"Imagine being present at the birth of a new industry. It is an industry based on groundbreaking new technologies, wherein a handful of well-established corporations sell highly specialized devices for business use and a fast-growing number of start-up companies produce innovative toys, gadgets for hobbyists and other interesting niche products. But it is also a highly fragmented industry with few common standards or platforms. Projects are complex, progress is slow, and practical applications are relatively rare. In fact, for all the excitement and promise, no one can say with any certainty when--or even if--this industry will achieve critical mass. If it does, though, it may well change the world."I wrote the letter below to Scientific American, about the article.
From Aaron Sloman Mon Jan 1 02:50:06 GMT 2007 To: firstname.lastname@example.org Subject: Bill Gates on Robotics -- the need for better requirements analysis The real reason for lack of progress in robotics. ------------------------------------------------- I think Gates is correct about many things in his article, especially his opening paragraph. However, as someone who has been working on this topic for over 30 years (including writing a book published in 1978: The Computer Revolution in Philosophy, now online here) I believe that he has made a mistake that is also made by most people who work in robotics. The mistake is believing that the *main* obstacle to progress is lack of technology, or lack of solutions to problems, whereas in fact the key problem is a lack of understanding of what *the problems* are and how the many problems differ and are related. Thus most people are attempting to design systems without analysing requirements thoroughly. One example is the need to characterise the many functions of vision: too many people think of vision as not much more than recognition of objects, ignoring its role in control of intricate actions, in perceiving and understanding processes, in seeing possibilities for future processes, and in understanding causal relationships, e.g. in an old-fashioned clock. Those capabilities are not only involved in acquiring physical competences, but also form part of what makes it possible to become a mathematician. The ability to learn and understand school geometry is not normally noticed as a requirement for a domestic robot. We can deepen our understanding of requirements by examining in far more detail than roboticists normally do the actual variety of competences displayed by humans and other animals (e.g. children learning to play with construction kits before they can talk, nest-building birds, hunting mammals). However, simply observing does not reveal what the competences are, nor what the problems are that need to be solved. Discerning what is being achieved, even in apparently simple actions, often requires deep interdisciplinary analysis based in part on experience of doing AI and philosophy. I've tried to illustrate that in connection with the role of ontology extension in development, in  (among other papers and presentations on my web site).  http://www.cs.bham.ac.uk/research/projects/cogaff/crp/ Aaron Sloman. The Computer Revolution in Philosophy: Philosophy, science and models of mind. (1978).  http://www.cs.bham.ac.uk/research/projects/cosy/papers/#pr0604 Ontology extension' in evolution and in development, in animals and machines. (PDF presentation, based in part on work with ornithologist Jackie Chappell.) Yours sincerely Aaron Sloman Web: http://www.cs.bham.ac.uk/~axs/ Honorary Professor of Artificial Intelligence and Cognitive Science University of Birmingham, UK
As someone who has been working on requirements for robots for more than 30 years, I believe that Gates has made a mistake (as have most in the field) in thinking that the main obstacle to progress is a lack of solutions to problems. In truth, the key obstacle is a lack of understanding of what the problems are, how they differ and how they are related. Most robotics engineers are attempting to design systems without thoroughly analyzing the requirements necessary to make them functional. One example is the need to characterize the many functions of vision. Too many people think of vision as simply the recognition of objects and ignore its role in controlling intricate actions, perceiving and understanding processes, seeing possibilities for future processes, and comprehending causal relations. We can deepen our understanding of requirements by examining in far more detail than roboticists typically do the actual variety of competences displayed by humans and other animals. But simple observation will not reveal the nature of such competences nor what problems need to be solved for a robot to achieve them. Discerning what is being achieved, even in apparently simple actions, often requires deep interdisciplinary analysis based in part on experience in artificial intelligence and philosophy.