Universal tests of intelligence for humans and animals
Scientists are developing a new intelligence test that can be taken by any living creature. It will allow for comparison of intelligence between humans and animals.
According to Monash University News, until recently it has been pretty difficult to figure out intelligence without the test being based around language, which has made it impossible to gauge how intelligent a chimpanzee is compared to a human, or a computer to a lowly worm.
Associate Professor Dowe, along with Dr Jose Hernandez-Orallo of the Universitat Politecnica de Valencia in Spain think that a total test of intelligence can be created that goes further than the current limits of language.
The test would utilize the method of operant conditioning, where the subject has to figure out the puzzle by trial and error.
For example, if the test were tic-tac-toe, the subject taking the test, having never seen the game before, would initially have to discover that the game is won by getting three in a row on a three-by-three grid, before playing it.
The best part is that correct answers would be rewarded in a manner that is appropriate for the user, like money for a human or a banana for a chimpanzee.
Subjects taking the tests who are performing well get upgraded to a more difficult test, while those performing poorly will get downgraded to an easier test, as measured by the principles of algorithmic information theory. This triggers the Minimum Message Length (MML) principle of machine learning, econometrics and statistical and inductive inference. This may remind some readers of the GRE, a standardized test that uses a similar principle.
The paper which details the work is currently the most downloaded article in the prestigious journal, Artificial Intelligence, but it’s behind a pay wall. It has also been covered in The Economist, in the March 5 issue from this year.
Information provided by: Monash University