The Problem Isn’t The Robots, It’s Us
Richard Feynman was a legend in scientific circles. One of the preeminent physicists of the 20th century—even other top minds considered him a magician—he is almost as well known for his jokes and pranks as he is for his groundbreaking discoveries.
When Feynman was a young scientist, Eugene Wigner compared him to Paul Dirac, a giant at the time well known for his autistic qualities, saying that “he’s a second Dirac, only human this time.” The quote is telling, not least because Wigner was Dirac’s brother-in-law.
While Dirac was clearly a genius, Feynman was truly transcendent. Although Feynman won the Nobel prize in physics, he was also a pioneer in nanotechnology and computing, did important work in virology and became an accomplished painter. Much like Feynman, as robots replace human jobs, we must learn to do them anew, only human this time.
What Is Intelligence?
Intelligence has always been hard to define. IQ tests have been around for at least a hundred years and they certainly measure something—scores have been shown to have a 30%-50% correlation with professional success—but fail to capture the whole picture. Feynman himself is said to have had an IQ score of 125. Good, but not exceptional.
If intelligence is an intangible quality in humans, then it is even harder to denote in computers. We know computers can do certain tasks, but at what point do can they really be considered to possess human intelligence? It’s a startlingly complicated question and it’s been debated for decades.
The most accepted answer comes from a 1950 paper by Alan Turing. He devised a simple test—called the Turing Test—which is devilishly simple. You just have a human judge converse with both machines and people and see if the judge can reliably tell the difference.
Alas, it turned out that the Turing test was relatively easy to pass. One program, called ELIZA, was able to fool people as early as 1966. Another, named PARRY stumped even experienced psychologists. Yet, those programs actually weren’t very useful for anything but passing the test. As it turns out imitating intelligence is not that hard.
The Infinite Monkey Theorem
Many believe that the real test of humanity is not logic or computation, but the arts. However it’s long been known that even great works, such as the the collected works of Shakespeare or Tolstoy’s War and Peace could be created mindlessly.
The concept, known as the infinite monkey theorem posits that infinite monkeys tapping away at infinite computers will eventually create not only Tolstoy and Shakespeare, but every other work human civilization has produced. In effect, with enough computing power, producing great works becomes a problem of curation rather than creation.
This, of course, is no longer theoretical. Computers already perform creative tasks like writing articles and composing music. In fact, they do it so well that even expert critics are sometimes fooled. Yet as Jaron Lanier argues in You Are Not a Gadget, this line of reasoning misses the point in much the same way the Turing test does.
Tolstoy and Shakespeare did not create because they were doing a job, but because they had a specific intention to relate human experience. Feynman was driven by similar motivations, which he made clear in his second memoir, The Pleasure of Finding Things Out.
We achieve greatness not through our ability to perform tasks, but through specific intent. It is our ability to imagine and dream that makes us special.
How Technology Fails
Computers do what they do not because they are motivated by experience, but because we design them to perform specific tasks in a specific way. Perhaps not surprisingly, given how incredibly powerful our computers have become, they excel at the jobs we give them. If fact, they often do them much better than humans do.
Yet they are not perfect and as Tim Harford has pointed out in the Financial Times, they often need humans to correct them. Computerized techniques like big data analysis are good at figuring out “what?” but not so good at discerning “why?” and the “why” is important.
If we tell a machine to find a series of correlations, it can instantly scour millions of data points and develop a working model. These models—Google flu trends is a great example—can be incredibly useful, but they do go awry. Correlation is not causality and at some point, we need to uncover causes to recognize and solve important problems.
Intelligence is vastly more than processing.
Don’t Be A Robot
Dirac became famous by solving a fairly obvious problem. (The Dirac equation essential reconciles Einsteins relativity with quantum mechanics). Yet Feynman excelled because he posed questions nobody else thought to ask. His genius was not only computation, but imagination.
For example, when he created the concept of nanotechnology at a physics conference in 1959, he didn’t use any complicated formulas— his entire speech could be read by an intelligent high schooler—but simply pointed out the possibilities of “room at the bottom”. He was also a great collaborator, which allowed him to explore new horizons.
Our problem today is not that we face a world of increasing automation, but that too many of us have grown accustomed to acting like robots, striving to perform rote tasks with efficiency and accuracy. We are educated to provide answers, not questions and when we enter professional life we are evaluated the same way.
Today, we carry smartphones with exponentially more processing power than Feynman and Dirac put together. We can choose robots to do a number of jobs more cheaply and efficiently than a human ever could. Yet they remain tools, means to an end rather than ends in themselves. Robots cannot live our lives for us.
That’s the challenge of freedom. At some point, you have to decide what to do with it or it is wasted.