Will supercomputing intelligences outsmart human-level intelligence?
March 18, 2011
Natasha Vita-More: “The very same technology that proposes to build superintelligences could also dramatically enhance human cognition…. The coincidental and subsequent developments of inventive projects arrived at through digital media, virtuality, and immersivity have furthered the scope of human experiential enhancement as artificial intelligence technologies are fostering arguably viable developments. This overlap of computational and physical forms an evolutionary crossing point. Could the human become a super AI?”
Doug Lenat: “The bottleneck is software, not hardware…. The limitations of WATSON, Google, SIRI, etc. are ones of breadth of inference, not quantitative performance metrics…. For many years now, dozens of us have been building CYC, a repository for the common sense knowledge and general inferencing strategies…. We are on the verge of a sort of Singularity in the building of CYC: it now knows enough, and can infer enough, to carry on interactive dialogues in English, opening up the possibility of having millions of people helping it to cross that finish line in 2012.”
Michael Vassar: My work is focused on the exploration and integration of the visions of the Technological Singularity developed by Vernor Vinge, Raymond Kurzweil and Eliezer Yudkowsky…. I associate these visions of the Singularity with three stages in the likely evolution of information processing, the combined impact of which will most likely make the 22nd century resemble the 20th less closely than the 20th century resembles the Cambrian. The earliest, Vingean stage is of particular importance, because the development of superhuman collective intelligences is likely to mark the end of the period during which deliberate human decision making can enable human values to directly influence humanity’s future.”