The collection of human technologies exhibits a form open-ended change and evolution, and this makes it difficult to detect technological innovations, especially if they are unanticipated.
One illustration of this problem is the on-going ad-hoc revisions to the classes of technology in the United States Patent Classification. A new approach to this problem combines methods from machine learning and natural language processing to mine millions of public documents describing patented inventions. For example, one can construct a high-dimensional technology feature space by training a doc2vec model on patent records, and then embed those patents in that technology space. The remarkable property of the resulting technology space is that the proximity of two patent records in the space reflects the similarity of the statistical features of the word combinations in the two documents. This technology space opens the door to a natural, objective, and operational method for classifying technology: simply identify clusters in the patents in technology space. Analysis of millions of US patent records shows that these clusters provide natural descriptions of technology classes, and they are especially useful for detecting innovations that were unanticipated. These methods can be generalised to detect unanticipated innovations in other on-going streams of digital data.
Please register on Eventbrite or contact email@example.com for more information.
Prof Mark Bedau
Professor of Philosophy and Humanities, Reed College
Mark A. Bedau is Professor of Philosophy and Humanities at Reed College, and for the past twenty years he has been Editor-in-Chief of the journal Artificial Life. He was co-founder and Director of the Initiative for Science, Society, and Policy (Denmark), Visiting Professor at the European School of Molecular Medicine (Milan), and co-founder of the European Center for Living Technology (Italy). His research interests include emergence, evolution and adaptation, the nature of life and intelligence, the evolution of technology, applications of machine learning and NLP methods for understanding and designing complex systems, and the social and ethical implications of new and emerging technologies. He has published extensively on these topics, and co-edited four books: Emergence: Contemporary readings in philosophy and science (MIT Press, 2008), Protocells: Bridging nonliving and living matter (MIT Press, 2009), The ethics of protocells: Moral and social implications of creating life in the laboratory (MIT Press, 2009), and The nature of life: Classical and contemporary perspectives from philosophy and science (Cambridge University Press, 2010).