close this bookVolume 4: No. 34
View the documentFunding news
View the documentAI news
View the documentHPC software development
View the documentBBS and Internet services
View the documentElectronic publishing
View the documentJob opportunities
View the documentDiscussion groups
View the documentPersonal finance
View the documentComputists' news

UAlberta's Chinook has taken the world checkers title by playing the #2 human player -- Don Lafferty -- to a 1-18-1 draw. Chinook ran on a Silicon Graphics Challenge XL with 16 CPUs, twice as many as two years ago. Jonathan Schaeffer says that Chinook knows 250B positions and thinks ahead at least 22 moves (vs. 18 two years ago). #1 player Marion Tinsley, 67, had to stop playing after six draws, for health reasons related to pancreatic cancer. Tinsley estimates that he looks ahead 30 moves, and sometimes as much as 65. [Dean Takahashi, SJM, 8/26/94.] (Perhaps we should measure width rather than depth. Depth is easy if your evaluation function is good enough to narrow the search.)

ATR (Japan) plans to build an artificial brain using "nanoscale self-assembly techniques." Researchers at the Human Information Processing Research Labs (Kyoto) want to grow the circuits with cellular automata controlled by genetic algorithms on parallel Darwin machines. When the genetic trails from two automata collide, a "synapse" is formed. Such a "neurite network" device might hold 10B cells with neural nets of 1M cells. A 16-neuron module is to be simulated later this year. [Colin Johnson, EETimes, 8/22/94, p. 35.]

Marvin Minsky is suggesting that the world's 50K expert systems be combined on one chip, along with 100MB or so of commonsense knowledge. Add a Pentium and he says you'd have all the processing power most users would need. The project could be done in five years, if it had unlimited funding. [Rory J. O'Connor (rjoconnor@aol.com), SJM, 8/31/94, 8D.] (Sounds like the CYC project, which Minsky has long favored. I don't see much urgency to combining heterogeneous expert systems, and the time estimate seems much too optimistic. I don't want to read too much into a brief mention in the newspaper, though.)

An overview of tools for constraint-satisfaction problems (CSP) has been compiled by the AI Applications Institute (UEdinburgh). http://www.aiai.ed.ac.uk/~timd/constraints /csptools.html. [Tim Duncan (timd@aisb.ed.ac.uk), comp.constraints, 8/10/94. David Joslin.]

leanTAP is a small, complete, and sound theorem prover for classical first-order logic based on free-variable semantic tableaux written in Prolog. The minimal Prolog source code is only 360 bytes:

prove((E,F),A,B,C,D) :- !,prove(E,[F|A],B,C,D). prove((E;F),A,B,C,D) :- !,prove(E,A,B,C,D),prove(F,A,B,C,D). prove(all(I,J),A,B,C,D) :- !, +length(C,D),copy_term((I,J,C),(G,F,C)), append(A,[all(I,J)],E),prove(F,E,B,[G|C],D). prove(A,_,[C|D],_,_) :- ((A= -(B);-(A)=B) -> (unify(B,C);prove(A,[],D,_,_))). prove(A,[E|F],B,C,D) :- prove(E,F,[A|B],C,D).

leanTAP's standard distribution comes with a rudimentary user interface and some test examples. To join the mailing list, contact Bernhard Beckert (beckert@ira.uka.de) or posegga@ira.uka.de. Other info is on http://i12www.ira.uka.de/~posegga/leantap/leantap.html, or FTP all the Lean* files from pub/posegga/ on sonja.ira.uka.de (129.13.31.2). [sci.logic, 8/4/94.] (Does 360 bytes put it in the same class as Kepler's and Newton's laws, Einstein's Lorentz contraction, Maxwell's equations, ...?)

"An alternative view [to logic-based expert systems] might be called the 'proverbs model.' By this view, natural selection, over the course of geologic eons, has introduced millions of genetically-programmed *details* into the human nervous system, anticipating particular sorts of survival-challenges that may arise, and venturing effective ways of reacting to them. Compared to Euclid's geometry, there will be many more 'axioms' and relatively fewer 'theorems', so the *logic* will probably be comparatively trivial -- almost all the expertise will be in the data structures." -- Jorn Barger (jorn@mcs.com), comp.ai.imho FAQ, 7/27/94.

The 7/94 CACM is a special issue on intelligent agents, with 16 articles from DAI and other perspectives. [Stefan Baldi (baldi@sparc.ebs.de), DAI-List, 7/22/94.]

Proceedings for the 13th Int. Distributed AI Workshop (July 27-29) are available for about $25 from Mark Klein (mklein@boeing.com), (206) 865-3412, (206) 865-2965 Fax. 18 papers. [comp.ai, 8/9/94. David Joslin.]