Technological Unemployment Makes the Jobs Apocalypse Worse
The employment apocalypse is getting worse tens of thousands of positions; including many well-paying middle class jobs, are vanishing before our eyes. Whether we like it or not, technology is driving this hollowing out of the workforce and it will soon get worse.
The latest casualties are at Bank of America (NYSE: BAC), which has announced plans to open 50 to 60 branches with no employees. To make matters worse, B of A plans to close 50 to 60 staffed locations, Reuters reported.
Bank branches without employees are nothing new. Automatic Teller Machines have been around since 1967, and night deposit drawers for far longer. What is new is a branch bank that consists only of technological interfaces. Such banks will soon become more common, JPMorgan Chase (NYSE: JPM) is testing next generations that allow customers to do all their banking on a tablet.
Not to be outdone Wells Fargo (NYSE: WFC) has set up an artificial intelligence team, Reuters reported. The purpose of this team will be to apply AI solutions to banking.
The Jobless Future of Banking
It is the internet that allows these institutions to dispense with so many employees. Currently, a person with a connected computer or smartphone can perform almost all bank transactions online.
The only banking one cannot do online is to make a cash deposit or withdrawal. Unfortunately, those transactions can easily be done at ATM or a supermarket cash register.
The apocalypse is already in full swing on Wall Street back in 2000 Goldman Sachs (NYSE: GS) employed 600 stock traders. The bank now employs just two equities traders and 9,000 computer scientists, The MIT Technology Review reported. Consumer banking is next, Goldman Sachs recently deployed Marcus a website where algorithms process and rtificial intelligence package consumer loans without any human involvement.
Eliminating employees like tellers also allows banks to get rid of other layers of employment including supervisors and human resources personnel. Nor is it just banks, Walmart (NYSE: WMT) has eliminated thousands of HR people.
Risk Management through Technological Unemployment
Organizations like Bank of America have several strong incentives to use the new digital tech to kill as many jobs as possible. The first is to reduce costs algorithms don’t draw salaries or receive benefits. Nor do they demand bonuses for performing certain feats.
Beyond costs there is the reduction in risk by switching to a digital process. Humans are more likely to make errors, steal, lie or deliberately slow the process to create more work for themselves. Banks are particularly vulnerable here because they have large amounts of cash.
Such risk reduction also motivated Walmart’s decision to eliminate 7,000 accounting positions last year. Many of those positions were cash counters and accounts receivable clerks, positions highly vulnerable to both fraud and theft.
There is also the problem of dishonesty as Wells Fargo discovered. It found that hundreds of employees; at dozens of branches, fraudulently signed thousands of people up for credit cards and other financial products they never ordered. The motivation for this horrendous behavior was to get bonus money.
Dishonesty is far from the only unethical behavior banks have to contend with. Humans can suffer from a wide variety of prejudices including racism, sexism, xenophobia and homophobia. Discrimination against borrowers has long been a major dilemma at banks.
Just as they do not steal or cheat, algorithms do not discriminate. A mortgage website that operates like Marcus is not likely to turn down a qualified African American applicant; because of skin color, and get the bank sued. Getting rid of human bankers and underwriters greatly reduces the risk of fraud and discrimination, which have historically been huge problems in the mortgage business.
Watch Out Machines have figured out how to Learn
Disturbingly worse is about to come because there are now machines that can learn many jobs. One of the first is Libratus; the poker-playing artificial intelligence that recently whipped four top professional card players at No-Limit Texas Hold Em.
What’s scary about Libratus is that it was not programmed to play the game like the poker bots that infest gaming websites. Instead its developers simply fed the AI the rules of Texas Hold Em and let Libratus learn the game by playing it.
“We give the AI a description of the game,” Libratus’s co-creator Carnegie Mellon University grad student Noam Brown told Wired. “We don’t tell it how to play. It develops a strategy completely independently from human play, and it can be very different from the way humans play the game.”
The truly frightening aspect of this story is that Libratus learned to play the game better than the human pros. It actually beat a team of four professional poker players in a three week tournament that involved 120,000 hands of poker.
Why Humans don’t stand a Chance against the Learning Machine
“Once you face Libratus, there’s nothing worse any human could ever do to you,” one of the beaten players Jason Les admitted to The Christian Science Monitor. “Every human is going to seem like a walk in the park.”
“It was an absolute beat down,” professional player Doug Polk; another of Libratus’s victims, said of the Brains vs. Artificial Intelligence Poker Tournament at the Rivers Casino in Pittsburgh in a Forbes interview. “The human team lost at a rate of approximately 2.5 times what the program lost at before.”
The best way to think of Libratus is as a learning machine. It employs three different methods of learning that humans use, reinforcement learning is old fashioned trial and error Libratus keeps doing a task over and over again until it gets it right.
That gives an AI a huge advantage because it can go over the process as many times as necessary until it learns. AIs don’t take bathroom breaks, sleep, stop for meals or go on vacation.
An algorithm called counterfactual regret minimization allows Libratus to perform a much wider number of plays than humans. A human player normally utilizes a few basic strategies while Libratus can use hundreds of different strategies. The poor human simply cannot keep up with that.
Finally there’s something called an end game solver which lets Libratus run the games over and over again as many times as it needs to. That means it can analyze a hand of poker hundreds or thousands of times until it figures what it did wrong. While human poker players were sleeping Libratus was busying analyzing their play with its algorithms.
How a Poker-Playing AI Threatens Banking and Finance Jobs
The frightening part is that the technology behind Libratus can be applied in a wide variety of banking, accounting and finance jobs. Its’ recent triumph demonstrates that is possible because poker is an imperfect information game.
Machines have long been able to play games like chess and Go because all the information is readily available. They’ve often failed at card games like poker because the amount of information available is limited. A player can only see his cards, he has to guess what his opponent is holding.
If Libratus can make decisions about poker hands it can make them in other imperfect information situations such as issuing loans. A loan application only contains part of the information about the borrower, such as his income and credit score.
Other factors such as future earning potential, actual cash flow or the viability of a business model might be unknown. A Libratus for loans would be able to over thousands of applications and past loans and identify patterns that may tell it those factors. It might be able to identify characteristics of individuals more likely to default or profitable customers the lender has been ignoring.
Beyond lending Libratus can be applied to many other finance jobs including stock trading, currency trading, risk management, financial analysis, market analysis, commodities trading, and hedging to name just a few. What happens when Libratus starts running a hedge fund or writing insurance policies?
The jobs apocalypse is about to get far worse because the machines have figured out how to learn. We had better get ready because AIs like Libratus will soon destroy hundreds of not thousands of highly paying jobs.
