The intelligent camera

photoIn his book “The Singularity is Near” futurist Ray Kurzweil points out that people are always asking when is artificial intelligence coming, but that it’s already all around us.  It’s just that every time it gets developed, we call it something else.  A real case-in-point is the modern digital SLR.

My father’s camera was a Ciroflex twin lens reflex camera, the “poor man’s Rolleiflex.”  It focused manually with a thumb dial and that was about all it did. There was no flash and no light meter.  For exposure you had to use either the guesstimates that came with the Kodak roll film instructions or an external exposure meter.  My father would mount this camera on an enormous set of external floodlights during birthday parties.  My sister and I still laugh at this since these  slights were so bright that some of our childhood friends may still be seeing after image.s all these years later.

My first SLR was similar in its capabilities, a Praktica from East Germany.  Then I saved up my pennies and bought myself a Leica M3, arguably the best camera ever made.  So let’s say it’s 1968 and I wanted to photograph a flower, but also wanted the building behind it in focus.  The process was 1. consider the scene and decide on an f-number that’s going to give you the needed depth of field, 2. set the f-number, 3. choose the point that you want to use for focusing, 4. move the exposure time dial which was linked mechanically to the exposure meter until the the needle lined up in the center,  5. look at the exposure setting and decide whether it was fast enough to hand hold, 6. focus the lens on the flower using the split screen parallax range finder, 7. note the distance to the flower and then mover the lens focus so that both that distance and the distance to the building were both bracketed by the appropriate f-number lines on the lens barrel.  8. push the shutter.  I used to be pretty good and pretty fast in executing this set of events seamlessly.

Moving on, in 1985 Minolta, with the introduction of the Maxxum 7000, developed “autofocus.”  I remember clearly at the time that many reviewers though this pretty much useless.  I mean, who cannot focus a camera? Today we may be moving towards a crisis on the other side.  With the abandonment of split screen parallax views on the focusing screen, manual focus is becoming harder and harder, even under the many conditions where it might still be preferable.

OK, so today with my Canon T2i, I choose depth of field mode, “A-Dep.” My camera chooses the exposure time for me based on my chosen f-number.  This is based on the placing nine focal points into simultaneous focus.  If this cannot happen, the camera tells me.  Focus is achieved automatically.  Really, if I don’t want to fuss with it,  all I have to do is push the shutter button.  If you think about it, the only thought process required of me was the knowledge that A-Dep was the best way to take this picture.  The little person that lives inside my camera is pretty smart – smart in all sorts of  amazing ways!

This is really a clear example of artificial intelligence, as long as we don’t quibble over the semantic question as to whether an algorithm-based program incapable of learning is true artificial intelligence.  But this definition is pretty satisfactory for most folks, and, perhaps of equal or greater importance,  nonthreatening in that we don’t expect the intelligence in our cameras to declare “cogito ergo sum,” or to emote in response to the subjects they are photographing by, say, crying or laughing.  Most importantly they are unlikely in this stage of development to pass the Turing test.*  But they really do take great pictures!

* The Turing test is a test of a machine‘s ability to exhibit intelligent behavior in such a way that it is indistinguishable from intelligence.  In Turing’s original conception of the test a human asks questions of another human and a machine.  The judge cannot see the participants and they only communicate via a single mode such as a key board and monitor.  That is the test is designed such that the machine does not have to replicate human speech.  If you’ve ever watched the closed captions with the news, you’ll realize that this last point would still be critical.

Close Menu