Tuesday, August 21, 2007

Intelligence Amplifiers vs Amplified Intelligence

One of the standard gimmicks of science fiction is the “intelligence amplifier”, a wonderful gadget that makes you smarter.

Great idea. But if we want to really build one, instead of creating it in fiction, we need to know more. For starters we need to know what we mean when we say intelligence.

Like most really simple concepts, intelligence turns out to be a real bear to define effectively. The obvious definition is that “intelligence is what makes you smart” – which is massively unhelpful. A more sophisticated definition is that “intelligence is what intelligence tests measure.” Which is tautologically true and again absolutely unhelpful.

So let”s take a look at the typical intelligence test. It contains questions on completing sentences, drawing analogies, matching patterns, inferring the next item in a sequence and, of course, mathematics. Given modern computer power, it is not that difficult to design software that will function as an electronic cheat sheet. That would undoubtedly increase anyone”s score on an intelligence test, but intelligence is supposed to measure a quality that is useful in the real world. How often do you have to complete sentences or decide what a box will look like when it is folded up?

To further complicate the picture, the current wisdom is that there are different kinds of intelligence. Not only is there the sort of intelligence measured by the grab bag of concepts on a conventional intelligence test, but there is also “emotional intelligence”, which measures how well we deal with others. Considering all the super-bright, super-maladapted social retards out there, I'd argue that emotional intelligence is about as important in life success as conventional intelligence.

This isn't getting us anywhere, so let”s look at a simpler question: What's an amplifier?

The obvious answer is “an amplifier is something that makes things louder.” Or in a slightly more sophisticated version an amplifier increases a signal of any sort.

However unlike “intelligence”, “amplifier” has an easily stated, but not-quite-obvious-definition. Technically an amplifier is a device that controls one power stream with another.

Since I”m of a certain age, let's start with a triode vacuum tube, the first real electronic amplifier. (A basic transistor works the same way.) A triode tube, for those of you who have never seen such a beast, is essentially a light bulb on steroids. You've got the glass envelope (the bulb) enclosing a filament (the thing that lights up and burns out in a light bulb) and two other components: the plate and the grid. The plate receives the electrons boiled off the filament and the grid, which can be charged, sits between the two.

When the tube is turned on electrons flow from the filament to the plate, passing through the grid. That produces a nice, strong flow of current (power) from the filament to the plate. But if you turn on the grid you interrupt the flow of current. Even a very small amount of energy applied to the grid can strongly reduce the filament-to-plate flow of energy and enough charge on the grid can completely shut off the flow, like flipping a light switch.

Now if you modulate the energy supplied to the tube”s grid in some way – say by speaking into a microphone – you also modulate the energy flowing through the tube, but on a much larger scale. Even a minuscule amount of energy from the microphone can produce an enormous change in the energy flowing through the triode. Enough to drive a bank of powerful loudspeakers say. Voila! The sound is amplified.

With that out of the way, let”s go back and look at the notion of intelligence again. All forms of intelligence seem to be made up of two components. There”s reasoning ability, or how well you can work with information, and there”s knowledge, or what you know about the world around you. In the case of a conventional intelligence test you have to be able to define words, know the basics of mathematics and understand such ideas a “car” and “miles per hour”. Even the folding box questions require knowledge of how objects fold. (There”s also memory, but for reasons that will become clear in a second, we”ll fold that into “knowledge.”)

So far no one has figured out a good way to increase the reasoning ability of a normal person, although there are some drugs that apparently have a small effect. We can teach logic and general semantics, and we can provide people with heuristics (rules of thumb) to help them analyze situations, but all those things only work if the person is able to reason in the first place. In other words, they are basically forms of knowledge.

If we can't increase reasoning ability, can we increase knowledge? The answer here is not only “yes”, but “hell yes!”. In fact the story of human civilization is in large part the story of increasing knowledge.

Until recently the difficult for the average person was getting hold of the knowledge. We spend enormous sums to impart that knowledge through schools and universities, with highly varying degrees of success. We have libraries packed with it, but unless you were able to visit those libraries that knowledge wasn't available to you.

A more fundamental difficulty is that we seldom want to know all, or even most, of what humanity knows about anything. We just need enough information, in an appropriate form, to solve specific problems or perform specific tasks. Our educational system is focused on providing a broad background on subjects, which may or may not answer our questions. To find the answer in a library requires knowing enough about the subject to know where to look.

I say “until recently” because computer networks, especially the Internet and its successors are changing all that. A few years ago Jerry Pournelle predicted that in a couple of decades we”d have all of human knowledge available through our personal computers. It looks like “all human knowledge” is stretching it, but certainly a good part of it is becoming available.

Already I can make a couple of mouse clicks and type a few words and get answers to most of the answerable questions I can think of. I can't discover the Secret of the Universe (and “42” doesn't qualify unless you know “base what and modulo what”?) but I can find out how to do anything from open-heart surgery to making fancy wooden boxes, I can communicate (or at least listen in on) experts discussing nearly any subject I can name, and I can ask them questions. (Which they may or may not answer.) If I want to learn a subject from quilting to tensor calculus there are tutorials available on the web to teach me – and usually if I can”t understand one tutorial there will be two or three others to choose from.

This is a long way from perfect. We still suffer bandwidth limitations that impede our access to that information. A bigger impediment is the old question of knowing how to find the information. Often this becomes a matter of spending a couple of hours trying to guess the magic words that specialists use to describe the things I'm interested in. The search and organizational features of the Internet are still pathetic. The signal-to-noise ratio is bad. In short it takes special skills in everything from critical thinking to search engine operation to get the most out of the system.

But all this is changing. Searches are getting better as more sophisticated search engines and completely new search methods are developed. Organization is improving. And bandwidth, processing power and accessibility are exploding.

For now most of us access this information sitting at a desk. If you want you can get a laptop with a wireless connection that lets you work just about anywhere. The next step is going to be an appliance we can always carry with us, with input and output devices that don't tie us down.

Suppose instead we had something that looked like a pair of sunglasses for a display and a microphone for voice input (or better yet something we could control with brainwaves). Let's also suppose that it had the bandwidth to maintain several high-capacity connections at once. Let's back this up with software that could monitor hundreds of channels of information and bring the stuff we might be interested in to our attention automatically. We could be accessing several sources of information, interacting with four or five people and monitoring a whole bunch of additional information, all while dealing with the real world.

At that point you”re using the relatively limited amount of brainpower possessed by a human being to control the enormous flow of information in a sophisticated computer network, the result is not an intelligence amplifier, but amplified intelligence – which is almost the same thing.

Almost, but not quite. The output of an amplified signal depends largely on the quality of the signal being amplified. In other words if you apply amplified intelligence to a moron you don”t get a smarter person, you get a more knowledgeable moron. This is easily demonstrated by a quick scan of Usenet news groups.

One of the problems you find in the news groups is that some of the morons have laid hands on a lot of knowledge. They don”t understand what they've read, they can”t really apply it, and because they are morons they don”t understand their output is not merely useless and annoying, it can be downright dangerous. (Check out some of the processes for making explosives circulating on the web.)

Traditionally one of the things that held people with severely impaired judgment in check has been their lack of knowledge. Most of the time they didn't know enough to be dangerous because the processes necessary to gain knowledge pretty effectively winnowed them out. (This wasn't an absolute protection, as the many examples of well-educated jackasses demonstrate, but it was some help.) Now the morons can find that knowledge on the Internet, but they still don”t have the reasoning ability to use it, or even really understand it in most cases.

If you wanted to, you could use the knowledge on the Internet to build anything from a web page to an airliner. Fortunately, AFIK none of the morons have tried to build an airliner yet, but the number of absolutely awful web pages out there gives you some hint of the magnitude of the potential problem.

Or try a more nightmarish case. The information on gene splicing is widely available. It”s also pretty well known that there is probably a way to create a strain of smallpox that is not only resistant to vaccines, but can re-infect the same person multiple times. Suppose some moron decides to try it out?

The good news is the same amplified intelligence will make it possible to find a cure for such a super-bug a lot faster. The other good news is that even simple gene splicing requires not only knowledge, it requires experience as well. The techniques aren't all that complicated, but making them work requires a fair amount of hands-on laboratory experience. The web doesn't amplify experience, except indirectly. (There”s also the fact that the remaining stocks of smallpox virus are carefully controlled.)

Are we doomed? Will we wind up surrounded by morons with megaphones? Not necessarily. Certainly the morons we will have with us, but we can build much more complex amplifiers than the simple triode example described above. We can take the input energy stream and modify as we amplify it by the use of things like filters. We can amplify only part of the energy stream and suppress the rest.

In principle we can do exactly the same thing with amplified intelligence. Through good design and “smart” software we can suppress some of the noise and amplify the real signal – the information we truly want.

Where does amplified intelligence lead us? Darned if I know. One of the things not even amplified intelligence can do is predict the future.

Not yet, anyway,

No comments: