Artificial intelligence is a pervasive part of modern life - used to predict corn yields and map disease outbreaks, among other applications. Sam Kean reviews "AIQ" by Nick Polson and James Scott.

Photo: Luke MacGregor / Bloomberg

Just a few years ago, the New York City government was getting 20,000 tips annually about landlords illegally converting building spaces into apartments - far too many to investigate. So the housing department turned to an unlikely source to help zero in on probable violations: artificial intelligence. After churning through millions of data points, a computer program suggested that inspectors look for two signs: spikes in utility bills and sanitation problems. Sure enough, inspectors who heeded these clues started catching five times more violators than before.

Anecdotes like this are Rorschach blots for our feelings about AI. Some people find them thrilling - pointing to a clever new tool for catching scofflaws. Others find them unnerving - should we blindly obey what the computers tell us to do, even if we don't understand why they're telling us to do it? As data scientists Nick Polson and James Scott reveal in "AIQ: How People and Machines Are Smarter Together," both sides are right.

Their book has two goals. First, to document the reach of AI in everyday life: It's now used to predict corn yields, track gender bias in films, map disease outbreaks and sort vegetables, among hundreds of other applications. But "AIQ" also aims to demystify AI. Each chapter focuses on an algorithm in a different industry, tracing the roots of each back to an innovator from before the computer era. One chapter goes back to Isaac Newton's stint at the Royal Mint, another to Florence Nightingale's medical reforms in Crimea, and so on. Each historical figure had a problem to solve and developed a novel solution. And luckily for later generations, the solution proved useful in fields far beyond the original.

The authors discuss, for instance, one of the problems that bedeviled Newton at the Mint: "anomaly detection" - the extreme variation in the weight and amount of silver used in each coin. For quality control, rather than weigh coins individually, the Mint would set aside a few thousand at random and weigh them all together - a process called the Trial of the Pyx. Unfortunately, Newton didn't quite understand the mathematics of interpreting these results (a rare mistake for him), and it fell to later mathematicians to make the Trial accurate. But once they did, it opened a bonanza for later researchers: The authors write that "averaging lots of measurements together is the most important idea in the history of data science," and can be used for everything from detecting credit-card fraud to hunting for radioactive bombs at ports.

Grounding AI in tried-and-true methods makes it seem less alien: Computers are simply faster ways to solve familiar problems. Hence the book's title, a portmanteau of AI and IQ - the point being that we need both.

'AIQ', by Nick Polson and James Scott;
St. Martin's, 262 pages, $26.99;
Photo: WSJ

Still, Messrs. Polson and Scott recognize that this can be an uneasy blend. How does Netflix "know," more accurately than my friends, what movies I'll enjoy? How does Facebook "recognize" my mug in photos? Such connections can seem all the more spooky because, in many cases, the programmers who designed the systems don't know the answer. As the authors note, "the role of the programmer [in AI] isn't to tell the algorithm what to do. It's to tell the algorithm how to train itself what to do." It's AI alchemy - the program gets the right answer, but no one can say why.

In response to this concern, the European Union recently implemented a regulation requiring all algorithms to be "explainable" by human engineers. This sounds reasonable, even necessary sometimes: In Florida's Broward County court system, AI sentencing programs can lead to longer terms for those flagged as likely to commit future offenses, based on demographic data. But because the algorithm judging them is secret, the accused can't really challenge its recommendations. The authors denounce this practice of "treat[ing] a sentencing algorithm in the same way you would treat a microwave oven, by just punching in some numbers and walking away."

Such algorithms cry out for explanation. In other cases, however, the EU policy could harm people. AI programs can already outperform dermatologists at scrutinizing pictures of moles and other blemishes and picking out skin cancer. We have only the faintest idea how they work. But if they're saving lives, who cares? Medicine needs an infusion of AI more than any other field, Messrs. Polson and Scott argue. The human body is extraordinarily complicated, and we're probably missing tons of genuine connections between diseases and seemingly unrelated symptoms. Doctors also spend up to one-third of their time manually keying patient records into computers - a colossal waste of their skills.

In response, the authors propose "Dr. Alexa." This smart speaker would listen in on consultations, update patient files in real time and help sort through possible diagnoses. Back-end AI would then pool this data with those of other patients and tease out connections that no human would ever spot.

It sounds straightforward, and the technology to build this system exists today. But Dr. Alexa needs data to train herself, and the authors claim hospitals are "paranoid" about letting AI loose on patient records. Whatever the reason for this attitude, few AI scientists work in health care, a situation the authors call "a moral embarrassment. We live in an age ... when your propensity to click on an ad for dog food is analyzed on supercomputers ... [Yet] we still rely on numbers that Florence Nightingale could have crunched with pen and paper to quantify the risk that your kidneys will fail."

This passage highlights the moral core of "AIQ": "When it comes to the important decisions in life, we can and should combine artificial intelligence with human insight and human values." If ceding some decisions to artificial brains still seems scary - well, modern life is scary. Politicians and CEOs make terrible decisions all the time, sometimes willfully. AI will make mistakes, but is the status quo superior? After all, "biased and ill-informed decision-making algorithms," the authors note, "are no less pernicious just because they run on little grey cells rather than little silicon chips." Uncanny or not, AI is more intelligent than IQ sometimes.

Mr. Kean is the author, most recently, of "Caesar's Last Breath: Decoding the Secrets of the Air Around Us."

Important → This essay may offer insights into the various cognitive areas of perception, response, and "instinctual intuition" that are related to cognitive dimorphic judgement abilities. The brain is extraordinarily complex, and within its physiology, structures, and chemistries lie the various Personality Filter subroutines that could be explored scientifically for Sexually Dimorphic Tapestries by those moved to do so. ← Important