What is connectionism? is it something of the past or of the future?
Connectionism, in the words of the Internet Encyclopedia of Philosophy, is "an approach to the study of human cognition that utilizes mathematical models, known as connectionist networks or artificial neural networks. Often, these come in the form of highly interconnected, neuron-like processing units."
I have mentioned it before in this blog, but only in passing.
The point is to abstract from the actual functioning of the human brain to study a hypothetical brain that humans can model, in the way that a planetarium may suffice when the night sky is obscure.
The IEP suggests that connectionism had a "heyday," in the late 20th century, but that the heyday is now past.
This doesn't mean the approach turned out to be a dead end. Quite the contrary, IEP says that the idea has passed its heyday (few scientists call themselves by the label anymore) because its central insights have been widely accepted and integrated into other research programs.
One significance of connectionism is that in the study of artificial intelligence, the Turing model of digital computing, had come up against a dead end by the 1980s. Turing's system contained two germane elements: the separation of the central processor from the read/write memory; and the computation over discrete configurations of digits. Neither of those has a lot to do with that which humans regard as intelligence. Those features created a dead end, and connectionism has allowed for a work-around.
If HAL does start killing us, we'll know who to blame.
Comments
Post a Comment