During the “automation scare” in the 1950s, people were intrigued yet suspicious about the power of computers. Would they someday send us into permanent unemployment? Could robots eventually take over the world?
The original goal of artificial intelligence was to build a person out of silicone. Although scientists made quick progress in building computers and robots that amazed us, they never came close to actually replicating the human brain.
“Machine learning” was first introduced 50 years ago. This concept focuses on computers’ ability to “learn” not through human programming, but through experience and pattern identification. For example, machine learning allows a computer to become a masterful chess player by observing good and bad moves and learning from mistakes. While a computer looks at every possible move up to 20 to 40 moves ahead, humans use more conceptual skills to decide how to make moves.
In a recent PBS NOVA documentary “The Smartest Machine on Earth,” researchers explore powerful new tools in computing, like “Watson,” a supercomputer with a brain, or central processing unit, that can process 500 gigabytes, or the equivalent of a million books, per second. In 2009, Jeopardy producers came to IBM to size up Watson’s abilities.
You might remember when 74-win Jeopardy! champion Ken Jennings went head-to-CPU with Watson a few years back. During Watson’s “training” for the match, the scientists at IBM had to constantly expose Watson to large amounts of possible answers to questions so that it would have sufficient rules and logic to come up with correct answers.
Having studied thousands of questions, within a few milliseconds Watson analyzed every possible answer. It learned to make statistical judgments based on how pieces of evidence work together in the database of information scientists gave to it. In the end, the competition was close, but Watson pulled ahead and won the show.
Scientists who worked with Watson point out that there are two ways of building intelligence. We can either write down the recipe or let it grow by itself. It’s clear that we don’t know how to write down the recipe, according to scientists. Machine learning enables computers to grow their own intelligence.
Scientists continue to debate the ability of machines to truly displace us. There is so much that we know that we don’t even realize we know, such as the fact that ice is cold and sandpaper is rough. Our common sense knowledge seems too complex to program into a computer. Human intelligence is deeply rooted in language and emotion.
Without experience or emotion, can computers ever understand the world the way we do?
They don’t connect to human cognition on an emotional level, such as the way a symphony or a play can move us. Language and context are barriers for understanding. For example, figuring out the meaning of a sentence like “I shot an elephant wearing pajamas” is very difficult for computers. Was the shooter wearing pajamas or was the elephant? Was a camera or a gun used to do the shooting?
Of course machine learning goes far beyond winning trivia game shows. It’s driving a computing revolution. According to an article in Wired magazine , today artificial intelligence isn’t trying to re-create the brain. Instead, it relies on machine learning, massive data sets, sophisticated sensors, and clever algorithms to master discrete tasks. “In short, we are engaged in a permanent dance with machines, locked in an increasingly dependent embrace.”
Machine learning makes it possible to predict the weather days in advance. It lets companies like Amazon or Zappos suggest products for you based on what you’ve chosen before. It allows doctors to better diagnose medical conditions.
It’s even helping us communicate with people through speech recognition, which was once though impossible. Computers are now trained with millions of patterns of human speech, and the accuracy continues to improve. There are even apps for the iPad and iPhone that you can use to quickly record something and translate it on the spot when talking to a person who speaks another language.
IBM imagines a time when a computer will operate like the one in Star Trek—as information-seeking tools that communicates with us to ensure we get what we want. This thinking signals a shift in the way we use and accept computers in our lives, compared to the fear and suspicion we felt half a century ago.