Mind Matters News and Analysis on Natural and Artificial Intelligence

Can Big Data Beat the Humans Who Compile It?

A computer pioneer bets no: Human intelligence augmented by artificial intelligence will always beat artificial intelligence alone. Is he right?
Share
Facebook
Twitter
googleplus Google+
arroba Email
Frederick Brooks
Fred Brooks

Computer pioneer Fred Brooks believed in the power of artificial intelligence but with an important qualification, which he outlined when he received the 1994 ACM Allen Newell Award: Intelligence amplifying (IA) systems can beat artificial intelligence (AI) systems “at any given level of available systems technology.” In other words, systems that augment our natural intelligence are more powerful than artificial intelligence alone:

Someday a computer may beat the world champion in chess. When that day comes, I should like to see the world champion equipped with a powerful and suitable IA chess tool, and then play against the AI system. I’ll bet on the IA team. Fred Brooks, “The Computer Scientist as Toolsmith II” at Acceptance Speech,

Just three years later, “that day” came: IBM’s Deep Blue beat reigning world chess champion Garry Kasparov. But, in the decade or two after Deep Blue’s victory, chess champions equipped with IA tools consistently beat the AI systems. So, Fred Brooks’ bet proved true both on “that day” and for the next twenty years.

But does the IA team always outperform a “given level of available” AI systems technology? No. And it probably didn’t surprise Brooks when in 2017 the AI engine Zor won the freestyle Ultimate Challenge chess tournament. This tournament allows human-computer teams to play both against each other and against computer-only systems. Zor’s victory led one commentator to declare in 2018 that “Computers are now superior to…man + computer teams.”

So did Fred Brooks’ famous inequality principle, IA>AI, hold true only in the short term? Let’s look at some other important IA systems.

Big Data is the premier example of an IA system. At the click of a mouse or the tap of a finger, billions of pieces of information that have been pre-searched, pre-sorted, and pre-indexed for us are available to augment our ability to make intelligent decisions. Do you want to know which university has the most influential professors and alumni? Would you like to know which neighborhood near your new workplace would be the best location in which to buy a home? General information in response to these and countless other questions is now at your fingertips, thanks to Big Data. That leaves you free to focus on questions only you can answer such as, in the case of a new home, does it matter if there is a public tennis court or a farmer’s market within walking distance?

Considering these trends, if Big Data progresses the way chess did, is it only a matter of time before humans are cut from the team? For example, will the Big Data about our purchasing trends be automatically fed into robot-operated factories that use genetic algorithms to make similar products, without any human intervention?

Before we envision scary scenes from The Matrix, we need to remember that Zor, Deep Blue, and any future AI systems are inevitably designed by humans. Developing an algorithm to beat humans or humans-plus-computers in chess requires a lot of other humans doing careful engineering. It was not done without human intelligence. Not only that but the program designed to play chess won’t suddenly become a champion at Monopoly—or even very good at checkers.

The bottom line is that Brooks’ Bet and his IA > AI inequality principle is a good reality check in the face of fears and hype about what AI will do in the future. AI is powerful and, when designed to mesh directly with the needs and intellect of humans, it becomes even more powerful. The best example of this to date is the Big Data revolution and the amazing augmentation of our intelligence that it provides.

See also: Study shows eating raisins causes plantar warts (Robert J. Marks)

and

Too Big to Fail Safe If artificial intelligence makes disastrous decisions from very complex calculations in health care, will we still understand what went wrong?

Jed Macosko

Prof. Jed Macosko holds a B.S. from MIT and a Ph.D. from UC Berkeley. He was a postdoctoral researcher under Prof. Carlos Bustamante, and also under Prof. David Keller. The Macosko-Holzwarth lab is currently focused on the mechanical properties of cancer. Prof. Macosko first collaborated with Dr. George Holzwarth in 2004 to explore how multiple motor proteins pull a single cargo in living cells. Prof. Macosko also partners with the lab of Profs. Keith Bonin and Martin Guthold to develop a new drug discovery platform and with Dr. A. Daniel Johnson of the Department of Biology to develop new teaching technologies.