Algorithms don’t have a sense of humor… yet

A good way to calm people who get into a panic when the topic of self-developing artificial intelligence technologies comes up from now on will be to tell them what happened to Tesla Motors NASDAQ share value on April 1.

In the spirit of April Fool’s, the company published a story on its corporate blog and informed the world via Twitter that it was launching the new Model W. The story explained that the W stood for watch, and was accompanied by a photograph of a watch out of which an enormous Big Ben (or to be entirely correct, the Elizabeth Tower) emerged, accompanied by a text making fun of the upcoming launch of the Apple Watch, and claiming: “Studies have shown the Model W will dramatically improve your health. If you work out. And it’s available in platinum!”

Not content with this, the company also announced via its YouTube and Twitter accounts that its cars would now include ticket-avoidance mode”, whereby when a law-enforcement officer of parking warden approaches the vehicle, the windshield wipers would come on and the care will drive slowly round the block “until the ticket issuing threat is no longer detected.”

So far so good. A bit of light humor that can do the company’s reputation as a youthful, innovative outfit any harm. But what happens when the algorithms of NASDAQ get in on the act? A huge amount of trading these days is done automatically via algorithms that are able to detect movements in the market and respond to them. They also detect news coverage of the company, and in this case, semantic analysis, identified the cause of the movements as “launch of new model W, and issued orders to buy huge amount of Tesla stock that briefly sent the company’s share price up by 1 percent, a company that is worth 24 billion dollars… just do the math!

In short, the automated algorithms fell for a joke that they were unable to understand because they had been designed by humans for other humans. If having a sense of humor is evidence of intelligence and this is what we can expect from state of the art machines and programs designed to manage the center of world trade, then we can probably sleep safely in our beds for a few more years before a Terminator from the future comes to kill Sarah Connor… :-)

(And no, this is not in any way proof that artificial intelligence is a long way off. But I did think it was funny)

(En español, aquí)