The Singularity Summit at Stanford took place yesterday. Here's a quick rundown of the event and some impressions from my notes. The singularity is described as that point in the future (2029 according to Raymond Kurzweil) where artificial intelligence exceeds human intelligence. The term is borrowed, somewhat imprecisely, from physics and math. Gravitational pull reaches a singularity at a black hole, for example.
Ray Kurzweil opened with his keynote, laying out his arguments for his "Law of Accelerating Returns." He says you can fit an exponential curve to virtually any history of events in information technology. Detractors have said that it's not an exponential curve, it's a logistic S-curve. They're wrong, says Kurzweil, because he's only talking about IT -- this isn't rabbits and foxes. "Yes there are limits, but they're not very limiting." Kurzweil believes that the mind is computational, and that we can "reverse engineer the brain." Anticipating Bill McKibben's argument that we might come to a point of technology where we say "enough" because we've got what we need to live happily, Kurzweil says it won't be enough until we overcome all diseases, and, by the way, that includes death.
Douglas Hofstadter spoke next. Hofstadter has been interested in this topic since the late 90's and organized this symposium's predecessor, The Spiritual Robots Symposium held at Stanford in 2000. The inspiration were two recent books, Kurzweil's The Age of Spiritual Machines, and Hans Moravec's Robot: Mere Machine To Trancend Mind. Hofstadter hoped then to initiate serious scientific and philosophical consideration of these issues, and he feels that the 2000 symposium was a failure, as participants didn't engage the issues, but mostly just presented their own work. He made a plea for the same today: "The scientific world is not taking this seriously." He wants to see "serious skepticism." He appreciates and studies the work of Kurzweil and others, but has trouble taking them seriously, in part because they're "marred by too much science fiction." He reaches a point where he has to ask, "how secure can I be of the sanity of this person?" In Kurzweil's case, there's the startling fact that Kurzweil does this work because he wants to live forever. Hofstadter didn't name other names, but Vernor Vinge's original essay, and some of the other writings proposed as being seminal for this field, clearly lack scientific or philosophical rigor. Not to dismiss sci-fi altogether, but singularity enthusiasts seem largely ignorant of other relevant fields -- those within science, which I think Hofstadter was
getting at, and those outside of science, in particular philosophy of
science and technology, history of technology, and science and
technology studies. Hofstadter made the technical point (as have others) that exponential curves don't have a knee, as Kurzweil claims. And they don't reach a singularity either! An exponential curve has the same shape regardless of scale. Indeed, I would think that one hundred years ago, or at any time for that matter, people could have plotted various markers of progress on a graph, saw an exponential curve and said, whoo boy, it's getting really steep now! Everything's going to change if it gets much higher! Perhaps the singularity, in this sense, is just an illusion that's there for anyone who looks, whenever they look.
Back to Hofstadter's plea for serious scientific discussion of the questions at hand. Did the new summit answer his request? I don't think so. The next several speakers were specialists in various tangential fields and most made little apparent effort to orient their talk toward this audience and topic.
Nick Bostrom of the Future of Humanity Institute presented on his theories of risk, in particular what he calls "existential risk." Sebastian Thrun of the Stanford Artificial Intelligence Lab gave an entertaining presentation about the DARPA autonomous vehicle challenge, which his team won this year. Cory Doctorow of BoingBoing evangelized against digital rights management. Eric Drexler of Nanorex evangelized on nanotechnology. "Bogus criticism has fallen out of fashion," he said, then disappeared for the rest of the day. I guess Dr. Drexler doesn't take questions. Max More of the Extropy Institute talked about something or other, futuristically. Christine Peterson of the Foresight Nanotechnology Institute told us more about nanotech.
Things got back on track somewhat with an on-topic talk by John Smart of the Acceleration Studies Foundation, who started with a joke about the preceding speakers' bad powerpoint: "I have 90 slides" for 20 minutes. Only it wasn't a joke. He raced through a survey of a history of acceleration studies (somewhat like this paper of his). It's useful work, presumably, been there's not enough depth, and I think his scope is too narrow. Smart often states that it takes "courage" to pursue singularity or acceleration studies. To me, this doesn't help his cause; it just highlights the fact that many of these folks are believers looking for pseudo-scientific support.
Eliezer Yudkowsky of The Singularity Institute spoke about his "Friendly AI" research and the coming "intelligence explosion" caused by recursive AI -- AI that can reprogram itself. He makes the point that we're not smart enough to predict what humans will do, so how can we hope to predict what smarter AIs will do? He has no solution, only that it's "very very very difficult." And he's working on it, I guess. Yudkowsky is undoubtedly very talented in his field, but again the breadth of his approach seems weak. There's much ad hoc amateur philosophizing, as opposed to serious engagement with professional philosophy. Some of Yudkowsky's statements also typify the scientistic and patronizing attitude of many singularity enthusiasts. They're in charge now, and everyone needs to get on board. Far be it from them to do their homework. One of Yudkowsky's concluding slides said "Someday the human species has to grow up. Why not sooner rather than later?" But grow up according to whom? Many wise people throughout history -- philosophers, scientists, religious thinkers -- have studied the problem of how do we grow up and transcend our human limits and achieve meaning. Some would say they've succeeded, or at least advanced the study quite a long way. Why should we take seriously a bunch of blinkered computer scientists and physicists who dismiss this history altogether?
The final speaker was Bill McKibben, who spoke eloquently about human meaning, hyper-individualism vs. ecology and community, and other large issues regarding the singularity and posthumanism. What are the practical applications of this work, he asked. I won't go into summarizing his points as they're largely an extension of his book, Enough, and I'm in pretty much full agreement with them. His critique wouldn't satisfy Hofstadter, I suspect, as it comes from the ecological and humanities side of things. I don't think that detracts from McKibben's arguments; it's just that there's room for criticism from many other directions as well.
Next up was a bit of panel discussion and then questions from the audience. I skipped out of these because of a splitting headache and a desire to visit the Stanford bookstore. Audio files of the event are supposed to appear on the site within a few days, and I plan to listen to the remainder in case I missed any interesting Q&A.