Claes Strannegård

Professor, Data Science division, and vice Head of Department responsible for utilization and collaboration at the Department of Computer Science and Engineering.

My main research interests are cognitive science and general intelligence. I like the multi-disciplinary character of cognitive science and have done research in mathematical logic, cognitive psychology, and general AI. For several years my research goal has been to develop systems with basic forms of general intelligence. I co-founded five research spinoff companies from Chalmers and the University of Gothenburg and worked with those companies for many years.
  • I have taught courses in logic, AI, basic programming, basic linguistics, and innovation & entrepreneurship. 
  • I designed, coordinated and taught at the Master's program Intelligent Systems Design at Chalmers in 2006-2010. 
  • I have supervised thesis projects at the PhD, MSc, and BSc levels. Please contact me if you want to discuss a thesis project with me.
I have worked with computer scientists, psychologists, neuroscientists, logicians, linguists, and mathematicians in the following fields:

  • Mathematical logic. During my PhD at the department of Philosophy at the University of Gothenburg and my post-doc at Utrecht university I did research in the metamathematics of arithmetic. My main result in this field is a completeness theorem that generalizes and strengthens several theorems in provability logic, e.g. Gödel's incompleteness theorems, Berarducci’s completeness theorems, and Shavrukov’s theorem on embeddability of Magari algebras.

  • Cognitive psychology. I led a research project on computational models of human reasoning that was funded by the Swedish Research Council (2013-2015), where we obtained the following results: 
    • Deductive reasoning. We conducted psychological experiments and collected accuracy and latency data on logical reasoning tasks. Then we constructed proof systems for logical reasoning with bounded cognitive resources and showed that minimal proof length correlated well with both accuracy and latency. 
    • Inductive reasoning. We used experimental data from IQ tests with number sequences and progressive matrix tasks. Then we constructed cognitive models based on Kolmogorov complexity with bounded cognitive resources. These models turned out to match or surpass human performance. 
    • Combined reasoning. We developed a cognitive architecture for symbolic processing that was able to learn basic arithmetic, logic and grammar from scratch using streams of examples only.
    Our research group won the best paper awards at the Artificial General Intelligence Society conferences in Beijing 2013 and Quebec City 2014.

  • General AI. Currently I am working on a control algorithm for a generic artificial animal, whose ability to learn, make decisions, and survive in different environments could match, e.g. a fruit fly. We approach the problem with a mix of logic (dynamic ontologies consisting of formulas of modal logic) and reinforcement learning. This research was funded by the Swedish Research Council (2014-2017) and the Torsten Söderberg Foundation (2018).
After my post-doc I received an innovation prize from the Swedish Academy of Engineering Sciences (IVA) and a grant from the Swedish Business Development Agency (Nutek) for developing an idea about logical analysis of industrial systems. This eventually led me away from academia and into the world of start-ups. Over a 15-year period I worked with five IT-companies as a co-founder and CEO/Director. The most successful ones were these:
  • Safelogic. This company focused on logical analysis of integrated circuits by means of automatic theorem proving. During my time as CEO the company grew from 1 to 20 employees. Safelogic was merged with the sales and marketing company Jasper Design Automation of Mountainview, USA. It developed into a leading company in the electronic design automation industry with several tech-giants as customers. The merged company was acquired by Cadence in 2014.
  • Optisort. This company focused on classification of physical objects by means of neural networks. The company -now called Refind- is a market leader in USA and Great Britain in the field of automatic battery sorting into environmental categories. Refind provided the technology for the fish sorting project FishFace that won the Google Impact Challenge Australia in 2016.

Published: Tue 09 May 2017. Modified: Wed 21 Feb 2018