As we all know, letting computers do everything for us will cause them to one day take over the earth and kill us all, or if we are lucky, just make batteries out of us. But some foolish, uneducated people out there have neither watched the Matrix nor Terminator movies and are working diligently toward making computers do people jobs. While someday these robots and programs will become our overlords, in their present crude state they are content to destroy us with their hilarious incompetence.
Here are some things that anyone with a lick of sense would not put in the hands of a computer, and the people who are putting them into the hands of a computer.
#5. Stock Trading
Trading stocks is a tedious, mass-number-crunching game. Why not make computers do it for you?
Because they will fuck it up.
Just a few months ago, on May 6, the Dow dropped about 1,000 points in minutes, which is usually about the kind of rate that kicks off something like the Great Depression. Stockbrokers barely had enough time to get their office windows open before the word spread that it was a false alarm. The cause? Robots.
"Mr. Raymond? It turns out it was just a computer glitch... oh."
Someone realized that human traders, no matter how sheeplike or pessimistic, couldn't have freaked out fast enough to drive the Dow down 1000 points in 10 minutes. Computers are just designed to do everything faster and more efficiently, including stock panics.
Stock trading algorithms, or "algos," as financial people fondly call them are responsible for 70 percent of trades today, which means that most of the stock market is literally just computers making deals with each other.
"Mr. Vaio? You wanted no cream, right?"
These programs trade in millionths of a second, millions of trades a day, and are even trained to prey on other algorithms by recognizing their patterns, stealing their stock picks, and selling them back to them. They even get names, like, "Dagger," "Barcode" or "Crystal Triangle," that apparently come from the Drug Dealer Brand Name Generator.
Malfunctioning algos have shut down the London Stock Exchange and briefly paralyzed the New York Stock Exchange among others. In one incident, a guy not double-clicking fast enough triggered a bug in the program.
When you're controlling billions of dollars, you don't have time to use an antiquated semi-automatic mouse.
Instead of slowing down to work out the bugs, finance companies keep rolling out bigger and more aggressive algorithms. The bank that crashed the NYSE didn't even know they'd done it until they got a (probably very angry) phone call the next day.
Some firms have gone even further and put the algos in charge. At Rebellion Research, computers and humans have actually switched roles. The program (named "Star") is the boss, analyzing over a decade of stock market data and coming up with a long term strategy, giving humans a list of buy/sell instructions every morning. The humans then make the trades at the computer's command, and are not allowed to modify them in any way.
It's not clear from the article whether the program is named after Star Jones pre or post surgery.
More freaky than the fact they're taking orders from a computer is (1) it's actually working - Star is beating the market and bought "defensively" before the stock market crash, and (2) they've started calling it a "he". The creators (servants?) of the program say things like, "He just loaded up on value stocks," and "I've learned not to question the AI." We will all be batteries soon enough.
First the stock market. Then this.
The economic collapse, a lot of people owe money these days, and that sure is a lot of people for banks to threaten every day. It's just not possible to hire that many threatening humans, so debt collection agencies have been using robocallers for years, which are about as inaccurate as mass mail (seriously, how many of you ladies have gotten penis enlargement ads).
"Would... I like... to... increase... the... size of... my..."
I have had my cell phone for three years and I am STILL getting debt collection calls for someone named Annette X (or Ax?) who apparently owned the number before me. Annette, if you are reading this, call the fucking bank already.
Since people like Annette apparently change their numbers and ignore these calls all the time, debt collectors have been moving up to the next step - lawsuits. Previously they didn't want to go here because it was insane to pay a lawyer to collect someone's piddling $250. But now? Robot lawyers.
Some terrible, terrible people have come up with computer software that allows one lawyer to easily file 5,700 lawsuits a year. You just feed the software a Social Security number, date of birth, address, and amount owed, and presto. It files all the paperwork and that person is sued. If they don't show up (and most don't), then jackpot! Free money.
But if they do show up, you may be fucked. One woman targeted by mistake showed up in court to prove she didn't even owe the debt. They wanted $9,000 from her. She got $8.1 million from them.
You'd think surely someone would be checking these records at some point, either at the bank before they sell it to the collection agency, or the collection agency before they hand it to the law firm, or the lawyers before they put it in their lawsuit robot. But no. The sworn affidavit, the piece of paper they need to show in court to take your money, is often signed by people who either don't read them, or in at least one case, by a person who is dead.
Despite being dead, Martha Kunkle managed to sign thousands of affidavits used to pursue thousands of debtors. Which seems like a good idea at first. If the image of your creditor returning from the dead as a restless payment-seeking banshee doesn't frighten you into paying your debts, what will? But alas, it's illegal.
It's getting harder to make money at news, and what with human reporters constantly fussing about "food" and "supporting a family," news outlets have been trying to figure out ways to cut them out of the equation.
It's easy to get a robot that eliminates reporters, it's harder to get one that takes their place.
You probably already know about Google News using some AI to collect headlines of the day, and you might have heard of AOL's plan to start writing stories based on search engine algorithmic analyses of what people want to read. That's small potatoes.
Google News and AOL News.
Statsheet wants to create a program to write entire sports blogs from basically scanning box scores, blogs that readers will think are written by a human. This article includes a sample. These automated blogs might someday be read by an algorithm like Infonic's or Reuters' which scans and analyzes hundreds of news articles a day to tell you what people think of different companies (in Infonic's case) or athletes, or political issues, or anything you don't want to read about yourself.
With computers doing all the writing and reading, I guess we will just go play video games or something.
But we'll probably want to tune in, because computers writing news can be pretty hilarious. Techmeme's 2007 experiment in algorithmically scanning the web to predict the next big headlines before they happen resulted in them putting out front page headlines about Anna Nicole Smith's hospitalization when she had in fact been declared dead.
That and other failures prompted them to stop being so idealistic about AI and hire an editor.
Making even less headway is automated video journalism. Do you like Siskel and Ebert style movie reviews? Then you'll hate NewsAtSeven, an AI journalism project by the Intelligent Information Laboratory at Northwestern University that feeds algorithmically generated lines to animated anchors using the shittiest text-to-speech program they could find. The bantering beginning at 0:45 is cringe-inducingly amusing, a contradiction you can only fully understand by watching the video.
To be fair, Techmeme and the NewsAtSeven team are clearly aware of how far they still have to go. More bravely clueless about the failure of his automated journalism venture is Tom Costello, who already failed at making a search engine "better than Google" with Cuil, and has now made the remnants of that into the amazingly bizarre Cpedia. Cpedia is a sort of automated Wikipedia that gathers information from the web and brings it together into entries that make no fucking sense. Stung by criticism along those lines, Costello replied with a weird blog post about how he was beaten as a child when trying to speak proper Irish and that's why it's okay for Cpedia articles to make no sense. Touche, Tom.