Neural Networks Help Computers Recognize Cat Videos on YouTube (and Help Us Understand the Human Mind)

We spend a lot of time in this class talking about a few types of networks, types that have been thoroughly researched for years; real life social networks, social media networks, and networks of communication hardware (telephones, the internet).

What we haven’t talked much about is the vast network of neurons that composes our brain.  This has been a lively area of interdisciplinary research over the past few years, with computer scientists and neuroscientists developing various forms of artificial intelligence based on the operation of the human brain.  The article that I link to below specifically discusses Google’s new version of its “Jelly Bean” voice-recognition software, included in the newest Android OS.

When a user gives a voice command to his or her Droid phone, an image of the audio waveform is sent to Google’s servers, where it is then processed by an algorithm based on human neural networks.  What I think is interesting about this research (and its application) is that two important things are happening at the same time.  First, neuroscientists are finding that the human brain can be meaningfully understood as a collection of networks.  Second, computer scientists are finding that, with this new understanding of the brain, they can create computer programs capable of tasks once thought to be the sole purview of the human brain, such as the voice and language recognition so crucial to the performance of Jelly Bean (and more specifically, the recognition of cats in YouTubes videos).

While this second area of research has the most obvious potential for having a great impact on our daily lives, it is the first area of research, on the brain itself, that I find most fascinating.  Outside of economics, my main academic interest is in the philosophy of mind, and I see the potential for a network-based understanding of the brain to fundamentally change our understanding of what the mind is.

For example, one of the primary problems in the philosophy of mind is our inability to explain subjectivity and identity (why I experience myself as myself, why I understand myself as separate from the rest of the world and other people, why I understand myself at five years old as the same person I am now).  A week or so ago, my Philosophy of Mind professor, Gerald Vision, was lecturing on John Locke’s memory-based solution to this problem (essentially that our thoughts are all connected to each other by first-person memory) and noted the strongest criticism to Locke’s argument is that memory is “gappy” – we tend not to recall everything that’s ever happened to us, though things we don’t currently remember must be included in our conception of our self (otherwise, we could reject any action we don’t remember as not our own, even if it is simply something that happened long ago).  However, if we conceive our mental states as a connected network (rather than a simple linear progression of thoughts) we might be able to solve this problem of identity by defining it in terms of a connected component of mental states.  I don’t necessarily remember what I had for lunch three days ago, but I remember being in a van driving out to Harrisburg three days ago after lunch, and at the time I was in the van I certainly remembered what I had recently had for lunch.  In that way, my current mental state is still connected, by some pathway, to the mental state of having lunch three days ago.  It’s a silly, simple example, but it makes the point.  On top of this, the idea of cycles, redundancy, and degrees of connectedness might be used to explain why some thoughts and memories remain more prominent in our minds than others.

So not only are networks useful for developing our understanding of the material brain, but they have the potential to help us understand the more abstract, difficult concept of the mind.

Anyway, here’s the article on Jelly Bean!