Handbook of Expertise and Expert

Performance: Its Development, Organization, and Content. In K. A. Ericsson,

N. Charness, P. J. Feltovich, and R. R. Hoffman, editors, The Cambridge handbook
f expertise and expert performance, pages 3–19. Cambridge University Press, New

York, NY, US, 2006. ISBN 0-521-60081-2 (Paperback); 0-521-84097-X (Hardcover); 978-0-521-60081-1 (Paperback); 978-0-521-84097-2 (Hardcover).

S. Franklin. History, motivations, and core themes. In K. Frankish and W. M. Ramsey,

editors, The Cambridge handbook of artificial intelligence, pages 15–33. Cambridge University Press, Cambridge, UK, 2014. ISBN 978-0-521-87142-6.

L. S. Gottfredson. Mainstream science on intelligence: An editorial with 52

signatories, history, and bibliography. Intelligence, 24(1):13–23, Jan. 1997. ISSN

01602896. doi: 10.1016/S0160-2896(97)90011-8. URL

http://linkinghub.elsevier.com/retrieve/pii/S0160289697900118.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 13/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 95

References

S. Legg and M. Hutter. A Collection of Definitions of Intelligence. In B. Goertzel and
P. Wang, editors, Advances in artificial general intelligence: concepts, architectures

and algorithms: proceedings of the AGI Workshop 2006, number v. 157 in Frontiers in artificial intelligence and applications. IOS Press, Amsterdam ; Washington, DC,

2007a. ISBN 978-1-58603-758-1.
S. Legg and M. Hutter. Universal Intelligence: A Definition of Machine Intelligence.

Minds and Machines, 17(4):391–444, Nov. 2007b. ISSN 0924-6495, 1572-8641. doi: 10.1007/s11023-007-9079-x. URL http://link.springer.com/10.1007/s11023-007-9079-x.

N. J. Mackintosh. History of Theories and Measurement of Intelligence. In R. J.

Sternberg and S. B. Kaufman, editors, The Cambridge Handbook of Intelligence, pages 3–19. Cambridge University Press, Cambridge, 2011. ISBN 978-0-511-97724-4. URL http://www.cambridge.org/us/academic/subjects/ psychology/cognition/cambridge-handbook-intelligence.

L. Muehlhauser. What is Intelligence?, June 2013. URL

https://intelligence.org/2013/06/19/what-is-intelligence-2/.

S. Urbina. Tests of Intelligence. In R. J. Sternberg and S. B. Kaufman, editors, The

Cambridge Handbook of Intelligence, pages 20–38. Cambridge University Press, Cambridge, 2011. ISBN 978-0-511-97724-4. URL http://www.cambridge.org/us/academic/subjects/psychology/cognition/ cambridge-handbook-intelligence.

J. O. Willis, R. Dumont, and A. S. Kaufman. Factor-Analytic Models of Intelligence.

In R. J. Sternberg and S. B. Kaufman, editors, The Cambridge Handbook of

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 13/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 96

References

Intelligence, pages 39–57. Cambridge University Press, Cambridge, 2011. ISBN 978-0-511-97724-4. URL http://www.cambridge.org/us/academic/subjects/ psychology/cognition/cambridge-handbook-intelligence.

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 14/13 Matthieu Vergne vergne@is.naist.jp NAIST

SLIDE 97

References

Formal definition of intelligence

Legg and Hutter [2007b]

Achieve goals in a wide range of environments: Observations oi + rewards ri + actions ai (o1r1 → a1 → o2r2 → a2 → . . . ) Agent = π(ai|o1r1a1...oi−1ri−1ai−1oiri) (probability of actions given history) Environment = µ(oiri|o1r1a1...oi−1ri−1ai−1) (probability of observations and rewards given history) Total reward to be maximised = V π

µ (various models)

Priority to simple µ (Kolmogorov complexity K(µ)) Intelligence =

µ∈E 2−K(µ)V π µ

Artificial Intelligence and Expertise: the Two Faces of the Same Artificial Performance Coin 14/13 Matthieu Vergne vergne@is.naist.jp NAIST