Presentation on theme: "Possibility of True Artificial Intelligence. Weak AI: Can machines act intelligently Artificial intelligence pursued within the cult of computationalism."— Presentation transcript:
Possibility of True Artificial Intelligence
Weak AI: Can machines act intelligently Artificial intelligence pursued within the cult of computationalism stands not even a ghost of a chance of producing durable results … it is time to divert the efforts of AI researchers - and the considerable monies made available for their support - into avenues other than the computational approach. (Sayre, Three more flaws in the computational model. Paper presented at the APA (Central Division) Annual Conference, 1993)
Weak AI: Can machines act intelligently Impossibility of AI depends on its definition 1. AI is the quest for the best agent program on a given architecture. Clearly possible. 2. AI makes machines think. An under-specified problems Turing Test: An experimental answer
Weak AI: Can machines act intelligently Disability Objection Turing lists: Machines can never … Be kind, resourceful, beautiful, friendly, have initiative, have a sense of humor, tell right from wrong, make mistakes, fall in love, enjoy strawberries and cream, make someone fall in love with it, learn from experience, use words properly, be the subject of its own thought, have as much diversity of behavior as man, do something really new. How has this argument held up historically? Computers can play chess, check spelling, steer cars and helicopters, diagnose diseases, … Computers have made small, but significant discoveries in mathematics, chemistry, mineralogy, biology, computer science, … Algorithms perform at human or better level in some areas involving human judgement: Paul Meehl, 1955: Statistical learning algorithm outperform experts when predicting success of students in a training program or recidivism of criminals GMAT grades essays automatically since 1999.
Weak AI: Can machines act intelligently Mathematical objection Certain mathematical questions are in principle unanswerable by a particular formal system. Gödel (1931), Turing (1936) Formal axiomatic system F powerful enough to do arithmetic allows construction of Gödel sentences G(F) G(F) is sentence of F, but cannot be proved within F If F is consistent, then G(F) is true Lucas (1961): Machine are formal systems hence they are not capable of deriving Gödel sentences while humans have no such limitation Penrose (1989, 1994): Humans are different because their brains operate by quantum gravity
Weak AI: Can machines act intelligently Counterarguments to the Lucasian position Gödel’s incompleteness theorem does not apply: Gödel’s incompleteness theorem only applies to formal systems that are powerful enough to do arithmetic, such as Turing machines. Computers are not Turing machines, they are finite. Hence they are describable in a very large propositional logic system, where Gödel’s incompleteness theorem does not apply.
Weak AI: Can machines act intelligently Counterarguments to the Lucasian position There is no problem in that intelligence agents cannot establish the truth of some sentence while other agents can. “J. R. Lucas cannot consistently assert that this sentence is true” If Lucas asserts this sentence, then he would be contradicting himself, so he cannot assert it consistently and the sentence needs to be true. Why should we (who can assert this sentence) think less of Lucas because he cannot assert this sentence? Humans cannot add 100 Billion 100 digit numbers in their lifetime, but computers can. Humankind did very well without mathematics for millennia, so intelligence should not be made dependent on mathematical ability.
Weak AI: Can machines act intelligently Counterarguments to the Lucasian position Computers have limitations, but so do humans Humans are famously inconsistent
Weak AI: Can machines act intelligently Argument from informality of behavior (Dreyfus 1972, 1986, 1992) Claim: “Human behavior is for too complex to be captured by any simple set of rules.” “Computers cannot do more than follow as set of rules, therefore they cannot.” Known as Qualification Problem in AI Claim applies to “Good Old Fashioned AI” (GOFAI) Dreyfus critique morphed into a proposal of doing AI together with a list of “insurmountable” problems All of these problems were addressed by AI research
Strong AI: Can Machines Really Think “Not until a machine could write a sonnet or compose a concerto because of thoughts and emotions felt, and not by the chance fall of symbols, could we agree that machine equals brain - that is, not only write it but know that it had written it” (Geoffrey Jefferson, 1949, quoted by A. Turing) Argument from Consciousness, but relates to Phenomenology Study of direct experience Intentionality Whether the machine’s purported beliefs, desires, and other representations are actually “about” something in the real world
Strong AI: Can Machines Really Think Turing proposes “a polite convention” that everyone thinks We are interested in creating programs that behave intelligently, not whether someone else pronounces them to be real or simulated. We avoid the question: “When are artifacts considered real?” Is an artificial Picasso painting a Picasso painting? Are artificial sweeteners sweeteners? Distinction seems to depend on intuition
Strong AI: Can Machines Really Think Turing proposes “a polite convention” that everyone thinks Searle: “No one supposes that a computer simulation of a storm will leave us all wet … Why on earth would anyone in his right mind suppose a computer simulation of mental processes actually had mental processes (1980) Functionalism: Mental state is any intermediate causal condition between input and output. Biological Naturalism: Mental states are high-level emergent features that are caused by low-level neurological processes in the neurons and it is the properties of the neurons that matter. Give different answers to the challenge by Searle
Strong AI: Can Machines Really Think Mind-Body Problem: How are mental states and processes related to bodily states and processes Descartian dualism Monism or materialism (Searle: “Brains cause minds”.) Raises the further questions of Consciousness, Understanding, Self-Awareness Free will Brain in a vat experiment Brain prosthesis experiment: Replace brain parts over time with silicon and see what happens to self-consciousness
Strong AI: Can Machines Really Think Chinese Room Transformed to the following axioms 1. Computer programs are formal, syntactic entities 2. Minds have mental contents, or semantics 3. Syntax by itself is not sufficient for semantics 4. Brains cause minds Conclusion from 1, 2, 3: programs are not sufficient for minds Chinese room argument argues for 3
Social Consequences of Successful AI Berleur and Brunnstein: Ethics of Computing, 2001 People might loose their jobs because of automation People might have too much or not enough leisure time People might loose their sense of being unique People might loose some of their privacy rights The use of AI systems results in the loss of accountability The success of AI might mean the end of the human race
Social Consequences of Successful AI Arthur C. Clarke, 1968 People in 2001 might be “faced with a future of utter boredom, where the main problem in life is deciding which of several hundred TV channels to select”