Why even a moth’s brain is smarter than an AI

Emerging Technology from the arXiv, MIT Technology Review, Feb 19, 2018
Commentary by Stephen Downes

A moth can recognize odour after just few exposures. The olfactory learning system in moths is completely mapped, but artificial neural networks do not perform the same task as well as moths. The difference, explains the article, lies in how the two types of neural networks learn (that is, how the two types of neural networks create and adjust connections between neurons). The artificial neural network uses back propagation. In other words, it is given feedback from a training set. The moth doesn't. "The successful recognition of an odor triggers a reward mechanism in which neurons spray a chemical neurotransmitter called octopamine into the antenna lobe and mushroom body." I don't know how well this work with students (probably not so well) but it's interesting to note that this discussion of learning doesn't involve language, learning content, or the social construction of meaning, or any of the trappings of traditional learning theory.

Views: 0 today, 451 total (since January 1, 2017).[Direct Link]