Vernor Vinge Does Not Yet Admit There Are Laws of Physics
This video at the end of this post (rather choppy in its editing and distracted in its focus, unfortunately), has been making the rounds under the headline "Moore's Law Won't Create the Singularity on its Own." (Eg, here.) Actually, Vernor Vinge is continuing to assert his expectation of a "Singularity" style event or process in "the near historical future," and attributes fully half of the rationale to Moore's law. So, misleding headline.

For readers who aren't familiar with the concept (are there any left?) the "singularity" describes a mechanism of intelligence generationally increasing itself, using increased intelligence to also decrease cycle time, leading to a spiralling increase in intelligence which shortly becomes theoretically infinite. The term references the gravitational well of a black hole, also called a singularity.

The basic theory (originally articulated in a clever little paper in 1993, which is available here) is simple. If you're smart enough to increase your own intelligence, whether using computational support or biological enhancement, then you can use the increased intelligence to further increase intelligence. The feedback loop reinforces itself to infinity.

In fiction, this is usually imagined as a combination of computer software and hardware that iteratively redesigns itself, each version reaching for godlike intelligence. (Ken McLeod and Charlie Stross have popularized (though they themselves did not coin) the phrase "The Rapture of the Nerds" to describe this singularity).

While some kind of intelligence, or sentience, or sapience that transcends our own is practically inevitable, the actual spiralling off to infinity has an inherent problem: physics.

Material science continues to make amazing advances in miniaturization which inherently leads to faster, more powerful processors, higher density data storage, and a general infrastructure that would lend itself to wild advances in "intelligence." However, these advances are limited. There will come a point when quantum physics prevents traditional circuits from getting any smaller (we're pretty close to this boundary now). There will come a point when laws of thermodynamics prevent processors from getting any faster. Science fiction writers imagine quantum computing as "the answer" but for a variety of reasons, that's even less likely to actually solve issues of scale or performance for anything other than a few very specific (although world-threatening in their own right) tasks.

I suspect there are other limitations as well. It's a well understood process of human development that actual human intelligence only begins to form as the brain *reduces* its complexity and interconnectedness. Simply throwing more and faster hardware at the problem actually makes it worse. (Perhaps like throwing more software developers at a slipping project.)

Do I think that we are the theoretical maximum of possible intelligence? Of course not, that would be absurd.

Do I think that there is a prospect of increasing our own human intelligence dramatically or creating something that is more intelligent than we are? Absolutely.

But I resist the term Singularity and the imagined consequences because I firmly believe that the problem is harder than imagined, and there are constraints of physics and logical optimization of networks that will affect a hard cap on what is possible.