Most Singularitarians feel that a Singularity should be technologically possible sometime between 2010 and 2030. Oxford University's Nick Bostrom agrees. We can probably push forward that date if we put enough effort into it. A Singularity is by no means inevitable - nuclear war, a nanotechnological arms race, and human-indifferent superintelligence are all serious risks which could prevent the creation of benevolent superintelligence, and possibly wipe out the human race as a whole.

Why so early? Creating superintelligence from scratch requires two things - hardware and software. Hardware with lots of computational power and memory bandwidth, and software capable of reaching human-equivalent standards in the domain of general intelligence. The better your hardware, the easier it is to write an intelligent software program.

As futurist and inventor Ray Kurzweil has shown, the availability of cheap computing power has been increasing exponentially since 1965, without any signs of slowing down. The raw computing power of some of today's supercomputers is already rivalling the computing power of the human brain. The human brain performs an estimated
10^17 operations per second (ops/sec), while IBM's Deep Blue supercomputer performs at roughly 10^15 ops/sec.

Eric Drexler, founding father of the field of nanotechnology, has shown how we will soon be able to create machines that build products atom by atom - including computers that perform 10^19 ops/sec, fit in your shirt pocket, and consume power equivalent to a light bulb. Such cheap computing power would literally throw the doors open to
advanced Artificial Intelligence, extremely high-resolution brain-scanning methods, and other technological marvels.

What about all those promises made by AI researchers in the 60s and 70s? Firstly, these researchers couldn't have created AI even if they had the correct theory - the amount of computing power they had available was comparable to that of an ant brain. Secondly, the field of cognitive science - the study of the brain and intelligence - was extremely undeveloped at that time. Thirdly, early failures are not absolute indicators of the implausibility of a given technology.
Lay theorists tend to underestimate the role of computing power as an AI enabler due to common biases.

next->