Internet users, those few of you out there, know that many websites require passing a CAPTCHA test before submitting information. That’s where you have to read distorted text and type it into a box. CAPTCHA, an acronym for "Completely Automated Public Turing test to tell Computers and Humans Apart," has been a mainstay of computer security for years. (For more on the Turing test, see ENV‘s Michael Egnor on "Can a Computer Think?") CAPTCHA takes advantage of humans’ ability to recognize text even when it is grossly distorted or broken, or placed against complex backgrounds. Software has been stymied by that ability — till now. New Scientist reported that a California team named Vicarious has broken it, at least partially.
Previous attempts at breaking CAPTCHA were unsuccessful because they thought like computers. The new algorithm thinks more like a human brain. Using neural networks, the software builds up the image in stages, or "nodes," first looking for patterns in small areas of the image and expanding the search in larger segments. Millions of nodes may be iterated. When shapes are detected, they are compared with what the software has learned about characters:
The strength of each neural connection is determined by training the network with solved CAPTCHAs and videos of moving letters. This allows the system to develop its own representation of, say, the letter "a", instead of cross-referencing against a database of instances of the letter. "We are solving it in a general way, similar to how humans solve it," says [Dileep] George [co-founder of Vicarious]. (Emphasis added.)
According to Science Now, the company claims a greater than 90% success rate. Fortunately for Internet users, the company has not released the algorithm — or else computerized bots might wreak havoc on Internet security.
Vicarious spokesmen consider their feat a breakthrough in artificial intelligence (AI). Science Now reporter John Bohannon reserves some skepticism about the "breakthrough" claim, since the algorithm is unpublished. In fact, he sent Vicarious some CAPTCHAs that tricked the algorithm. The designers had to tweak some things to train it, then it succeeded; for instance, it didn’t recognize Cyrillic characters ("We haven’t trained our system on other languages yet," the spokesman said.) Bohannon recognizes, though, that if the claim holds up, this could represent a milestone on the road to designing computers that see the world like humans do.
This story does not mean that computers are developing minds. Dr. Egnor makes it clear that computers and software are artifacts of human minds. The computer is not independently exercising its own will; it is following a process that was impressed into it by intelligent design. Humans are "very clever" in their ability to make "astonishing artifacts," Egnor points out. Breaking CAPTCHA may be a milestone in artificial intelligence, but it is not a breakthrough toward artificial mind. The computer executes an algorithm; it does not perform a mental act. As Egnor pointed out, passing a Turing test would only prove a clever human’s ability to fool another human. "The only thinking in a computer is the thinking imparted by the engineers and programmers who built and programmed the computer," he says.
Since intelligent design presupposes a mental act directed toward a purpose, AI is a misnomer. It should more properly be described as "artificial execution of human-designed algorithms."
This is really a story about biomimetics — a form of intelligent-design science. The engineers looked to the way a brain solves a problem and tried to imitate it. It took human intelligent design to design the computer. It took intelligent design to write the software. It took human ID to test it, tweak it and perfect it till it succeeded. It requires human intelligence to see a good design. It takes ID to formulate a purpose. Then it requires human intelligence and will to move things in a preferred direction for that purpose. Nothing is left to unguided processes. Even selection from random trials (falsely called "Darwinian" algorithms) employs human purposeful choice.
If human intelligence is responsible for the software success, where did human intelligence come from? And aren’t other animals intelligent, like whales, dolphins and birds? Sure, but to suppose that they and we evolved by unguided processes begs the question. We have no experience of intelligence (artificial or biological) arising by matter in motion. Some people may have faith that it can, but they cannot point to evidence that it did. No; we know by "Vicarious" experience that thinking, problem-solving entities are products of intelligent design. What allows you to "tell computers and humans apart" is that both are designed, but humans are also designers. Our experience of ourselves as designers, combined with the impotence of all material explanations, rationalizes the existence of a designer for our bodies and minds as an inference to the best explanation.