Peek at 1998 Moravec book, Chapter 3

3: Power and Presence

The arrival of intelligent machines is paced by the growth of computing power. A comparison between edge and motion detectors in the human retina with similarly functioning computer vision programs suggests that the retina does the job of 1,000 MIPS (million of instructions per second) of computing. The whole brain is 100,000 times larger than the retina, so is worth perhaps 100 million MIPS of efficient computation. The following diagram rates other entities.

Power and Capacity

Power rating of natural and artificial thinkers

MIPS and Megabytes: Entities rated by the computational power and memory of the smallest general-purpose computer needed to mimic their behavior. Note the scale is logarithmic on both axes: each vertical division represents a thousandfold increase in processing power, and each horizontal division a thousandfold increase in memory size. General-purpose computers (marked by an *) can imitate other entities at their location in the diagram, but the more specialized thinkers cannot. A 100 million MIPS computer may be programmed not only to think like a human, but also to imitate any other similarly-sized computer. But humans cannot imitate 100 million MIPS computers--our general-purpose calculation ability is below a millionth of a MIPS. Beep Blue's special-purpose chips calculate chess moves like a 3 million MIPS computer, but its general-purpose component can do only a thousand MIPS. Most of the non-computer entities in the diagram can't function in a general-purpose way at all. Generality is an almost magical property, but it has costs. A general-purpose machine may use ten times the resources as one specialized for a task. But if the task should change, as it usually does in research, the general machine can simply be reprogrammed, while the specialized machine must be replaced.

Information handling capacity in computers has been growing about ten million times faster than it did in nervous systems during our evolution. The power doubled every two years in the 1950s, 1960s and 1970s, doubled every 18 months in the 1980s, and is now doubling each year.

Computer power: 1900-1997

Power/cost of 150 computers from 1900 to 1997, rising 1000x every 20, now 10, years

Numerical data for the power curve
Faster than exponential: In three decades the doubling time has fallen from two years to one year.

Alas, for several decades the computing power found in advanced Artificial Intelligence and Robotics systems has been stuck at insect brainpower of 1 MIPS. While computer power per dollar fell rapidly during this period, the money available fell just as fast. The earliest days of AI, in the mid 1960s, were fuelled by lavish post-Sputnik defense funding, which gave access to $10,000,000 supercomputers of the time. In the post Vietnam war days of the 1970s, funding declined and only $1,000,000 machines were available. By the early 1980s, AI research had to settle for $100,000 minicomputers. In the late 1980s, the available machines were $10,000 workstations. By the 1990s, much work was done on personal computers costing only a few thousand dollars. Since then AI and robot brainpower has risen with improvements in computer efficiency. By 1993 personal computers provided 10 MIPS, by 1995 it was 30 MIPS, and in 1997 it is over 100 MIPS. Suddenly machines are reading text, recognizing speech, and robots are driving themselves cross country.

AI Computer power: 1900-1997

AI computers, rising from .1 to 1 MIPS in 1960, then from 1 MIPS to 100 in 1990s

The long stall: From 1960 to 1990 the cost of computers used in AI research declined, as their numbers increased greatly. The dilution absorbed computer efficiency gains during this period, and the power available to individual AI programs remained almost unchanged at 1 MIPS, barely insect power. AI computer cost bottomed in 1990, and since then power has risen instead, to several hundred MIPS by 1997. The major visible exception is computer chess, whose prestige has lured the resources of major computer companies, as well as the talents of special machine designers. Less visible exceptions probably exist in high value competitive applications, like petroleum exploration and intelligence gathering.

Progress in Computer Chess, a thin slice of AI

Computer chess rating rising steadily from 800 (infant) in 1956 to
over 2700 (world champion) in 1997

Agony to ecstasy: In forty years, computer chess progressed from the lowest depth to the highest peak of human chess performance. It took a handful of good ideas, culled by trial and error from a larger number of possibilities, an accumulation of previously evaluated game openings and endings, good adjustment of position scores, and especially a ten-million-fold increase in the number of alternative move sequences the machines can explore. Note that chess machines reached world champion performance as their (specialized) processing power reached about 1/30 human, by our brain to computer measure. Since it is plausible that Garry Kasparov (but hardly anyone else) can apply his brainpower to the problems of chess with an efficiency of 1/30, the result supports that retina-based extrapolation. In coming decades, s general-purpose computer power grows beyond Deep Blue's specialized strength, machines will begin to match humans in more common skills.

Related Articles

Ripples and Puddles, April 2000

When will computer hardware match the human brain?, December 1997