The technical intelligence of supercomputers like IBM’s Jeopardy champ Watson may one day surpass that of humans, but they will never have this.
“This entity is able to understand natural-language questions and identify the correct answer on any topic.”
If we were playing Jeopardy! two weeks ago and the above “answer” appeared in a category called, let’s say, “Philosophy,” the correct “question” would have been, “What is a human?” But that all changed when the newest IBM supercomputer, Watson, appeared on the game show last week and handily beat the reigning champions. Just like Deep Blue’s victory over Kasparov in chess, this is a significant milestone in artificial intelligence. Watson is the first computer that can decipher naturally phrased questions and produce, with statistically impressive accuracy, the correct answer.
Unlike traditional search engines such as Google, which simply match keywords and use algorithms to guess which result is of the highest quality, Watson has to decipher the meaning of the statement in order to start analyzing potential answers. Words like “bat” and “bank,” which have multiple definitions – a bat, for example, can be a flying creepy mammal or something you swing at a flying creepy mammal – require an understanding of context in order to start searching for the right answer.
Watson’s ability to answer open questions lies in its natural-language understanding and the ability to confront the inherent ambiguity that the human mind naturally deals with at a level below conscious perception. Its success has raised questions about the assumed supremacy of human intellect and reignited latent fears of the looming singularity – that point in the future when human intelligence is trumped by a technology driven super-intelligence.
Much of the popular analysis of Watson (here, for example) focuses on its technical limitations compared with humans – for instance, the amount of energy required to power the computers, and the sheer size of the computers needed to run the programs in contrast to the human mind, which is compact and generates its own energy.
But these limitations form a weak argument for why the human mind is better than Watson. Based on Moore’s law, which accounts for a doubling in computer processing efficiency every two years, technical features such as power and processor size of the human mind will stay ahead of Watson only for a finite period. It’s only a matter of time before Watson is as efficient in processing power as our brains are.
These purely technical perspectives are, however, missing a fundamental point. No matter the natural-language processing power Watson has and how much data it can crunch in milliseconds, it will never be able to create meaning from knowledge.
IBM has put together some interesting promotional videos about Watson’s development. In these videos, engineers explain, in laymen’s terms, how Watson was created and the magnitude of that achievement, as well as, most interestingly, Watson’s evolution and its early failures. In footage of early trials, the room is packed with engineers on the edge of their seat, hanging off of every synthesized syllable emerging from Watson’s speakers. Back then, Watson got a lot of answers wrong. The engineers were crestfallen; the men (yes, they all happened to be men) who had poured years into this machine were frustrated and upset.
As the engineers struggled to get the algorithms right, and more recently as they watched with pure elation as Watson won Jeopardy!, their faces captured the true power of the human mind – the meaning and feeling and emotion associated with the technical accomplishment, that only a human can experience and express.
In one of the IBM commercials featuring Watson, Groucho Marx makes a cameo, delivering his famous line: “One morning, I shot an elephant in my pajamas.” Then he drops the punch line: “How he got in my pajamas, I don’t know.” People find that funny because it is ridiculous to imagine an elephant in Marx’s pajamas, but there’s so much ambiguity in this statement that it would be nearly impossible for most computers to decipher.
Watson is able to do it. Watson knows how to disambiguate. But Watson is lacking something essential, the ingredient that will forever stave off the singularity and that ensures the human mind, and spirit, will stay in the top spot even if we lose at Jeopardy!. What Watson doesn’t have is the capacity to understand or express the emotional meaning of the statement. It has no empathy for the fear or sadness that may have gone into shooting an elephant one bright sunny morning.
In the near future, Watson will most likely graduate from game shows to hospitals, where it will help doctors answer challenging and ambiguous questions. But Watson will never be able to deliver the news to a patient. Its technical intelligence and its capacity to decipher problems and acquire facts may one day surpass our capabilities, but Watson will never have emotional intelligence. That will always be the realm of the human mind.