The explosion in data – Big Data – has signaled a new era, an era in which businesses, politicians, the legal system and others are making critical and life-altering decisions based on temporary and misleading correlations discovered by computers.
In his new book, “The AI Delusion,” Gary Smith, Pomona’s Fletcher Jones Professor of Economics, lays out why trusting important decisions to computers and Big Data is a big mistake. "The AI Delusion" will be available on October 1, 2018.
“People think computers are really, really smart,” he says. “Our admiration of computers has convinced many that artificial intelligence (AI) can make smarter decisions than humans.”
Computers excel at well-defined tasks like matching pixels and calculating correlations, says Smith. “They can search for words, spell-check words and find definitions of words, but they don’t – in any meaningful sense – actually know what a word means.” A human perspective is still needed when an algorithm is not enough. “Computers have no common sense, they don’t have wisdom, they don’t have critical thinking and they can’t think up theories.”
Computer successes have led to the exploration of a myriad of applications for artificial intelligence. The French government will spend $1.85 billion over the next five years to support research in the field. In the U.S., companies rely on computers to screen job candidates, assess loan applications, and pick stocks. During the presidential campaigns, computers steer resources, TV ads and staff power based on Big Data analyses – for better or for worse.
This is worrisome when AI may be used, for example, for algorithmic criminology. Courts all over the U.S. use black-box computer algorithms to make bail, sentencing and parole decisions, even though no one knows the reason for these decisions.
Smith suggests caution, “Thus far, artificial intelligence is designed to perform narrowly defined tasks, and it does them really well; but computers have a lot of trouble handling tasks that require genuine knowledge of what you’re doing.”
Smith argues that artificial intelligence still lacks integrative thinking and has trouble deciphering meaning or interpreting patterns without context. In order to improve AI, researchers are studying how to get computers to think more like humans. Researching how children learn is one way that the industry hopes to improve AI in the future.
“At this point in the development of AI, we should be very skeptical of turning important decisions over to computers,” says Smith.
“Computers are not going to take over the world, because they don’t know what the world is. The real danger is that we think computers are smarter than us and so we trust them to make important decisions that they’re not qualified to make.”