You employ a most dangerous argument when you seek to use a lawyer’s own words against them. In the present case, you have repeated back to me what I told you recently, namely that a wise person’s reasoned judgment must come to the same conclusion when presented with the same facts. You start with my conclusion, and you ask whether this means that a computer is wise. And taken a step further, will artificial intelligence become the ultimate embodiment of the wise philosopher, giving in to neither emotion nor temptation, providing consistently correct answers to all our problems?
Since it is my own words you employ, I cannot dismiss them lightly. Let me take you on a small journey before we arrive at our conclusion together. Let us travel using one of the most impactful inventions of the industrial age, the combustion engine powered car. The first modern automobile was invented in the 1880s by Karl Benz. If we could have an insight into horses’ thoughts at that time, I wonder if they would have run something like this: “Foolish humans. They have a perfectly reliable companion in us horses. We pull their loads, plow their fields, and take them to every corner of the earth. How much farther do they want to go, and how much faster do they want to get there? Can you ride two horses at once? Ten? And how will they feed them all? Mmmm, is that a carrot you’re holding?”
Little did our faithful equine friends know how insatiable are our human appetites. Today mere passenger cars can sport more than a thousand HP, a veritable horde of horses to blanket the landscapes these supercars roar past. Inside, their occupants are unseeing of anything they are speeding by beyond the tachometer and speedometer. My point is not to disparage the supercar or its owners, but rather to ask you this question: is a horsepower equivalent to a horse?
“No one is suggesting that, and they never were,” you say. “A horsepower is just a measure of power, the rate at which work is done.” All right. Is this a more fair comparison: would you say a car is equivalent to a horse? Before you jump to answer, recall that it take a combination of a car and a human to drive a car off a ferry landing and into a lake, something no horse would do. You could say this is the fault of the human blindly following the GPS, but the car still ends up submerged. If a person sought to perform a similar stunt on their horse, the person is the one likely to end up spluttering wet.
“I know you said we were going on a journey,” you say “but is it one that will take all day? Can we take a rest stop and stretch our legs?” I am reminded of what young kids the world over say to their parents moments after getting in the car, “Are we there yet?” And I will give the universal parents’ response, “Just a little bit further.” In place of our metaphorical rest stop and breath of fresh air let me just say that I am now preparing to compare the brain with the computer, our minds with the machine. I will even go so far as to give you the conclusion up front.
It is that today, and for the foreseeable future I think, wisdom is a uniquely human attribute. Wisdom consists not just in making a correct decision, but in making it with understanding of the reason why. Humans make wise decisions out of more than short-sighted self-interest because they can anticipate consequences, both to themselves and others. A computer algorithm can make the same decision, indeed it cannot make any other, but there is no concern with consequences.
We talk of cold computer logic explicitly to distinguish it from humankind’s reason, which is the result of overcoming our mess of constantly churning emotions. Though the resulting decision may be identical, the process by which we arrive at it couldn’t be more different. Some of the greatest advancements in artificial intelligence research came when we stopped trying to build algorithms to solve defined problems in their entirety and instead designed simpler systems patterned on human neural networks. This set the stage for deep learning and machine learning, where processing takes place by layers, each adding a piece of the puzzle and allowing higher levels of abstraction to emerge as a result.
In the case of both human infants and our fledgling machine systems, learning is the result of being confronted with data. Infants are amazing perception machines, taking in a flood of information from all their senses. This creates connections and associations among neurons in the brain, which are strengthened or weakened by further experiences. The computer systems need to be force-fed their data like stubborn babies, and they need truly vast quantities of it to start drawing abstractions that a baby can do with ease.
It is when we observe how the two systems develop and learn from their steady diet of data that we come to the great distinction between humans and computers. Computers do not lack for judgment. We humans build the rules in, or the rules for how rules will be learned, such that with a given input, the result will be the same. What computers lack is perception. They do not sense the external world as we do, neither in quantity nor in quality. In fact for any inputs to be made workable to the computer, they must be translated into the sterile binary code of zeros and ones. Humans do not lack for perception. It is the judgment that we must learn through painful experience and repeated trial and error.
Because computers judge consistently but are limited in their perception, while humans have vast perception of situations that we judge inconsistently, we remain far apart. And critically, when humans make judgments it is the result of two things taken together: all of our past experiences and our anticipation of future results. I have no doubt that we will steadily add to the situations in which computers outperform humans in tasks that appear to reflect perception, thinking, and judgment. But until a computer takes the leap of making a decision not just because of past experience but because it desires a future outcome, I expect no wisdom to come from it.