Understanding Computers Understanding People

Does this simple title make you pause for a second to figure out what meaning is intended? Should there be a semi-colon after "computers" or is the phrase a clever conjunction of adjective and gerund? The choice leads to different meanings. Your language ability allows you to recognize the problem and you can make a likely correct choice in interpretation. But can a computer system do that? Conventional wisdom says that computers have no understanding of ideas whatever. They only understand numbers. Conventional wisdom says that a really good spell checker, grammar checker or translator cannot exist until computers are capable of true comprehension. Predictions suggest that this will come in fifty years. Conventional wisdom is wrong.

ACI has developed a technology that allows computers to understand ideas as they do numbers. ACI's technology can "understand" words in their context and correct spelling based on that understanding. To appreciate this capability, consider this example:

On hearing the word "air" a computer could choose air, heir, err, e'er, ere, are, or Ayr. Which of these homonyms is correct? Of course, the right answer depends on how the word is being used, i.e., its context. In the demo that follows, the profound capability of ACI's technology is revealed. The demo was written as a Microsoft Word macro, and it is as simple as basic math for a computer.

Conventional wisdom also says that computers will never be able to deal with analogies or understand humor. Again, conventional wisdom is wrong. ACI's technology can deal with analogies of the type found in SAT tests - typically better than most humans and certainly millions of times faster.

"What is black and dangerous and lives in a tree?" The answer is, "A crow with a machine gun." ACI's technology allows the computer to understand the concept of a crow armed with an automatic weapon. (We have to admit that while the computer gets the joke it still doesn't think it's funny.)

People may not understand computers but computers are closing the gap by starting to understand people. To illustrate this in somewhat more detail, let's look at how ACI's technology treats word analogies as math problems and then see how it can disambiguate homonyms in Microsoft Word.