MadSci Network: Zoology |
How well would an educated primate do on a Turing style test? Speaking for educated primates everywhere, I think that I would do quite well on such a test. Rather than “educated” you are asking about those non-human primates, apes mostly, that have been taught some type of human communication skill, such as American Sign Language. How would they perform against Turing’s test. First, let’s look at Turing’s test. I’m not an expert in evaluations of intelligence, or computer artificial intelligence, and I would ask you to direct a question to those Mad Scientists for more detailed information. But, what Turing proposed was more of a standard then a test. Simply stated, artificial intelligence in a computer would be judged successful if the computer could engage in a conversation, via teletype or some other instrument, with a human who would be unable to discern that they were not conversing with another human. Note, that this is not a test of “intelligence” per se, and that it is specifically geared to evaluating human-computer interaction. It would have many limitations if it were brought outside the realm of computer science. For example, how well would a three year old child perform on this type of test? I brought up the example of a child for a specific reason. Apes that have been taught human communication are often compared to a child when evaluating their communication skills. I am aware of no experiment that tried to evaluate ape skill using a Turing type of test. I suspect that this is because a human interpreter would always be necessary when sign language is involved – either the human is communicating directly with the ape (and therefore knows that it is an ape) or another human is acting as an interpreter between the ape and human. When you have a human intermediary, you may be evaluating their interpretive skills more than you are evaluating the intelligence of the participants. Some apes have been taught a symbolic keyboard language, so that they communicate only by pressing symbols on a keyboard. For the most part, these experiments were constructed to evaluate the ape’s ability to learn grammar (nuances of meaning based upon the sequence of symbols) without the need of a human interface. In most cases, the apes in these experiments are communicating primarily with a computer (I wonder if they can tell the difference?). More valuable tests of non-human intelligence are those that do not rely on, or make use of, human language skills. Non-humans have been given intelligence tests from the very beginning of intelligence testing, with varying degrees of success. Who is smarter, the dog that learns to sit-up and beg for a treat, or the cat that will ignore you until you provide the treat anyway? These tests are constantly confronted with the problems of “What is intelligence?” and “How can we evaluate intelligence in a way that is independent of species/cultural perspectives?” These are deep and difficult questions, and I leave it to the experts in this area to explore their ramifications.
Try the links in the MadSci Library for more information on Zoology.