logo   Wittgenstein Award Laureate

0
nix nix nix Jürgen Knoblich Portrait Gerhard Widmer Portrait Markus Arndt Portrait Christian Krattenthaler Portrait Rudolf Zechner Portrait Joerg Schmiedmayer Portrait Barry Dickson Portrait Rudolf Grimm Portrait Walter Pohl Portrait Renee Schroeder Portrait Ferenc Krausz Portrait Heribert Hirt Portrait Meinrad Busslinger Portrait Peter Markowich Portrait Andre Gingrich Portrait Kim Ashley Nasmyth Portrait Peter Zoller Portrait Walter Schachermayer Portrait Georg Gottlob Portrait Marjori + Antonius Matzke Portrait Erich Gornik Portrait Erwin Wagner Portrait Ruth Wodak Portrait

Wittgenstein Award Laureate 1998 Univ. Prof. Dr. Georg Gottlob

Information systems and artificial intelligence Technical University of Vienna

Georg Gottlob ext

Institut für Informationssysteme ext

Georg Gottlob Research Interests ext

mail gottlob@dbai.tuwien.ac.at

ADAPTING LOGIC TO REALITY

The mathematician and computer specialist Georg Gottlob is investigating computational complexity, the optimization of decisionmaking and the avoidance of serious logical errors. An example is the representation of knowledge. ”How,” asks Gottlob, ”can you represent knowledge formally, logically? You need to include knowledge processing, because I want to do something. I therefore need algorithms that can represent this process.”

Classical logical statements (”If not B then …” and so on) are too rigid to describe knowledge of acting bodies (man, computers, robots). They are ”monotonous:” the facts and rules are pre-defined and their validity is not affected when new facts or rules become known. ”In practice things are different so we need a new logical formalism.”

Gottlob has been interested in this issue since he and colleagues encountered it in Stanford in the middle of the 1980s, when the area was new. From 1986 he has been teaching and researching database theory at the Vienna University of Technology. He is not dealing with ”normal” databases but with those that store logical rules and decision diagrams so that these can be interrogated. In this way, proofs can be checked automatically. ”But the required algorithms are so complicated that a complete check is no longer possible.” The task is thus to find a compromise between practicability and formal completeness. The level of accuracy required depends on whether you are trying to simulate the attempts of a fly to escape from a room (”phases of local optimization and random phases”) or the industrial steel production process of Böhler in Kapfenberg. Or whether you are testing the software used for missile control or monitoring of nuclear power plants, where errors can be critical.

The 1998 Wittgenstein Prize has given Gottlob the opportunity to use money in a variety of ways and he is very pleased with the flexibility. ”We have hired young scientists, awarded fellowships to students – including four from the University of Bratislava –, bought computers, travelled abroad and invited top-quality researchers to visit us, thereby paving the way for future collaboration.” A stay as Visiting Professor at Berkeley in the summer of 1999 was also facilitated by the prize money.

As an example of his many published results he cites his work on ”Constraint Satisfaction Problems.” Grossly simplified, such problems are said to be – subject to certain limitations – solvable if each variable can be assigned a value without breaking any of the constraints. Applications range from crossword puzzles to tasks in industrial planning.
Gottlob’s team has managed to show that an approach they developed is more efficient than previous methods. ”The problems can be parallelized to a large degree and can thus be rapidly solved by parallel processors.” The presentation of the results was followed in January 2001 by a research contract from DaimlerChrysler to develop programs to solve efficiently certain configurational problems relevant to the car industry.

Georg Gottlob