I don't know that a superintelligence will emerge from the web, per se, but I think it is becoming a sufficiently large body of textual information that an appr

I don't know that a superintelligence will emerge from the web, per se, but I think it is becoming a sufficiently large body of textual information that an appropriate AI program could use to bootstrap itself to amazing levels of intelligence. AI programs have a problem that they have to infer all their patterns just from text, which is fairly low-bandwidth. Humans (and robots) get much more bandwidth through vision, hearing, touch, and body feedback. This might not seem important, but it provides context around every piece of information that comes in, making it much easier to make sense of things and giving richer patterns to try to match against. Without that context, an AI would need to multiply the amount of text needed by 1M (1B? 1T? X?) to train itself the way a child does.