For the past few decades artificial intelligence researchers have generally believed that connectionism was the key to building a generalized AI system. Short version of connectionism: symbolic thought is the emergent result of connections between billions of neurons, each of which individually plays only a small, distributed role.
It's pretty surprising that researchers seem to have found evidence that individual neurons can identify objects as dissimilar as sports cars and dogs because the predominant theories expect that such high-level symbolic recognition would be distributed across a large number of neurons, not concentrated in any recognizable location.
In previous studies, Earl K. Miller, Picower Professor of Neuroscience, found that individual neurons in monkeys' brains can become tuned to the concept of "cat" and others to the concept of "dog."
This time, Miller and colleagues Jason Cromer and Jefferson Roy recorded activity in the monkeys' brains as the animals switched back and forth between distinguishing cats vs. dogs and sports cars vs. sedans. Although they found individual neurons that were more attuned to car images and others to animal images, to their surprise, there were many neurons active in both categories. In fact, these "multitasking" neurons were best at making correct identifications in both categories.
Of course, there are still multiple neurons involved in any of these recognition problems, but it's still striking that such distinct localized behavior can be observed on single neurons.