Home > Hardware

Computerworld -

Inventor, writer and futurist Ray Kurzweil

Inventor, writer and futurist Ray Kurzweil has been a pioneer in speech and character recognition, reading technology, music synthesis, virtual reality and artificial intelligence. He has founded nine businesses in those fields, including Kurzweil Technologies Inc. in Cambridge, Mass., and he's won numerous awards, including the National Medal of Technology.
In his recent book, The Singularity Is Near: When Humans Transcend Biology (Viking Adult, 2005), Kurzweil, 57, predicts that ultimately, human intelligence and computer intelligence will fuse and become indistinguishable. He recently told Computerworld how and when that might come about.
Your idea to reverse-engineer the human brain seems pretty far out. Until recently, we haven't had the tools to scan the brain with sufficient resolution. But there are five or six new scanning technologies. For the first time, we can see the brain creating our thoughts.
The amount of data we are gathering about the brain is doubling every year. As we get the data from particular regions, we can rather rapidly create detailed mathematical models of them. It's a conservative expectation that we will have a very accurate detailed simulation of all the regions of the brain by the late 2020s.
Ten quadrillion [10¹⁶] calculations per second is sufficient to emulate all the regions of the brain. Japan just announced two supercomputers that will achieve that by 2010.
The question arises, Are we intelligent enough to understand our own intelligence? Maybe that's a feature of complex systems -- they can't be so complex as to understand themselves. But it turns out that's not the case.
But why re-create the brain in software when we already have it in wetware? It's going to be very powerful, because we'll be able to combine what currently are advantages of human intelligence, particularly our pattern recognition, with ways in which machines are already demonstrably superior.
What's the future of the computer itself? Once we get past Moore's Law, we'll use 3-D molecular computing. [In the late 2040s], one cubic inch of nanotube circuitry will be 100 million times more powerful than the human brain. On the software side, machines [in the 2030s] will be able to access their own source code and improve it via an ever-accelerating, iterative design cycle. So ultimately, these systems will be vastly more intelligent than humans and will combine the advantages of biological and nonbiological intelligence. I don't see this as an alien invasion of intelligent machines; this is emerging from within our civilization.
Well before that, computation will be a worldwide