We talk a lot about the coming wave of artificially intelligent machines which will either drive us and clean our homes or kill us all off, but we seldom talk about how these things could keep us from doing stupid things likely clicking on or forwarding a fake news story, tweeting without thinking and ending our careers, or accidentally hurting our kids and friends. This was what made last week’s AI Day from Intel (News - Alert) so interesting for me. Yes, they had the typical comments on performance and smart robotic devices which catch everyone’s interest, but they also uniquely had a focus on AIs that could be used to make us better people. After the last few months of a nasty primary election here in the U.S., I’m pretty sure I know at least two people that could have used one of these things.
Let’s talk about how an AI could make you appear a ton smarter and less likely to hurt people you care about.
The Technology Problem
We all do things we later regret, ranging from shopping to tweeting under the influence, to responding in anger inappropriately, to forgetting that what sounds funny in our head may sound racist, sexist, or like grounds for termination to the folks that receive our poorly thought through email, instant message, social media post, or text. Sometimes just spending an extra few seconds thinking about what folks might actually think when they read what we wrote could save a friendship or a job and yet, I’m sure, every one of us has said or written something, likely a number of something’s, that we now wish we hadn’t.
This isn’t just an electronic issue either, as I’ve ended important relationships with letters that were ill thought through, and this was after I’d observed my father, who was particularly bad at this, destroy companies with his ill thought through correspondence. But, at least sometimes, because it took so long to write and then post a letter, there was at least the chance we’d think better of what we were doing and not actually mail the thing. Back then a lot of us wrote things that we never actually mailed because our slow brains actually kicked in before we took that final step of putting the letter in the mailbox. But technology made things faster; particularly with Twitter (News - Alert), where you can formulate and execute a comment often before you’ve even thought through what you are writing, let alone whether you should hit send.
So, technology, at least with regard to our doing stupid things, focuses first on speed, with the end result being that a lot of us are doing far more public stupid things. What we need is an alert system that warns us, before we hit send, of the implications of our action.
AI Guardian Angel
You know there are two modes for the AI being considered for autonomous cars; one is where you don’t drive at all and the car does all the work. The other is called “Guardian Angel,” and this is where you seem to be driving but, should you do something stupid or unsafe, the car instantly takes over and prevents it. A drunk driver, in that latter mode, would be like a child in a car seat with a steering wheel, completely disconnected from the car’s systems but still thinking they were driving and free to say “vroom, vroom” anytime they wanted.
This can be applied to social media and email. What Brian Krzanich, Intel’s CEO, seemed to be talking about is a system that would intercept your hasty post and first advise you not to send it, articulating the likely repercussions, and then actually help you formulate something else that made you look far smarter and would be more likely to accomplish the goal you intended rather than the one you did not, like getting fired.
This same AI could flag fake news before you clicked on it, or worse, before you forwarded it. This would work not by preventing you from seeing it, but by suggesting you take that extra step and make sure you either assured its accuracy or included the language “I found this interesting but, be aware, my AI says this may not be true” when you forward it to others, better protecting yourself. Like you, I have a number of relatives that apparently believe anything and I’ve given up chastising them about it, concluding, like I expect many on their distribution list, that they are incurable idiots. If an AI like this could make people that seem to relish in being idiotic look smarter, think what it could do to those of us who actually try to think before hitting send?
Over the next few years we’ll have AIs that translate what we say, with nuance in place, to any other language; AIs that can, with punctuation, do speech-to-text for us, speeding up our correspondence and decreasing dramatically our chances of looking stupid and ending our careers. It is somewhat comforting to know we will also have AIs that could be used to significantly offset, and perhaps eliminate, this risk so that instead of appearing insanely foolish we actually look smarter.
With so many warning that AIs will be the end of us, it is nice that firms like Intel are working to prevent that eventuality—even if it is only to make sure we don’t become our own worst enemy.