Robotic Nation Evidence

I gave a talk to a group of robotics folks on Monday night, and one of the concerns that came up in the Q&A session afterwards was that Moore's Law will somehow "run out of steam." Two of the problems that people brought up included, a) the ever-shrinking size of transistors cannot continue forever, and b) increasing power consumption cannot continue forever. One point made is that current supercomputers (the kind with 10,000 Pentium chips running in parallel) can consume 10 to 20 megawatts of power. Certainly a robot cannot consume 20 megawatts.

I understand both points, but I think they are both irrelevant. First, we KNOW it is possible to produce a high-performance, low power CPU. Each one of us has a brain that performs something on the order of one quadrillion operations per second, yet it consumes only 20 watts.

Second, Scientists and engineers make discoveries all the time, and things simply get faster and faster. 20 years ago a Cray computer ran so hot that the entire computer (as big as a refrigerator) was immersed in liquid FC-77 to extract the prodigious amounts of heat it created. Today you can get that same power in a little desktop computer cooled with a small fan. That's normal progress, and there's nothing going to stop that sort of progress.

Here are two articles that show current trends in making computers faster and more efficient:

Sun chips away at wireless chip connections

From the article:

It will take a lot of work, but Sun Microsystems says it is making headway on a technology that will allow chips to communicate without circuit boards or wires.

The technology, called "proximity communication," aims to let one chip transmit signals directly to another next to it, instead of through the tangle of pins, wires and circuit boards employed today. If successful, the technique could greatly alter many aspects of computer design.

Performance, for instance, could greatly escalate because the speed of transferring data among chips and the number of channels for the transfers would increase. Energy consumption could also decline. Just as important, overall costs could fall, because defective chips could be removed like Scrabble tiles.

Also:

The technique could also allow designers to remove the cache--the large pool of memory currently found on the processor--and put it on a separate chip. Caches were integrated onto processors to amplify bandwidth. Adding cache, however, bumps up manufacturing costs, as it greatly increases the number of transistors. With the bandwidth constraint gone, caches could once again be made independent without it having an impact on performance.

If you take out the cache -- representing millions of transistors -- you can use those transistors for something else. See this page for some thoughts.

This article talks about a whole new paradigm for computing:

Nanotech leads way to quantum computing

From the article:

Plastic chips and quantum computing could be among the new ways of keeping up with Moore's Law in the future, according to a new study on nanotechnology, the science of manipulating matter on a molecular level.

One point I made in my talk and in my article discussed how quickly airplanes advanced between 1903 and 1954:

Imagine that you could travel back in time to the year 1900. Imagine that you stand on a soap box on a city street corner in 1900 and you say to the gathering crowd, "By 1955, people will be flying at supersonic speeds in sleek aircraft and traveling coast to coast in just a few hours." In 1900, it would have been insane to suggest that. In 1900, airplanes did not even exist. Orville and Wilbur did not make the first flight until 1903. The Model T Ford did not appear until 1909.

Yet, by 1947, Chuck Yeager flew the X1 at supersonic speeds. In 1954, the B-52 bomber made its maiden flight. It took only 51 years to go from a rickety wooden airplane flying at 10 MPH, to a gigantic aluminum jet-powered Stratofortress carrying 70,000 pounds of bombs halfway around the world at 550 MPH.

That is the kind of progress we will continue to see in computing power over the next 50 years. We will see progress in transitor size and power consumption, packaging, etc. We will also see completely new paradigms arise. What these developments mean is a dramatic increase in robotic intelligence over the next several decades, along with dramatic changes in the world economy.

See also:

• Moore's Law continues
• Moore's law and hard disks
• Future processors
• The effects of Moore's law
• Robotic AI using Neural Nets