Evolving Intelligence (EI)

Aristotle defined human beings as the rational animal, identifying our intelligence as the characteristic that differentiated our essential nature from that of other animals. Intelligence is commonly recognized as one of the highest forms of complexity to emerge in biological systems. How did this astounding level of functional organization arise? Was intelligence simply the result of a lucky accident? Did its emergence require fine tuning of multiple historical factors? Or was the emergence of intelligence inevitable or at least likely? Is intelligence really unique to human beings, or might similar mental capacities be found, or implemented, elsewhere?

Stephen Jay Gould in Wonderful Life (1989) argued that intelligent beings like ourselves were an evolutionary fluke, famously suggesting that if one were to rewind the tape of life and start again one should not expect a similar outcome. Others, such as Daniel Dennett in Darwin’s Dangerous Idea (1995) and Simon Conway Morris in Life’s Solution (2003) argue that intelligence is such a useful commodity that evolution could not help but have converged upon it. Religion has long considered the origin of human intelligence to be a spiritual mystery, with some holding that mental capacities belong solely to us by virtue of possession of a soul.

Herbert Simon, the Nobel laureate who is recognized as one of the founders of the field of Artificial Intelligence (AI), saw the human mind as “Wonderful, but not incomprehensible.” In 1956, together with his colleague Allen Newell, Simon designed the first AI system—Logic Theorist—which discovered proofs for theorems in logic. This pioneering research inspired a generation of investigators, who worked to design complex computer systems to mimic human cognition. However, despite some remarkable successes, the creation of intelligent machines has proved to be remarkably difficult and AI research has not fulfilled its early promise.

The PI of this grant (Robert Pennock) was a graduate student of Simon and has suggested that one reason for the difficulties is that Simon’s original empirical work led researchers to begin at the top with human intelligence rather than considering how that could have emerged from simpler forms of intelligence in other beings. Put another way, the problem may be that too much emphasis was placed on Aristotle’s notion of the rational and not enough on the notion of the animal. Rather than focus immediately on higher order propositional intelligence (i.e. knowing that), surely it makes more sense to begin with behavioral intelligence (i.e. knowing how) that is more commonly shared across species. Many of Simon’s key theoretical insights about rationality, such as the importance of pattern recognition and satisficing behavior, can be fruitfully redeployed in this alternative bottom-up evolutionary approach.

In our view, the failure of AI may reflect the difficulty of top-down design of something so complex as high-level intelligence, which all scientific evidence indicates evolved from (was built on a scaffold of) other forms of intelligence that existed before humans evolved. Could it be, then, that the best or only way to AI is not to design it from scratch but, instead, to create circumstances that allow it to evolve? Might there not be general evolutionary principles that would lead to the emergence of such complexity in digital as well as biological environments?

[Excerpt from Project Description]

Created 1/15/06. Last updated 5/19/08.
Return to Dr. Pennock's Home Page