Just as the AS (Artificial Sensorium) of the machine already surpasses that of the human in various ways, the AM (Artificial Mind) of the machine already surpasses the mind of the human in various ways. Kurzweil writes in his 1999 book, "A computer can remember billions or even trillions of facts perfectly, while we are hard pressed to remember a handful of phone numbers. A computer can quickly search a database with billions of records in fractions of a second. Computers can readily share their knowledge bases." (pp. 3-4). The IBM computer which beat Kasparov in 1997 calculated 200 million chess moves/second ("Beyond Human", PBS TV, May 22/01 in Vancouver).
LOGIC AND MEMORY
Most people recognize great chess playing as a feat of mind, a manifestation of high intelligence. But try psychologically analyzing it. The logic is quite simple. There are only so many rules for making moves and a normal adult human can learn them in a matter of minutes. These rules are what we could call the logic of the game. The great achievement of a chess master is the ability to store MANY moves and at great SPEED, in memory. Isn't that the standard way of rating a computer? We ask what its speed is and how much capacity in memory it has.
If I walk through downtown Vancouver, the next day I will be able to report few details compared to the total number en route. A digital camera on a robot travelling the same route will have ALL of the details in memory. The problem for the machine is that it cannot give a MEANINGFUL report the next day in human terms. Perhaps to Andromeda Sapiens the entire record is meaningful and the details are unimportant but to Homo Sapiens in the Milky Way Galaxy, details are required and those details are particularly required in terms of our friend from the last unit, figure-ground perception. How many robbers held up the bank? What were they wearing? Did they carry guns? All of these details require figure-ground perception.
Details like that of the number of robbers also require logic ability combined with the memory and figure-ground/object recognition ability. The gamut of mental abilities pertaining to arithmetic/logic/mathematics becomes important too. We might also ask our roving robot how fast the robbers were going as they went down the street. Perhaps the robot has a timing device and knows that it took 30 seconds for them to make their way from the bank entrance to the getaway car. And its range finding/artificial depth perception tells it that distance was 1,000 feet.Thus it can use a mathematical ability to calculate the speed. (PS-From here on "logic" ability will often refer to the complete gamut of arithmetic/logic/mathematics or a/l/m).
And let us say we are on a battlefield where such powers of observation, logic and memory may determine who wins and who loses. Many possible outcomes must be calculated like the possible outcomes in a chess game. The importance of the artificial mind becomes very clear when examples like this are considered. What also becomes clear is the extent to which the abilities of an AM (or AI) are like those of a human.
Schneider (1999, p. 25) credits Charles Babbage and his "anaytical engine" from the 1800's as the "father of the computer" although the analytical engine was theory only and not actually built. The modern compuer is credited by some sources to the ENIAC at the University of Pennsylvania in 1946 or the Mark I at Harvard in 1944. But a 30 ton monster performing a mere 5,000 calculations/second is a far cry from the modern computer. Somewhere along the line came "Moore's Law" which states that computing power is growing exponentially by doubling every two years or less while cost remains the same. So we are getting twice the "bang for the buck" every two years or less Of course Moore's Law is not a law in the sense of an immutable relationship. It is a generalization or extrapolation of a trend and the trend may continue or not.
Also important from the perspective of machine psychology and its biomimetic/psychomimetic theme is how compact all of this computing power is becoming. We can reasonably expect to see humanoid robots carrying tbe onboard computing power which some say matches that of a human. Kurzweil (1999) estimates that the human brain could maximally perform 20 million billion calculations/second based on the idea that a synapse firing is a calculation and all synapses firing would give us that number. He guestimates that by 2020 that much computing power will be carried by a $2,000 pc (p. 103).
FIRING ON ALL SYNAPSES
The old expression "firing on all cylinders" becomes "firing on all synapses". The relationship between number of synaptic connections and computer-equivalency is crude. First, it is unlikely that a human could ever have all synapses firing at one time in a perfectly co-ordinated and intelligent manner. Secondly, there are uni-cellular organisms with no synapses at all which seem to manifest their own kind of intelligence. Intelligence may exist also at an intra-celllular level. Certainly the DNA gives us a kind of intra-cellular computer and it is intelligent enough to produce another bio-computer which we carry on our shoulders.
Computing power beyond the "synaptic power" of the human brain would not have to be carried by humanoid robots in a roundish pumpkin which is fashionable for humans. The humanoids can be connected by wires or wirelessly to large mainframe computers. Michael Mayer, Manager of Pervasive Computing for IBM says "Before 2010, we will have computers with human brain capabilities." See
<http://www.ocregister.com> re "Digital Revolution" by Jim Krane, AP, Sept. 06/01. Professor Hugo de Garis from Kyoto Advanced Telecommunications says his "RoboKitty" will have the processing power of the human brain by 2007. See <http://www.theage.com> , "Artificial Brain Brings Food for Thought" by Garry Barker in "The Age" journal, 2000 AD. This would not be a bright human but the synaptic power of the machine would be in the human range says de Garis.
We have more than enough hardware now to develop SHAIs because mass storage of data files and program files is almost unlimited. Imagine a huge collection of computer disks on a shelf carrying world knowledge.
A robotic arms reaches out and inserts a disk into the computer as required. If the computer knows how to "data mine" or read all of that knowledge, it will be a most impressive SHAI so there we have a THIRD KIND OF SHAI, after the OCW-bot and Visualizing-Logic machine. This would be a READING MACHINE, a machine which can read with comprehension, operationally defined as the ability to give answers to important questions after the reading material is scanned in. A number of research centres are now working on such problems. The general problem is called "data mining". A specific and important aspect of it is reading text, ie sentences and phrases in a human language. The data mining expression allows us to include reading tables, charts, maths etc. The online discussion <[email protected]> covers related issues.