I found an interesting book at my local book exchange,
The Society of Mind by Marvin Minsky. It is a series of short essays about how the mind works, and article 7.2, Uncommon Sense, just struck me as amazingly obvious, but I never thought about it that way before.
We've all heard jokes about how stupid present-day computers are. They send us bills and checks for zero dollars and zero cents. They don't mind working in endless loops, repeating the same thing a billion times. Their total lack of common sense is another reason people think that no machine could have a mind.
It is interesting to note that some of the earliest computer programs excelled at what people consider to be "expert" skills. A 1956 program solved hard problems in mathematical logic, and a 1961 program solved college-level problems in calculus. Yet not til the 1970s could we construct robot programs that could see and move well enough to arrange children's building-blocks into simple towers and playhouses. Why could we make programs do grown-up things before we could make them do childish things? The answer may seem paradoxical: much of "expert" adult thinking is actually simpler than what is involved when ordinary children play! Why is it easier to program what experts do than what children do?
What people vaguely call common sense is actually more intricate than most of the technical expertise we admire. Neither that "expert" program for logic nor the one for calculus embodied more than a hundred or so "facts"--and most of them were rather similar to one another. Yet these were enough to solve college-level problems. In contrast, think of all the different kinds of things a child must know merely to build a house of blocks--a process that involves knowledge of shapes and colors, space and time, support and balance, and an ability to keep track of what one is doing.
To be considered an "expert," one needs a large amount of knowledge of only a relatively few varieties. In contrast, an ordinary person's "common sense" involves a much larger variety of different types of knowledge--and this requires more complicated management systems.
There is a simple reason why it is easier to acquire specialized knowledge than commonsense knowledge. Each type of knowledge needs some form of "representation" and a body of skills adapted to using that style of representations. Once that investment has been made, it is relatively easy for a specialist to accumulate further knowledge, provided the additional expertise is uniform enough to suit the same style of representation. A lawyer, doctor, architect, or composer who has learned to deal with a range of cases in some particular field finds it relatively easy to acquire more knowledge of a similar character. Think how much longer it would take a single person to learn to deal competently with a few diseases and several kinds of law cases and a small variety of architectural blueprints and a few orchestral scores. The greater variety of representations would make it much harder to acquire the "same amount" of knowledge. For each new domain, our novice would have to learn another type of representation and new skills for using it. It would be like learning many different languages, each with its own grammar, lexicon and idioms. When seen this way, what children do seems all the more remarkable, since so many of their actions are based upon their own inventions and discoveries.
So they will never teach a computer to be an eyas, because they'd have to teach it to regulate its own temperature, play with its food (feet especially), flap its wings, jostle its nest mates, sleep in a huddle, poop, take bird baths in the gorge, catch food that's been dropped to it while flying, stoop turkey vultures and mourning doves, perch on Mercury and other buildings in the area... or any of the other wonderful things an eyas/fledge can do...
Here's to child's play!
