Seán Ó hÉigeartaigh was recently interviewed by RealClearTechnology on the future prospects and risks of artificial intelligence, alongside the Machine Intelligence Research Institute‘s Luke Muehlhauser and NYU’s Professor Gary Marcus:
“Most of the AI applications we are familiar with represent “narrow” intelligence,” O’Heigeartaigh said. “Like Deep Blue and chess, they are hard-coded to do a specific task, and may be super-human at that, but cannot adapt to new tasks without significant rewriting — i.e. the extent to which they can learn is very limited. One goal of the field of AI, for a long time, has been to develop more ‘general intelligence’ — algorithms that can learn about their environments and adapt to a variety of different tasks.”
