Saturday, January 14, 2017


Alexa, The World's First Artificial Stupidity

A six year old girl in Dallas asked the Amazon Artificial Intelligence Alexa to buy her a dollhouse and some cookies. Apparently, there was no fail-safe set to prevent such things, even though the settings exist in the device. Yes, because of a parental oversight, one of the most popular AI's on the market today was confounded by a six year old girl. Not high marks for AI in the 21st century. The parents were puzzled when an expensive doll house and 4 pounds of cookies showed up a couple days later. Eventually they questioned the girl, and she confessed to sharing her desires with Alexa. It is unknown if the girl is grounded or what at this time. (read the story here:)

Then, to add insult to injury, a news reader in San Diego apparently set off an avalanche of doll house and cookie orders by merely reading the story on the air. It is unknown at this time exactly how many Alexa-equipped households were ordering dollhouses and cookies. Maybe Amazon should have better setup instructions? Way to boost the economy, eh? So the six year old's parents weren't the only stupid people in the world, apparently, there's a whole pile of them in San Diego! (read the story here:)

Not an auspicious beginning for AI. The problem appears to be, the designers keep trying to make it more user-friendly, and easy to use, and wind up making it unusable. Now some sort of security codes or passwords or something have to be implemented to prevent the thing from just randomly picking up words off the TV or the Radio and acting on them. The problem in past was in the voice recognition, it had to be "trained" to identify a users voice, vocabulary was limited, and complex commands like "Alexa, buy cookies," were a little beyond it's capabilities. With the technological obstacles removed, it now becomes a problem of common sense. Was no one on the design staff able to foresee this might become an issue? If I tell Alexa that I really want a Rolls Royce, will it actually attempt to have one delivered? How do you draw a line that it cannot cross without human intervention, without rendering it so inconvenient that you're better off just using your computer or smart phone? I don't believe these challenges to be insurmountable, but we're going to have to do a lot better than this moving forward. AI is here to stay, how do we integrate into our already tech-heavy lives without letting it take over? Perhaps "AI" is too strong a term, the thing isn't really intelligent, it's basically a voice-activated personal assistant. Strangely, the hardest part has been getting the voice activated interface to work reliably. Now that we seem to have that obstacle removed, maybe we should be trying to make the thing more intelligent.

Comments: Post a Comment

Subscribe to Post Comments [Atom]

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to Posts [Atom]