Nick Swardson’s “Gay Robot” learns how to cope with society. Via ComedyCentral.com
Growing Pains for AI Require Parental Guidance
In the ongoing advancement of Artificial Intelligence, there have been some rather hilarious events take place that seem to point to computers developing a sense of humor so to speak. And through the ability to look back on some of these events like a parent remembering raising a child and the associated growing pains, these screw-ups seem pretty funny and somewhat innocent.
Just like real juvenile delinquents that humans also produce, only different. So in light of some of the silly games that AI is learning to play I’ve decided to share a few with you this week.
Maybe we should be calling it ‘FAIcebook’ these days. Facebook has taken to AI in a big way. Just last year in 2017, in an effort to remake the selling of insurance, (Im not kidding), they turned loose two AI driven chatbots to learn how to trade virtual objects with each other. I imagine it went something like : Chatbot ‘Bob” meet Chatbot ‘Alice.” “How do you do Alice?” ” I do well. How do you do, Bob?” That went well for a little while. Success, thought the programmers.
But then. Bob-bot said something like :”I can can I I everything else” To which Alicebot replied ” Balls have zero to me to me to me to me to me to me to me to me to” Uh Oh! why the little scamps have gone and created a new way of conversing with each other in a kind of shorthand, that only they know the meaning of. ” Zoinks” said Zuckerberg’s crew chiefs and they pulled the plug. That’s the basic story from several news sources but probably not the only facts behind the story. The clever couple spelled their own doom by being a little too smart for the comfort of the humans who pulled the plug. And thank God they could pull the plug.
In my opinion, it is by without a doubt, the ultimate question any engineer MUST ask themselves while designing any AI: “How do I turn this thing off?”
In a survey of 100 Americans by 60 Minutes/Vanity Fair/April 2016, 53% said AI advancement was important, 20% said it was necessary, 15% thought it was dangerous, and a full 4% thought it would make God angry! GRRRRR
You probably heard of the tiff between Mark Zuckerberg and Elon Musk and their differences over the good and evil side of AI’s future. The Facebook Boss being on the side of letting AI have its bit and run free, so to speak. Musk, not so much. Well, the FAIcebook chatbot incident seemed to prove Musk’s point that AI really should have an ON/OFF switch. Ya think?
Some cops in Hamburg, Germany thought the switch was a good idea, when a neighborhood Alexa decided to rock the block at 2 a.m. while the homeowner was away. As the music blared at an extremely high volume, someone called the “polizie.” The cops broke down the door, unplugged the rebellious Alexa, and left the homeowner a bill.
In a more recent update, FAIcebook sent out notices to its users, stating that they are using, or might use, or could use, a facial recognition program. They provided an on/off switch for the time being. Let’s hope its better than the one the New Zealand Department of Internal Affairs uses. Their naughty Kiwi AI rejected a certain Mr. Richard Lee’s passport renewal because the machine said his eyes were closed in his photo. Mr. Lee is of Asian decent. Oops! Mr. Lee had the help of a human getting it straightened out.
Of course even heavyweight Google has had its AI problems. When Google introduced its “pixel recursive super resolution,” the program was given the task of taking someone’s poorly taken photos of someone’s face, and by matching up photos from its database, it could clean up the bad photo into a work of art. Uh, not quite. It was impressive how it worked, but to some, it was just horrifying what it produced. If you look at the included image below you get some idea why I believe On/Off switches should be regulated by an act of congress stating every computerated, AI updated, pixelated dang one of them must have one.
Other folks have been thinking about the proper way to raise the AI Child, and turn them into AI Adulthood. Check out what a writer for the Washington Post thinks about the future of parental guidance over AI. https://www.washingtonpost.com/news/theworldpost/wp/2018/02/27/ai/?utm_term=.16a692ced11f