Part 4: The Cognitive Machine

In some ways, it’s not surprising that jobs are being lost to computers. Going back decades, we understood and anticipated that computers and robots would replace us on assembly lines and in completing menial office tasks. So, it’s not a shock that the fastest declining human skill in the workforce is “manual dexterity, endurance, and precision,” according to the World Economic Forum (2018). “Dexterous” professions include mechanics, physical laborers, and butchers, among others.

At Yotel, a hotel in midtown Manhattan, a robotic luggage concierge named Yobot lifts up to 500 pounds at a time and deftly moves, tracks, and stores an assortment of bags for guests. Guests still have the option of checking in with a human employee, but Yotel reports that the vast majority choose to use automated kiosks and Yobot instead. (Plus, this assuages any pressure to tip the bellhop.) I did the same when I stayed at the hotel in 2017.

We also know that computers are extremely adept at completing tasks that follow if-then-do patterns and solving problems with clear inputs and known end-states. Think of the structured conversation we used to have with a gate agent that an automated kiosk easily reproduces — “Where are you going? How many bags are you checking?” Similarly, much of our messaging with companies online is a conversation with a chatbot — not a human — that can communicate expeditiously and effectively within a structured environment.

Yet, we didn’t necessarily anticipate that computers might threaten the jobs of analysts, coordinators, and managers, positions that require a relatively high degree of cognitive ability. The algorithms that power computers are now so sophisticated that machines are increasingly less dependent upon human knowledge. More specifically, we have entered an era of “machine learning,” whereby computers learn by engaging in a series of trial-and-error processes and then perform self-analysis to determine how they can improve. In machine learning, artificial intelligence is leveraged to find and apply data set patterns in order to teach itself. New York Times reporter Craig S. Smith likens machine learning to child development: “When a mother points to a dog and tells her baby, ‘Look at the doggy,’ the child learns what to call the furry four-legged friends. That is supervised learning. But when that baby stands and stumbles, again and again, until she can walk, that is something else” (2020). That is what computers are doing now.

The power and growth of machine learning are mesmerizing. For example, technologists were shocked when Google’s artificially intelligent AlphaGo defeated one of the world’s top Go players in March of 2016. For context, Go is a popular game in East Asia where players take turns placing their white or dark stones onto a grid, and if one player surrounds a collection of the other stones, they capture those stones. Unlike chess, Go becomes more complicated as you play because stones are constantly being added to the board to the point where it can be difficult to know who is winning. The game is over when all the stones have been placed, and the winner is the player who has captured the most stones.

Lee Sedol, eighteen-time international champion, participated in a five-game showdown with AlphaGo. AlphaGo won four of the five matches, but its most striking move was number 37 early in the second game. AlphaGo made an unusual move regarded as “suboptimal” by judges, and onlookers initially thought it had made an error. Sedol spent an unusually long time considering his next move — he seemed shocked to encounter such a nonsensical move. Eventually, however, the players found themselves creeping toward the black stone of move 37, which ultimately gave AlphaGo the edge and allowed it to clock its second consecutive second win.

Fan Hui, European Go champion and judge of the match, was aghast. “It’s not a human move. I’ve never seen a human play this move,” he said after the match. Since humans have played millions of games recorded and posted online, AlphaGo was able to assign probabilities to all recorded moves and learn to play better. AlphaGo can now play by itself, learning from each new game and generating huge volumes of game data.

Humans simply cannot do this — our brains cannot accommodate the sheer amount of information, calculations, and exactitude that a computer can. Yet, despite the enormous implications of machine learning, much of the public seems to think of computer advancement simply in terms of routine manual and cognitive performances. According to the PEW Research Center, 77% of U.S. adults believe it is very or somewhat likely that fast-food workers performing routine manual skills will be replaced by robots or computers in their lifetimes, and 65% of U.S. adults believe that insurance claims processors performing routine cognitive skills will be replaced as well. Interestingly, only 30% of Americans believe that their own jobs would become automated in their lifetimes. “Not me!” they say.

But researchers believe otherwise. Computers are now usurping non-routine “cognitively demanding” skills, as well, including those that require us to manipulate new information.

Consider this partial list of “best jobs” for a communications professional and their estimated salaries from the Bureau of Labor Statistics (BLS) for 2019:

• Meeting/Event Planner — Makes announcements, composes press releases, writes descriptions and biographies for event literature, and creates online content about meetings with an average salary of $54,880.
• Business Reporter — Analyzes developments within businesses, industry, and the economy in general for websites, television stations, newspapers, and magazines in a language understandable by the general public with an average salary of $62, 400.
• Marketing Manager -Analyzes consumer reactions to a brand’s products based on factors such as price, consumer experience, packaging, and accessibility with an average salary of $149,200.

Communicating. Analyzing. Advising. Managing. Aren’t these “human” skills? When you think of a “communicator,” do you think of a robot? Yet, AI can perform all these skills. AI can make announcements, format press releases, create content, analyze stats, perform calculations — all with the click of a button.

All this is stunning, but how effective is machine learning at managing and retaining human employees? According to Harvard Business Review, machines can be highly effective sales managers when given sufficient data. Companies that use artificial intelligence to establish key performance indicators and set strong targets “can become higher performers with larger sales, better margins, and a more motivated salesforce” (Chung, Huber, Murthy, Sunku & Weber, 2019.) Researchers found that the use of analytics to set sales goals could even help to retain top sellers for longer by maintaining a challenge.

While a sales manager might use an employee’s past performance to set an ultimately unrealistic target, algorithms are more successful at setting appropriate sales targets because it is “using advanced analytics to identify the true drivers of business outcomes and are applying big data and machine learning to understand customer demand at an unprecedented level of accuracy and granularity” (Chung et al., 2019). Top salespeople may be less likely to leave the company altogether if they perceive that their sales quota is attainable (and not the subjunctive whim of an administrator).

And, yet, there remain opportunities for human managers, because creativity and social-emotional skills are very difficult, if not impossible, for machines to attain. So, a human sales manager might review an AI-produced sales target, but leverage her social and emotional skills to finesse a sales quota for a salesperson undergoing a difficult period in their personal life, such as the death of a loved one. In other words, the manager might anticipate that the salesperson will bring in fewer sales during this period and thus adjust the machine-generated quota downward, but expect higher performance in three months. As a result, the manager leverages her humanity to create a more viable sales-quota strategy going forward.

Cooperative partnerships with machines are happening across many professions. Take, for example, radiology. Physicist and machine learning researcher Max Tegmark writes, “if you go into medicine, don’t want to be a radiologist who analyzes the medical images and gets replaced by IBM’s Watson, but the doctor who orders the Radiology analysis, discusses the results with a patient and decides on the treatment plan” (2017, p.123). Humans are able to look at situations holistically and creatively by taking social and psychological considerations into account.

In another case, Neuroscience News reports that an app called Coughvid can detect coronavirus with 70% accuracy from the sound of a cough (2020). A pulmonologist who simply administers tests is much less valuable than the one who can provide a diagnosis and individualized medical advice with the help of a machine. Curious problem solvers will be able to add value and distinguish themselves from machines by offering enduring creativity.

The World Economic Forum predicts that by 2025, robots will handle 52 percent of current work tasks, almost twice as many as now, and that by 2022, “even work tasks overwhelmingly performed by humans today — communicating, interacting, coordinating, managing and advising — will begin to be taken on by machines” (2018). These are not skills we typically associate with robots, but they could soon recalibrate our view of long-standing workforce positions. And it should recalibrate our view of how K-12 instruction is formulated.

Clearly, there is mounting evidence that machine learning will soon affect every field, discipline, and worker in the world, no matter what rung on the corporate ladder they find themselves on. The challenge for educators, then, is to nurture in students the skills they need to compete and collaborate in this computerized world. If we focus our teaching on rote and highly structured information, we are essentially teaching students to perform like computers (which is a competition humans cannot win). What we must do is teach them to work with computers — to leverage their creativity and social-emotional skills to add value to the information machines produce.