i'm a bit jetlagged. i'm in mexico city today, and just arrived from shanghai last night. i'm doing a really mediocre job of listening to an english translation of a spanish conversation, with the spanish conversation at full volume in the background. i tripped on steps right in front of me. i struggled to order breakfast off the hang tag card. why?
well first, it's my job :)
second, i'm tired - and when you're tired your ability to see things properly, to process information, can be affected. you struggle to find the word to describe a thing; you speak more slowly and with greater hesitation; your depth perception might be affected and you reach for something far away that's actually close.
we have complex systems for categorizing and naming things - for recognizing stuff. we have multiple words for an object - i'm wearing a shirt, blouse, top with pants, slacks, trousers, chinos and a pair of shoes, flats. i'm adorning myself with jewelry, ring. this assumes i'm thinking in nouns - identifying an object by its essential thing-ness. what if i look down and see gold with white with gold and more gold? would that tell you what i'm wearing? think about something round: it could be a ball, it could be a globe, or an orb. it could be a balloon, a sculpture, a bubble, a sphere; it could contain something or be empty; it could be solid; it could be made of nearly anything; it could float in the air or on the water or it could sink; it might roll or sit still. something round could be an infinite number of things. how do we decide if it's a golf ball or a balance ball or a mylar balloon or a snow globe?
on the one hand, this confusion is terrific - it's how we can achieve lateral thinking. what is this thing like? what can we learn from those other, like things? what can be borrowed or stolen or given back? we reason by analogy, by the familiar, we distinguish based on particulars, we limit the scope of what we think about to the context unless we are given permission or instruction to do otherwise. my mother used to say that she could hand an infant a purple clock and tell her it's a green toaster and how would the infant know? part of our ability to identify something is in our acquisition of language, our social negotiations over the names and provinces of things.
robots, on the other hand, live in a constant state of abstract, lateral thinking, when it comes to identifying objects. looked at from multiple angles, one object could be another. a tray could be a table; a book could be a shield; a couch could be a cow. so what's a young robot about town to do? or worse, what's a young soldier robot to do? via
Wired and the
DoD:
Recognizing and identifying an object from a video input turns out to be a very difficult problem. The problem stems from the fact that a single object can be viewed from an infinite number of ways. By rotating, obscuring, or scaling a single object, one can create multiple representations of an object - which makes the problem of matching the object to a database of objects very difficult. The problem expands exponentially when objects that need to be identified have never been viewed before. Combine these limitations with the wide variety of objects which might be identified, and the problem becomes intractable.
what will the military do with these shoddy, lateral thinking 'bots? apparently they take forever to identify objects and don't even do so accurately most of the time. they want to detonate your cat and pet your C4. which in some circles is perfectly acceptable; but in the don't ask, don't tell American military - we don't want to see too much creativity, it's sloppy. precision, march! the objective of the DoD project is thus:
To create an object recognition system based on the newer psychological models of object recognition by using a series of different algorithms to identify a variety of objects in different orientations. Such a system would be extremely beneficial for robotic control/intelligence and would allow for an exponential expansion of robotic capabilities and intelligence
and the next thing you know, they become self-aware and stuff. the most critical identification of an object any being can make is when they see their reflection, point, and say, "me."
Courtesy of Scientific American: a robot dragonfly that flies for 3 minutes at about 5m/second and carries an onboard camera. someday soon it will be able to hover like a hummingbird and fly backwards. in the movies, robotic insects that have cameras are usually sort of sinister and creepy and climbing all over things to expose the hero to his nemesis - but this one seems so delicate and elegant that surely it will come solely to its programmer to put it to evil ends... sigh.
Posts
