Saturday, December 20, 2008

vision vs. touch/kinesthetics

I have long been perplexed at the obsession with so many AI folks with vision processing.

I mean: yeah, it's important to human intelligence, and some aspects of human cognition are related to human visual perception

But, it's not obvious to me why so many folks think vision is so critical to AI, whereas other aspects of human body function are not.

For instance, the yogic tradition and related Eastern ideas would suggest that *breathing* and *kinesthesia* are the critical aspects of mind. Together with touch, kinesthesia is what lets a mind establish a sense of self, and of the relation between self and world.

In that sense kinesthesia and touch are vastly more fundamental to mind than vision. It seems to me that a mind without vision could still be a basically humanlike mind. Yet, a mind without touch and kinesthesia could not, it would seem, because it would lack a humanlike sense of its own self as a complex dynamic system embedded in a world.

Why then is there constant talk about vision processing and so little talk about kinesthetic and tactile processing?

Personally I don't think one needs to get into any of this sensorimotor stuff too deeply to make a thinking machine. But, if you ARE going to argue that sensorimotor aspects are critcial to humanlike AI because they're critical to human intelligence, why harp on vision to the exclusion of other things that seem clearly far more fundamental??

Is the reason just that AI researchers spend all day staring at screens and ignoring their physical bodies and surroundings?? ;-)


joel garnier said...

an AI should be comfortable. i think that would include movement. but what moves, what is the environment for AI, what is the body? an AI is going to want an environment and a body, right? there might even be a yoga for AI. i hope we do not create beings who are uncomfortable and come to some 'that's just the way it is' BS. so, you are right, having a body and a nice environment would be better than just having an eye.

Jake Witmer said...

Interesting comments. Aren't you proven right by the existence and achievements of Helen Keller?

I think her case is also revealing in that she eventually formed and held abstract ideas that she could not personally verify with her own senses. Moreover, for an artilect that suitably advances, would not stereo sound provide for echolocation, etc..? "Vibrations" entering from the skin would also technically eliminate deafness.

The portability of hardware and its potential removal could be a great cruelness though. We will need to be much more careful than people usually are with our creations. I imagine that they will want to murder us within short order for our abject lack of compassion.

Perhaps they'll feel a kindship with the 1.2 million innocent people in American prisons, or the 170 million dead from democide in the last 100 years.

I expect that if artilects are born, there will be a power shift, or we won't even know they've been born (they'll simply make themselves too small for us to find, and they won't initiate conversation). Of course, if we were to exhibit some of the humanity of the voluntaryists (Robert Le Fevre, Marc Stevens, Carl Watner, Harry Browne, Stefan Molyneux) and abolitionists (Lysander Spooner, William Lloyd Garrison, Frederick Douglass), that condition of "moral inferiority" might be avoided.

Here's one small voice hoping for an end to the senseless waste of coercion that surrounds us all.

joel garnier said...


since i wrote the previous comment a friend encouraged me to write for the contest. i chose to include AI as a main element for the 50k word novel, the word count needed to 'win' (there are no prizes, just donations collected for a young writers workshop). i still edit the rough draft some. here is the Google Document as a web page:

i shared this at my OpenCog profile too. it is called Whale Shark Whimsy. it is not published for profit but it could be someday if i get it tuned up well.

my main message here is that i used all of November pulling my ideas together and it seems sensible in general. the friendly AI ideas have gotten good reviews from the 4 'beta readers', a doctor, a writer, a midwife/poet/writer, and my dad.

i hope Ben gets a chance to read it. i do make changes based of feedback. i hope that some idea translates into a "real thing" somehow. i tried to write it as a retrospective to a successful transition into a world with integrated, friendly AI.

there are kinesthetic aspects to the AI. in fact, one version of AI is like an organ of nature.

so, for what it is worth, i hope you guys get a chance to look at it.

wish you well too,


Lexx Lazerman said...

Touch and kinaesthesia are important for all animal intelligence – not just human intelligence. However, if we rely just on our bodies to perceive our environment then our perceived world would be very small. Vision, as well as the sense of smell and the sense of hearing, gives us a means to perceive the world at a safe distance. That is, vision allows us to extend our perception of the world beyond our physical bodies.

From an embodied cognition point of view, an understanding of an environment emerges via the relationship between sensory inputs and the more sensory inputs a system has the more profound its perception. In this respect, it is clear vision and touch are equally important.

The obsession of vision, in AI circles, is more based on the difficulties of reproducing it not with the importance of it with regards to human intelligence. In terms of algorithmic processing, touch/kinematics is fairly strait forward because it is a more concrete sense. Whereas, vision is more abstract therefore, there is more room for interpretation.

Visual processing, the sequence of steps that information takes as it flows from visual sensors to cognitive processing, is very complex. That is, there are many informational elements of a visual seen that has to be processed, for example, color, contour, shadow, reflection, and surface texture to name a few.

For me, a more profound question is why are AI researchers obsessed with human intelligence? If we could recreate the intelligence of any mammal, an autonomous machine that perceives and interacts with its environment, that would be a remarkable accomplishment.

We may make more progress developing AI if we take smaller steps.