Unformatted text preview: ARTIFICIAL INTELLIGENCE 1 GOALS GOALS
• Motivation – Are there social implications from Artificial Intelligence? • Provide a brief review of Artificial Intelligence. • Explore the issues concerning the construction of HAL-like machines. • Initiate discourse on digital tears: - emotion and the computer interface - can computers actually HAVE emotions? • Strong AI - can “machines” have sentience?
Any sci-fi buff knows that when computers become Any self-aware, they ultimately destroy their creators. From 2001: A Space Odyssey to Terminator, the message is clear: The only good self-aware machine is an unplugged one. We may soon find out whether that's true. ... But what about HAL 9000 and the other fictional computers that have run amok? "In any kind of technology there are risks," [Ron] Brachman acknowledges. That's why DARPA is reaching out to neurologists, psychologists - even philosophers - as well as computer scientists. "We're not stumbling down some blind alley," he says. "We're very cognizant of these issues." cognizant
View by Bruce Sterling. Wired Magazine (May 2004; Issue View 12.05). "Since when do machines need an ethical code? For 80 years, visionaries have imagined robots that look like us, work like us, perceive the world, judge it, and take action on their own. The robot butler is still as mystical as the flying car, but there's trouble rising in the garage. In Nobel's vaulted ballroom, experts uneasily point out that automatons are challenging humankind on four fronts. First, this is a time of war. ... The prospect of autonomous weapons naturally raises ethical questions. ... The autonomous second ominous frontier is brain augmentation, best embodied brain best by the remote-controlled rat recently created at SUNY Downstate in Brooklyn. ... Another troubling frontier is physical, as opposed to mental, augmentation. ... Frontier number four is social: to ... human reaction to the troubling presence of the humanoid. ... If the [First International Symposium on humanoid ... Roboethics] offers a take-home message, it's not about robots, 4 but about us." www.aaai.org/AITopics/html/ethics.html 5 Historical Development of AI AI
N e u r o n A b s t r a c t io n S y m b o lic C o m p u t e r S c ie n c e W eak A I K n o w le d g e R e p r e s e n t a t io n S e a rc h M e th o d s C o g n it iv e S c ie n c e N e u ra l N e tw o rk s P e rc e p to n s C o n n e c t io n is m HAL 9000 1997? S tro n g A I 6 Early Enthusiasm “We have invented a computer program We capable of thinking nonnumerically, and thereby solved the venerable mind-body problem.” -Simon, on Logic Theorist (1963) -Simon,
7 40 Years Later…
• “One of the most frustrating lessons computers have taught us is that many of the actions we think of as difficult are easy to automate and vice versa. In 1944 dozens of people spent months performing the calculations required for the Manhattan Project. Today the technology to do the same thing costs pennies. In contrast, when researchers met at Dartmouth College in the summer of 1956 to lay the groundwork for AI, none of them imagined that 40 years later we would have come such a short distance toward that goal.”
– D. Lenat, Scientific American, September 1995 8 Two Fundamental Paradigms: The Symbolic Paradigm Symbolic
• Dominant paradigm. • Knowledge is kept separate from the inference engine. • Early phase emphasized the inference engine. • Feigenbaum (1977) demonstrated that the knowledge base is more important. • Based on the Physical Symbol System Hypothesis (Newell and Simon, 1976).
9 Two Fundamental Paradigms: The Connectionist Paradigm Connectionist
• The “natural” approach. • Analog computation done by a linked network of very simple processors. • Emphasizes learning and discovering representations (e.g., pattern recognition). • No separation of knowledge from the inference mechanism. • Complex network of SIMPLE PROCESSORS (in contrast to Von Neumann Architecture)
10 Hybrid Schemes Hybrid
• Merger of the two paradigms. • Knowledge is power. • Neither paradigm is a necessary means for intelligent action. • Both paradigms are different abstractions on intelligence. 11 Four Approaches on Defining AI Four
• Thinking Humanly - The Cognitive Modeling Approach • Acting Humanly - The Turing Test Approach • Thinking Rationally - The Laws of Thought Approach • Acting Rationally - The Rational Agent Approach Russell and Norvig, 1995
12 Three Approaches to AI: Research According to Marvin Minsky Minsky
• Case-based (stored problems and solutions). • Rule-based (a.k.a., Expert Systems, brittle). • Connectionist
“Connectionists take pride in not understanding how a network solves a problem.” 13 Do We Need a Better Understanding of the Brain? of
• Do we need connectionism? • Do we need a theory of mind? “Many of the tasks that are easy for people to do (figure out a slurred word in a conversation or recognize a friend’s face) are all but impossible to automate, because we have no idea of how we do such things. Who can write down the rules for recognizing a face?” - Lenat, 1995
14 The Four “Common” Fields of AI Research Research
• Expert Systems (programs that mimic the decision-making and problem-solving thought process of human experts. • Robotics (Machines that can move and relate objects as humans can - includes vision). • Natural Languages (Systems that translate ordinary human commands into language computers can understand and act on). • Problem Solving (Programs that cover a broad spectrum of problems, from games to military).
15 End Of Overview End Now, to “Computational Creativity” 16 Can Computers be Creative? Can
Definition: Creativity • Creativity is a human mental phenomenon based around the deployment of mental skills and/or conceptual tools, which, in turn, originate and develop innovation, inspiration, or insight. (Wikipedia.org) • A mental process characterized by the inspiration of new and novel ideas – sometimes characterized by subconscious inferences. 17 Recall The Lady Lovelace Recall Objection Objection
“It has no pretensions to originate anything. It can do whatever we know how to order it to perform.”
- Lady Ada Lovelace 18 Recall Refuting Lady Recall Lovelace Lovelace
• “The Lady Lovelace Objection” is still very common today. • It is refuted simply by noting that one of the things we can tell them to do is to learn from their experience (e.g., game-playing). • The claim that the program’s ability to learn originated in the programmer is also refuted by noting that the programmer’s ability had to originate in other “programmers.”
19 Weak AI
Can machines be made to act as if they were intelligent?
20 The Turing Test The
In 1950, Turing defined intelligent behavior as the ability to achieve human-level performance in all cognitive tasks, sufficient to fool an interrogator. Such a computer would require the following:
• • • • Natural Language Processing to enable it to communicate in some human language; Knowledge Representation to store information provided before or during the interrogation; Automated Reasoning to use the stored information to answer questions and to draw new conclusions; Machine Learning to adapt to new circumstances and to detect and extrapolate patterns.
21 The Total Turing Test The
Add the following features:
• Computer Vision to perceive objects, and • Robotics to move them around. 22 The Loebner Prize The The Turing Test has The not been passed not
23 Does HAL Pass the Turing Test? 24 Prerequisites for HAL Prerequisites
Vision (eyes) “That’s a nice rendering, Dave. Could you hold it a bit closer?” 25 Prerequisites for HAL Prerequisites
Speech Recognition and Speech Understanding (ears) Understanding “Listen, HAL, there’s never been any instance at all of a computer error occurring in a 9000 series, has there?” “all” is not a place Homonym (there vs. their)
26 Prerequisites for HAL Prerequisites
Text-to-speech synthesis (voice) Douglas Rain (the voice of HAL) recorded all his lines over one weekend, speaking in an attractive, mellow, expressive tone quite different from the usual mechanical monotone attributed to computers in that period. 27 Prerequisites for HAL Prerequisites
Language usage (knowledge) “… the problem is not one of language, but of knowledge and the acquisition of knowledge. A computer would need to know a great deal to engage in even simple dialogue.”
- R. Schank
28 Other HAL Issues
• • • • • • Reliability Game playing (realized) Planning (“I can’t allow that to happen.”) Emotions Conscious experience Murder 29 HAL 9000 is a Bad Idea! HAL
Computer scientist insists on machines that make people smart (rather than making machines smart). Ben Shneiderman, a Professor of Computer Science and Head of the Human-Computer Interaction Laboratory at the University Maryland, will delivered the opening plenary address at CHI 98. He insisted that the best use of computers is to enhance human abilities, and that the goal of making computer user interfaces intelligent or anthropomorphic is misguided.
30 Carnegie-Mellon Conference Conference
• A one-day conference on October 19, 2001 • Assembly of experts (including Arthur C. Clarke) • Their mission: to answer the question of whether computers would help or hinder the building of a good world in the year 2050.
31 A SIGNIFICANT CONCLUSION CONCLUSION
• HAL will be realized within 50 years time - Just more blind optimism? - If true, what are the social ramifications?
Amer: “In talking to the computer, one gets Amer the sense that he is capable of emotional responses…. Do you believe that HAL has genuine emotions?” Dave: “Well, he acts like he has genuine emotions - umm of course he’s programmed that way to make it easier to talk to him. that But as to whether or not he has real feelings real is something I don’t think anyone can truthfully answer.”
33 Question Question
What IS emotion? “Since Plato, most philosophers have drawn a sharp line between reason and emotion, assuming that emotions interfere with rationality and have nothing to contribute to good reasoning. But current research in cognitive science [suggests] that emotion and reason are closely intertwined.”
- Paul Thagard, Waterloo
34 Definition of Emotion Definition
• Hundreds of philosophers and researchers have tried to create an exact definition of emotion. Unfortunately, the only common ground among a myriad of writers is the conclusion that emotion is not easy to define (Richins 1997). 35 Emotions - Two Issues Emotions
1. The User Interface 2. Genuine emotions as an emergent property? Question: Does the latter require “Strong AI?” 36 Strong AI Strong
Must machines that act intelligently necessarily possess real, conscious minds?
“I am putting myself to the fullest possible use, which is all I think that any conscious entity can ever hope to do.” - HAL 9000
37 Question Question
What IS conscious experience? • The most fundamental question in philosophy. • We have made literally NO progress answering this question in all of history. • In fact, nobody can say what the question of consciousness means in the first place. • The question may lie beyond the purview of scientific scrutiny (unless one can demonstrate how the subjective collapses into the objective, since science is centered upon the objective). Excellent references:
38 From Chalmer’s, “The Puzzle of Conscious Experience”
Conscious experience is at once the most familiar Conscious thing in the world and the most mysterious. There is nothing we know about more directly than consciousness, but it is extraordinarily hard to reconcile it with everything else we know. Why does it exist? What does it do? How could it possibly arise from neural processes in the brain? These questions are among the most intriguing in all of science.
39 Mazlish’s Four Discontinuities Mazlish’s
1. Earth - Heavens (Copernicus). 2. Human - Animal (Darwin). 3. Extension of Darwinism to mind 3. (Freud). (Freud). 4. Animate - Inanimate (Mazlish, 1970). All have been removed but the last.
40 FACT FACT
• At least 10% of this class has not gotten past the second discontinuity. • This percentage is much higher for the general public. 41 “It would be absurd, of course, to contend that there It are no differences between man and machines. This would be the same reductio ad absurdum as involved reductio in claiming that because he is an animal, there is no difference between man and the other animals. The matter, of course, is one of degree. What is claimed here is that the sharp discontinuity between man and machines is no longer tenable, in spite of the shock to our egos. Scientists, today, know this; the public at large does not…” - B. Mazlish B.
42 Nobel Laureate, Sir Francis Crick Crick
• British physicist/biochemist • Co-discoverer of the molecular structure of DNA (Nobel Prize, 1962) • Author of “The Astonishing Hypothesis: The Scientific Search for the Soul.” 43 CRICK’S DEFINTION of CRICK’S The Astonishing Hypothesis
• From Crick’s Glossary: “The hypothesis that a person’s mental activities are entirely due to the behavior of nerve cells, glial cells, and the atoms, ions, and molecules that make them up and influence them.” • In other words, the “astonishing hypothesis” is that the soul is MYTH.
44 The Fundamental Question of Philosophy Philosophy
What am I? Question: Can the construction of HAL-like machines help us to answer this question (or better yet, help us to figure out what it is we ARE asking?)
45 The Mind-Body Problem The
• Age old question: “How can we combine mind and brain – two quite distinct entities” • Descartes posited an independent entity (the soul) which interacts with the brain (known as interactionalist dualism – not widely accepted today) • Materialism rejects the soul concept, accepting only the existence of material things (one form of materialism is strong AI, which contends that a computer with the right program would be mental)
46 The Chinese Room Argument The
• Axiom 1: Computer programs are formal (syntactic) • Axiom 2: Human minds have mental contents (semantic) • Axiom 3: Syntax by itself is neither constitutive of nor sufficient for semantics • Conclusion: Programs are neither constitutive of nor sufficient for minds. 47 On Architecture On • Are Von Neumann machines homomorphic with neural networks? • If not, then could an “artificial” neural network yield strong AI? 48 In Summary In
• Are computers going to outsmart humans or are computers going to make us smarter? • Are computers going to help us address the age-old riddle of conscious experience? • What is the future of the user interface?
49 COMPUTER MALFUNCTION
View Full Document