Peter Naur and the Jennifer Aniston Neuron

Peter Naur has an impressive biography. He was a pioneer of programming languages, at a time when the idea of higher-level abstractions from the deepest-level (“close to the electrons”) instructions was new, unusual, and anything but a given.

But the notions he presented at this year’s HLF seem to me to fall foul of the dictum ascribed to Einstein, namely that one should make things as simple as possible but not simpler. Naur presented a simple model of how the brain works: Neurons, connecting synapses, and a one-to-one mapping to a conceptual network (his example had an Audrey-Hepburn-node connected to nodes such as “beauty” and “actress”). His claim: the brain really is that simple. The numerous articles on memory in the 2003 Encyclopedia of the Neurological Sciences: enumerated by title and quickly dismissed as “What is in these articles is merely confusion.”

Proof for the simple one-neuron-one-concept model: The discovery of the Jennifer-Aniston-specific neuron that Naur read about in National Geographic (Danish edition, No. 2/2014). This specific neuron, he tells us, only fired when the human under study was shown images of Jennifer Aniston. Just as in Naur’s simple model: one node, one concept.

(Naur also self-published his theory in 2008; available from his website on naur.com.)

The good thing about conference WLAN is that you can augment talks by googling interesting concepts immediately. In this case, a quick internet search returns quite a number of interesting articles about the Jennifer-Aniston neuron (which is a newer and undoubtedly more mass-media-compatible version of the much older concept of a “grandmother neuron” from 1969): Here is a New Scientist article with more information – including some numbers that would seem to indicate that the initial study involved so few neurons and so few different images shown that it is quite improbable anyone could have found a hypothetical uniquely Jennifer-Aston-specific neuron. And when it comes to the question of what such experiments mean, here is a nice article by Robert Krulwich putting things into perspective. Here is a description of a more recent experiment which finds less specific neurons; the discrepancy could be linked to different stages of encoding memory.

Apparently it’s not as simple as Naur makes it out to be. I, for one, would really have appreciated getting all this information – and more, since what I’m listing here is just the result of a simple internet search during the talk, and thus likely to be the tip of the iceberg. Or, put more generally: I would like the HLF talks to be as simple as possible, but not simpler.

A video of Naur’s HLF talk can be found here.