Expert: Saurabh Kudesia - 5/29/2006

Question
I am trying to evaluate a supposedly
"human thinking" machine program.

Typical IQ Test questions are typed in,
and correct answers are displayed. For example,
all apples are red, but no apples are square.
Are any square objetcs apples?

I've seen a computer answer these type of logic questions correctly, but I'm thinking this type of program could easily be written, and it is
not truly a "thinking machine" that is displaying human like intelligence.

I cannot reveal much more info than this right now, as there is a patent pending on this AI computer program, and I have signed a none disclosure agreement. I'm just trying to figure out if this demonstration proves anything?

Your comments will be appreciated ... thanks!"


Answer
Hello Dennis,


Let me elaborate further to your query, specially this section:

"I've seen a computer answer these type of logic questions correctly, but I'm thinking this type of program could easily be written, and it is not truly a 'thinking machine' that is displaying human like intelligence."

Although significant advancements have been made in the development of artificial organs, all these major breakthroughs appear to be insufficient for synthesizing human like intelligence. Computer can be programmed to play chess, beating everyone and even create certain signs of mind but does that mean that computers can be as intelligent as humans? Doesn’t human like intelligence require a degree of awareness and understanding of what one is doing? Do we really know the meaning of intelligence? Even the Deep Blue chess computer had 256 chips, which could evaluate and consider all legal moves from a certain position in one electronic flash, orchestrated by a 32-processor mini-supercomputer. It examined 200 million chess positions a second. Chess programs, on unaided general-purpose computers, average about 16,000 instructions per position examined. Deep Blue, when playing chess (and only then), was thus worth about 3 million MIPS, 1/30 of our estimate for human intelligence. What if this number grows to 90 million MIPS? Could Deep Blue be as intelligent and aware of its environment situation as we are? Is it possible for an intelligent looking robot to have the awareness and understanding of what it is doing? Artificial intelligence is still not the same as organic intelligence though it may be more prolific.

Since understanding seems to be a conscious activity, conscious thinking has a non-computational character. Of course, consciousness has many other dimensions. The perception of color red, for example, is something that requires consciousness, as does the sensation of pain. Also, calculations may or may not terminate. But what is the rule that defines that calculations have ended? Why does the sum of 2 and 2 end up with just 4? Is answer to this question complete or just part of a temporary arrangement to answer this question? How will you let the machine know that the answer has really been obtained? Are there any other rules, calculations or algorithmic procedures to ascertain this? On the second note, can any biological machine including humans be ever be able to tell when did solutions starts and ends?

So, you are absolutly correct when you assume that the machine can actually show the sign of intelligence when infact, it was just mimicking the program codes. The only way you can decide whether your program is really intelligent is to give a similar type of query, rather little deviated from the original one, that can test whether the program is able to draw conclusions, and therefore, is able to relate new situation with its experience, or not. The program can only be labelled as truly intelligent when it is able to take logical conclusion from one situation to another situation, without breaking the laws that determines its 'creative' ability, to draw another set of conclusions. This is what is known as the evolving characteristic of the organic thinking process.

All the best in your endevour.

Saurabh