Friday 30 January 2009

On Face Recognition

The organic cerebral source code that has evolved in our brains is a wondrous thing, and way beyond my humble comprehension. One of the more power subroutines of this complex neural code must be the one dealing with face recognition. We are remarkably efficient at picking out faces and instantly matching them from previous experience to immediately identify a myriad of friends, family and celebrities. This powerful process ensures that a couple of dots and a line are all we need to recognise a pair of eyes and a mouth. The consequences of this usefully evolved ability is the fact that we can also easily see faces when none are actually present. The Madonna on a pop tart, Jesus on a slice of a toast, a face on a mars, a monster in the curtains, and a ghost on the stairs seem to me more likely explained by this natural ability of our brains rather than crazy paranormal phenomenon.

The new face recognition software on iPhoto ’09 must work in a similar, but far less sophisticated way. My prediction, whilst installing iLife ’09 last night, and mulling this over, was that surely even the geniuses at Apple aren’t as clever as many hundreds of thousands of years of natural selection, honing and fine tuning our minds. Therefore surely iPhoto ’09 will also see faces where there are none.

The face recognition software turned out to be better than I had expected, despite the fact that I noticed on Stephen Fry’s twitter feed that he was having some problems getting it to recognise himself. However, as I suspected after sorting a large number of faces, iPhoto did eventually start seeing things. I’m not going to count, the cigar stall Indian, the statue of Horus and the picture of Carl Sagan on my T-Shirt, all of which iPhoto asked me to name. But as expected, iPhoto eventually found a face in the folds on my wife’s skirt, and in the craggy light and shade of a rock face. I write this not as a criticism of iPhoto ’09, which I was most impressed with, despite the fact that it did once ask me to confirm that a bearded and bespectacled gnarled old man was my beautiful wife. I’m writing this to ask a question; if we can now write software that not only efficiently apes one of our more complex brain functions, but is also prone to the same delusions as us, does this make genuine artificial intelligence or even artificial consciousness seem just a little more probable at some point in our species future?

2 comments:

Pete UK said...

what pleases me is that iPhoto is exhibiting just the same tendency towards false positives as our own brains. It rather suggests we are on the right lines!

Shadeburst said...

Crispian I believe that you are English. English people don't say "to immediately recognise." However I'll give you a Noddy Badge for spelling "recognise" with an S not a Z.