ARTIFICIAL INTELLIGENCE by David Warren-Smith
David readily admits that he has had a life long fascination for Artificial Intelligence (AI) and robotics and has often given this a lot of thought over the years (you are not alone there, David). As it is frequently stated, we are really in the process of producing the next evolutionary stage that will (in maybe around 2020 David estimates) surpass humans to become the highest form of intellegence anywhere that we know of - although by then even an SN7400 might outpace most of us!
David developed the idea that all logic designs can be reduced to just combinations of combinational logic functions and, by applying feedback, memory. This applies to relay logic, electronic logic, and even biological logic as in the central nervous sytems of insects, animals and humans.
David said he would confine his talk to asynchronous logic design as opposed to synchronous, clocked logic. Asynchronous logic is closer to what occurs in biological logic and different functional areas can operate at full speed without being impeded by the slowest parts as in synchronous logic.
Rather than use some intuitive random approach that is often used to design logic functions, David compared that approach with a much more systematic and reliable method using state tables, Karnaugh mapping, and Boolean equations (ie. state tables leading to a circuit and not the other way around) to design a simple asynchonous logic circuit using both relay and then NAND logic gates to produce an R-S flip-flop.
David expanded this basic design to produce a circuit to light either a red or green LED depending on which of two buttons was pressed first ie. the circuit determined which order the buttons had been pressed. The circuit happened to be implemented in an Altera PLD logic chip. The design provided inherent switch debouncing, a LED only lit when both switches were operated, and there was no unpredictable behaviour such as race conditions that occur in many "intuitive" designs.
This then led on to some discussion of the Aibo dog with some learning skills and apparently emotions, and the Honda humanoid robot, Asimo.
You can get more from David's Web site at http://users.senet.com.au/~dwsmith/ and look up concept1.htm Also try http://www.ai.about.com/compute/ai/library/weekly/aa051899.htm for a list of AI links, http://world.honda.com/robot/ for Asimo, and http://www.ai.mit.edu/humanoid-robotics-group/kismet/kismet.html for Kismet.
There was lots of good food for thought in all that.
. . . Rick Matthews
Post meeting writeup by Mr David Warren-Smith:
Pre meeting writeup by Mr David Warren-Smith:
What is intelligence and can a digital circuit have it?
My contention is that a decision making ability together with memory forms
the basis of all manifestations of intelligence, both naturally occurring
and man made. The concept of the need for a decision making ability and
memory was the basis of the discovery that made automated computers
possible. As a digital designer I know that the logic of a circuit is
independent of the implementation of the circuit. Therefore I contend that
this concept also applies to naturally occurring intelligence. We just haven
't quite caught up with nature yet. However, it is only a matter of time
before we do. All the indications are that robots, humanoid or otherwise,
will eventually outperform all naturally occurring intelligent life forms in
every endeavour possible.
In the case of biological intelligence there is also an ability to learn. A
baby has a built in instinct to cry when it needs mother's milk, mother's
affection or when a change of nappy is needed. From there, everything has to
be learnt, either by trial and error or by being shown or copied.
Intelligence is demonstrated when the baby can start to make practical use
of what it has learnt on its own initiative. What enables the child to act
intelligently is its decision making ability together with its memory of
what it has learnt. Attempts to copy natures animal brains, in the form of
neural networks, have been going on for years with a great deal of success.
The basic logic components in their digital form AND, OR, NAND and inversion
are sufficient to form a digital circuit that has the necessary decision
making ability and memory to exhibit intelligence. In this talk I will show
you how asynchronous digital circuits can be constructed from these
components that have both memory and a decision making ability. An
asynchronous circuit is one in which the circuit responds immediately to any
changes in the inputs to the circuit, without the use of a clock source to
drive it, and in which the behaviour of the circuit depends on the past
history of its inputs.
I will show you that a systematic synthesis technique can be developed for
the design of these circuits that requires only a basic understanding of
Boolean algebra. This will show you how digital circuits can have the basic
ingredients of intelligence built into them. I will provide a very simple
practical demonstration of how this works. The talk is based on a paper that
I have been writing on this subject. There is a considerable amount of
material on all aspects of artificial intelligence available on the
Internet. The implications for the evolution of all manifestation of
intelligence, the impending ascendancy of robots, etc. should not be