Intelligence is the Red-Herring of AI

Intelligence is the Red-Herring of AI

By Kiran Garimella   |   July 24, 2018
Kiran Garimella, Ph.D., is Chief Scientist and CTO at KoreConX, an Author, Speaker, Researcher (AI, ML, Blockchain) and an ex-CIO

One of the most contentious aspects of AI is the meaning of 'intelligence.' No one debates the meaning of the word 'strength,' or belittles the idea that machines can be stronger than humans, or even tries to re-define mechanical strength to mean some mysterious physico-spiritual capability that is unique to humans.

The debate around the meaning of intelligence when it crops up in any conversation on AI is extremely baffling - until we take into account the fragile psychology of humans. Somehow, we've convinced ourselves that cognitive abilities are the sole province of the human brain, while we grudgingly cede the physical realm to the machines.

Every encroachment on human cognitive abilities is fiercely contested. The first such encroachments by calculators somehow managed to sneak by the detractors of AI. These little electronic devices at first did basic mathematical operations faster and more reliably than humans. Schools still insisted, with some justification, that kids continue to learn the multiplication table on the grounds that calculators aren't readily available.

Whenever we put pen to paper to perform any arithmetic calculation beyond our unaided mental capacity, we now pull out our smartphone or call on Siri or Alexa! Have we lost our mental 'manhood'? How about algebra? Not only does my smartphone perform some algebra, but there are many online calculators that do it even better. Are we threatened by all this? Not me!

Whether it's robots performing microsurgery, machine learning algorithms beating chess masters, composing songs, or driving cars, machines with AI are definitely invading the sacred grounds of human cognitive ability.

But so what? Why this causes so much anguish—in fact why this should even be a cause for more than passing admiration—just baffles me. I'm equally puzzled by some people's attempt to play the 'yes, but' game of deprecating AI:

Computers can beat a grandmaster at chess, yes, but they couldn't possibly compose music.

Computers can compose music, yes, but they have to be taught all these rules, they cannot be creative by themselves.

Computers can paint, yes, but they can't drive cars.

Computers can drive cars, yes, but they can't cook.

Computers can cook, yes, but they can't analyze a legal document for errors.

Computers can identify errors in a legal document much faster and with more accuracy than lawyers, yes, but they couldn't really predict the outcome of a legal case.

Computers can predict the outcome of a case with astonishing accuracy, but they don't really understand case law, do they?

Yes, but... yes, but... yes, but... Nothing that computers do, it seems, can ever earn them respect! Never mind that most humans can't do more than a few of the above tasks with any competence either.

Red Herring

There are bizarre variations of this insistence that computers perform exactly as humans in order to be classified as intelligent. The more egregious example of this is the requirement for a computer (a robot) to go watch a movie and write a review that would be accepted for publication by the New York Times.

Must a robot go to the theater to watch a movie? Is the robot required to drive to the theater? Why not just analyze the movie file directly? Must the robot munch popcorn and drink soda while watching the movie? Perhaps even go to the restroom later? Must the robot open up a laptop and actually type in the review?

Most adults can perform these activities, yet not be able to type up a review that passes muster with the NY Times. Does that make all these adults unintelligent?

Debating the nature of intelligence and attempting to measure it is a futile exercise, whether applied to humans (remember IQ tests?) or software. Intelligence is not what is, but to paraphrase Forrest Gump (or more accurately, his mama), intelligence is as intelligence does.

That's where the real genius of Alan M. Turing becomes apparent. Forget your preconceived notions of intelligence and your fragile egos, he said in effect. If in a set of narrow tasks, a computer can fool you into thinking that it is a human, then it is said to behave intelligently (whether it actually is or not). If it acts, walks, and quacks like a duck, it is safe to consider it a duck—only in the context of those behaviors, of course, and not for eating!

In his original description of the now famous Turing test, Alan Turing was careful to design the test in such a way that the lack of technology in computers that mimics biological characteristics of humans would not affect the test. For example, both participants—human on one side of the screen and the other (unknown) participant on the other side—communicated only by typing answers and not by speaking, since the technology of the day could not produce artificial speech that might fool a human.

Accompanying these arguments to prove that AI isn't real (in some mysterious definition of 'real') is the deep-seated fear that AI (generally in the form of monstrous robots) would somehow develop consciousness (that terrifying "singularity") and run amok enslaving mankind and taking over the world.

What is conveniently forgotten is that what makes us look human, be human, and sound human—those universal characteristics that we all outwardly display without regard to race, religion, gender, geography, age, skin color, or level of education—have nothing to do with intelligence whatsoever! If there is one universal trait that makes us humans human is our unique ability to make rationality the handmaiden of emotion! And that is one trait we definitely do not want in our AI.

Advances in AI have forced the fundamental issue to the forefront: intelligence does not require a biological basis any more than strength requires a muscular basis. In this battle for dominance over the domain of intelligence between AI-enabled devices (or software) and detracting humans, the redefinition of intelligence by action and example is redounding more to the credit of AI than to the irrationality of the dissenting humans.

The lack of a clear definition of intelligence and our increasing inability to protect our human turf of intelligence should not threaten our spirituality or wisdom. Even if AI encroached on spirituality and wisdom in some way, it is a hallmark of true spiritual maturity and wisdom to accept that with equanimity and put it to good use for our benefit.