Cheaaaaat!
January 2, 2019
Big headlines the last few days about Artificial Intelligence becoming more human by the second. Stanford/Google researchers have determined that a machine was "hiding information" that would allow it to cheat at its assigned task.
It involves everyone's favorite Google toy, Google Maps, and the programming necessary to turn aerial views into street maps and then back again. To reduce the complicated scientific discussion to a simple statement for now, it didn't accurately detail the finer nuances from an original aerial view into the street map; thus when programmers asked it to go from the street map back to the aerial view, it included details that were only noticeable (for the most part) on the original aerial view.
I find this interesting for more than just the obvious reasons that suggest we can't trust the machines we are creating to help manage our world. That is fascinating stuff left to the Robert Heinleins and H.G. Wells of the world. I am more intrigued by the idea that our machines, like us humans, will look for short cuts to make a task less onerous. As the articles about this specific program point out, the AI reacted exactly as it was programmed--to find the most efficient way to complete a task.
At a time when educators bemoan the ease with which students choose shortcuts that amount to cheating in their eyes -- think Quizlet -- we are forgetting that humans inherently desire the shortcuts, the cliff notes, the psats, the plsats, the graphing calculators that get us to the end point more quickly. Machines have just become the most convenient way to achieve this efficiency. To then attack the very machine for seeking the same efficiency, well, that's like blaming Eve for taking Adam's rib.
Notice also how in the article about this AI's actions, we tend to want to shame it in the same way we might shame humans for their shortcuts: "The machine, not smart enough to do the actual job of converting these sophisticated image types to each other, found a way to cheat that humans are bad at detecting." The machine isn't smart or dumb, it simply reacts to the need--to create the most efficient way of producing the photos. In some ways that is most genius.
This all seems so applicable in a world where we told today's students, the infamously Gen Z kids, memorize where to find information more than they ever memorize the actual information. The world is full of short cuts, baby, so it only seems appropriate that I suggest how we can get from "teach" to "cheat" in three easy steps.
Step 1: To Teach starts with Transformation, but we use age-old methodology and miss that boat, so put Transformation at the end.
Step 2: Eacht now leaves the power of Computerization smack dab in the middle, but all praise the technology. Put it in front of the class. We've got the apps, the calculators, the websites to justify it.
Step 3: Ceaht doesn't Hype the process enough, so make sure hype follows the computerization. Why learn through blood, sweat and tears, when everywhere I look I see gaming cheats, cheat sheets, a cheat engine, even cheat days?
And thus you end up with a Cheat. We have turned a teachable moment into a cheat-able moment.
I honestly can't say if this blog is sincere or satirical. All I know is that I think I just put myself in the company of Bill Belicheck. I know of a few readers I have now lost simply by that association. That may leave 3 of you.
|