The New AI: General and sound and relevant for physics

Jan, 2003
21 pages
e-Print:
Report number:
  • TR-IDSIA-04-03

Citations per year

20102011201201
Abstract: (arXiv)
Most traditional artificial intelligence (AI) systems of the past 50 years are either very limited, or based on heuristics, or both. The new millennium, however, has brought substantial progress in the field of theoretically optimal and practically feasible algorithms for prediction, search, inductive inference based on Occam's razor, problem solving, decision making, and reinforcement learning in environments of a very general type. Since inductive inference is at the heart of all inductive sciences, some of the results are relevant not only for AI and computer science but also for physics, provoking nontraditional predictions based on Zuse's thesis of the computer-generated universe.
Note:
  • To appear in B. Goertzel and C. Pennachin, eds. 'Artificial General Intelligence': DATE-UPDATED = 01/05/2004
  • dungsproblem. Proceedings of the London Mathematical Society, Series 2, 41:230- 267, 1936. 68 V. Vapnik. Random processes and transformations. In Proceedings of the International Congress on Mathematics, volume 2, pages 264-275, 1950. 69 The Nature of Statistical Learning Theory J. von Neumann, New York, 1995. 70 C. S. Wallace and D. M. Boulton. Theory of Self-Reproducing Automata. University of Illionois Press, Champain, IL, 1966. 71 An information theoretic measure for classification P. J. Werbos. 72 P. J. Werbos. Beyond Regression: New Tools for Prediction and Analysis in the Behavioral Sciences. PhD thesis, Harvard University, 1974. 73 Learning how the world works: Specifications for predictive networks in robots and brains. In Proceedings of M.A. Wiering and J. Schmidhuber International Conference on Systems, Man and Cybernetics, N.Y., 1987. 74 L. Saitta, editor. Solving POMDPs with Levin search and EIRA. In K. Zuse Machine Learning: Proceedings of the Thirteenth International Conference, pages 534-542. Morgan Kaufmann Publishers, San Francisco, CA, 1996. 75 K. Zuse. Rechnender Raum. Elektronische Datenverarbeitung, 8:336-344, 1967. 76 Rechnender Raum. Friedrich A. K. Zvonkin and L. A. Levin & Sohn, Braunschweig, 1969. English translation: Calculating Space, MIT Technical Translation AZT-70-164-GEMIT, Massachusetts Institute of Technology (Proj. MAC), Cambridge, Mass. 02139, Feb. 1970. 77 The complexity of finite objects and the algorithmic concepts of information and randomness
    • S. Ulam
      • Comput.J. 11 (1968) 185