“But when they are connected in the right ways, groups of people and computers together can often get closer to perfect intelligence than either could alone.” Malone, who is the Patrick J. McGovern Professor of Management at the MIT Sloan School of Management and the founding director of the MIT Center for Collective Intelligence, explores the potential of such connections in his new book, Superminds: The Surprising Power of People and Computers Thinking Together, from which this excerpt is taken.
Will general AI be a form of collective intelligence?
We know that the human brain is itself a form of collective intelligence. It is made up of a group of billions of individual neurons that—when working as a group—act in ways that seem intelligent.
Perhaps one of the best ways to create a real general AI, therefore, is to create a collective intelligence that combines, inside a single system, many different kinds of artificial intelligence. In fact, Marvin Minsky, one of the fathers of AI, suggested as much in his writings about a “society of mind.” In Minsky’s view, a society of mind emerges from the interactions of many smaller “agents,” none of which is very intelligent as an individual but all of which, together, create an overall system that is intelligent.
A hint of what this might look like comes from IBM’s Watson system. When Watson plays Jeopardy, the system makes use of thousands of smaller agents, many of which work in parallel on different processors. Each of these agents is more complex than a single human neuron, but none of them alone is nearly smart enough to be a competitive Jeopardy player.
For instance, one of the questions Watson answered was “President under whom the US gave full recognition to Communist China.” To answer this question, some of Watson’s agents went to work proposing the names of US presidents as possible answers. Other agents started looking in encyclopedias and similar resources for information about “US,” “recognition,” and “Communist China.” Using a special encoding of this reference information together with common-sense knowledge, these agents proposed more answers to the question, probably including the names of Chinese and American officials involved in the announcement. Eventually, based on many different agents evaluating many different kinds of evidence, and each “voting” for the answers it thought were most plausible, Watson’s society of agents concluded that the answer with the highest confidence level was “Jimmy Carter,” which was, in fact, correct.
How can AI help make groups smarter?
In the long run, it seems to me very likely that—whenever it happens—real general AI will include something like Minsky’s society of mind: a combination of many different specialized forms of reasoning and intelligence that, together, produce general intelligence.
But what can we do in the meantime? Here’s the surprisingly important idea that many people still don’t really appreciate: Long before we have general AI, we can create more and more collectively intelligent systems by building societies of mind that include both human and machine agents. In other words, instead of having computer agents like those in Watson try to solve a whole problem by themselves, we can create cyber-human systems where human and machine agents work together on the same problem. In some cases, the human agents may not even know—or care—whether they are interacting with another human or another machine.
In this way, humans can supply the general intelligence and other specialized skills that machines don’t have. The machines can supply the knowledge and other specialized capabilities that people don’t have. And the groups of people and computers together can act more intelligently than any person, group, or computer has ever done before.
How is this different from the ways people are thinking about AI today? Many people assume that computers will eventually do most things by themselves and that we should put “humans in the loop” in situations where they’re still needed. But I think it’s more useful to realize that most things now are done by groups of people, and we should put computers into these groups in situations where that’s helpful. In other words, we should move from thinking about putting humans in the loop to putting computers in the group.
Excerpted from Superminds: The Surprising Power of People and Computers Thinking Together. Copyright © 2018 by Thomas W. Malone. Used with permission of Little, Brown and Company, New York. All rights reserved.