The Manifest Destiny of Artificial Intelligence
Will AI create mindlike machines, or will it show how much a mindless machine can do?
Applied Computer Science
Edward A. Feigenbaum, a veteran of AI’s first-generation, has declared that “computational intelligence is the manifest destiny of computer science.” The slogan “manifest destiny” once expressed the sea-to-shining-sea territorial ambitions of the young United States. Feigenbaum, by analogy, is telling us there’s no stopping AI until it reaches the level of human intelligence. (And then why stop there?)
Feigenbaum’s declaration reiterates Minsky’s prophecy from 50 years ago. They may both be proved right, one of these days. In the meantime, I see a different kind of territorial aggrandizement going on. AI is expanding into turf that once belonged to other specialities. It looks like the destiny of artificial intelligence may be to assimilate all the rest of applied computer science.
I came to appreciate how much the field has broadened as I was reading Artificial Intelligence: A Modern Approach, a recent textbook by Stuart Russell and Norvig. (The third edition came out in 2010.) It’s a splendid book, and I recommend it not just for students of AI but for anyone seriously interested in computer science. And that’s the point: Many of the ideas and methods introduced here would be quite at home in a text on algorithm design or optimization theory. Some of the more traditional AI themes are scarcely mentioned.
Russell and Norvig give a brief history of AI, where recent developments are introduced under the rubric “AI adopts the scientific method.” This characterization seems a bit heavy-handed. Are we to conclude that previous generations were unscientific, toiling in benighted pursuit of cognitive phlogiston? The book’s subtitle, “A Modern Approach,” reinforces this impression. I’m sure the intent is not to be dismissive or scornful; it’s just that the questions that animated AI research in its first decades no longer seem so urgent or central. But that’s not because those questions have all been answered.
Meanwhile, the brash new style of AI plunges ahead. The roaring success of all those “shallow” methods—such as treating natural language as a sequence of n-grams—is something I find both exciting and perplexing. Exciting because these are algorithms we can implement and programs we can run; AI becomes a technology rather than a daydream. Perplexing because the shallow methods weren’t supposed to work, and now we have to explain their unreasonable effectiveness. Perhaps this is how we’ll get back to those deeper questions that Minsky warns we are neglecting.
- Brown, P. F., et al. 1990. A statistical approach to machine translation. Computational Linguistics 169(2):79–85.
- Etzioni, O. 2011. Search needs a shake-up. Nature 476:25–26.
- Feigenbaum, E. A. 2003. Some challenges and grand challenges for computational intelligence. Journal of the ACM 50:32–40.
- Feigenbaum, E. A., and J. Feldman. 1963. Computers and Thought. New York: McGraw-Hill.
- Ferrucci, D., et al. 2010. Building Watson: An overview of the DeepQA project. AI Magazine 31(3):59–79.
- Green, B. F., et al. 1961. Baseball: An automatic question answerer. In Proceedings of the Western Joint Computer Conference, February 1961, Vol. 19., pp. 219–224.
- Hutchins, J. 1986. Machine Translation: Past, Present, Future. Chichester, U.K.: Ellis Horwood.
- McCorduck, P. 2004. Machines Who Think: A Personal Inquiry into the History and Prospects of Artificial Intelligence. Twenty-Fifth Anniversary Edition. Natick, MA: A. K. Peters.
- Minsky, M. 1961. Steps toward artificial intelligence. Proceedings of the IRE 49:8–30.
- Nilsson, N. J. 2010. The Quest for Artificial Intelligence: A History of Ideas and Achievements. Cambridge: Cambridge University Press.
- Russell, S. J., and P. Norvig. 2010. Artificial Intelligence: A Modern Approach. Third edition. Upper Saddle River, NJ: Prentice Hall.
- Samuel, A. L. 1959. Some studies in machine learning using the game of checkers. IBM Journal of Research and Development 3:210–229.
- Schaeffer, J., et al. 2007. Checkers is solved. Science 317:1518–1522.
- Wilks, Y. 1973. The artificial intelligence approach to machine translation. In Computer Models of Thought and Language, R. C. Shank and K. M. Colby, eds. San Francisco: W. H. Freeman, pp. 114–151.
» Post Comment