> However, the idea is, once you have a general intelligence running on a computer, you could easily scale it to create a super human level intelligence. I que

> However, the idea is, once you have a general intelligence running on a computer, you could easily scale it to create a super human level intelligence. I question that. I expect that "intelligence" (however defined) will not be a linear function of the scale of the system. It will be considerably less than linear. If it's log-log scaling, say, and you already had to go to a large scale just to get human-equivalent intelligence, how easy is it to scale to superhuman levels of intelligence? Don't assume that scaling will be easy. It very well might not be.