Abstract: Information and Computer Technology is all about bottlenecks and shortcuts. For every problem we solve, we create two new ones. Advances in software technology have focussed on speed of software development, but this has come at the cost of huge inefficiency, giving back the gains of Moore's Law - which itself is changing in nature as sequential processors give way to multicore systems, and the problems of distributed processing and parallelism become more critical. What has this got to with Artificial Intelligence or Cognitive Robotics? The message here is that solving problems of intelligence requires cooperative advances in hardware and software, not just robotic or other embedding hardware, but advances in computer hardware and software technology that are targeted to the kind of problem we deal with in Complex Intelligent Systems or Cognitive Robotics.

   Research in Cognitive Robotics can be traced back to Turing in the 1940s, who in proposing his test of intelligence said that he could achieve this by "buying the best sensors money could buy" and making the computer learn, himself developing one of the earliest self-organizing models. The key here is that learning language is as much about learning about the world, and how we interact successfully with the world, as about learning linguistic knowledge. Piaget in the 1930s and Block in the 1960s emphasized the functional aspects of this, that the things we can do or use become building blocks that can be used for more complex purposes, including as building blocks of language - the sensorymotor world is where we ground our nouns and verbs. Powers in the 1980s emphasized this need to learn a sensorymotor ontology, which Harnad in the 1990s characterized as the symbol grounding problem, while Cognitive Linguistics emerged as we saw that the nature of language reflected the nature of the world. The message here is that our brain and our learning is targeted at dealing with and surviving in our environment, and that interpretation of our social, cultural and linguistic environment is just part of this.

  This talk will present 30 years of unsupervised and multimodal learning research on discovering grammatical and phonological information from text and speech, about the importance of affect, emotions and drives, about the learning of non-linguistic information, body language and facial expression, and the grounding of ontology in both real and simulated environments. However the focus is not just about better technology, but about understanding the human side of the equation, and building human-machine systems that are optimized by understanding both sides of the problem. The message here is that we need to use biological methods to explore predictions of models, and as a result we use EEG to understand and optimize cognitive load and overal workload, and to develop new dimensions for computer interfaces, leading to physical robots and talking heads, and combinations of the two - in applications ranging from defence to education to health.

 

Bio: David Powers is Professor of Cognitive and Computer Science at Flinders University, in South Australia.  He has qualifications across mathematics, computer science, psychology and linguistics, including certificates teaching English as a second language, and technical analysis of financial markets. His PhD from the School of Electrical Engineering at the University of NSW (1989) focussed on how to get a computer to learn language like a computer does - and his diverse qualifications are all steps along this path. He was formerly Editor-in-Chief of ACM's SIGART Bulletin and helped launch their Intelligence Magazine; he was founding President of ACL's SIGNLL and in that role was responsible for launching CoNLL. He is currently Editor-in-Chief of the new SpringerOpen Journal Computational Cognitive Science and the associated book series Cognitive Science and Technology - these emphasize that we need to simultaneously approach the hard problems of Learning and Intelligence in ways that involve better understanding both the human and the system, and the journal is designed to support researchers in learning how to speak the languages of the different disciplines that bear on this.