Cutting Edge Computing

Algorithms Are Thoughtless December 30, 2014

I’m reminded of the movie I, Robot. Will Smith (as Del Spooner, but it certainly should have been Elijah Baley) distrusts advanced, human-like robots because he came to experience that while they make rational and largely correct decisions, on occasion the logically right, but humanly wrong determinations could be life or death under critical circumstances.

Of course, the late and lamented Isaac Asimov used the “I, Robot” science fiction series as a foil for logic puzzles surrounding his seemingly ironclad “Three Laws of Robotics.” Still, he was not unaware of the potential for wrong when the logic is right.

I came across this posting recently, commenting on a Facebook app that helps users build their “Year in Review.” Regrettably, this past year was not a bed of roses for this person, and the Facebook app selected images that brought back those unpleasant times.

Further, the Year in Review keeps rotating back to the person while in Facebook, serving as an unwanted reminder of events that don’t need reminders. The author called this “inadvertent algorithmic cruelty” (I wonder if there could also be “intentional algorithmic cruelty?”). Apparently, Facebook made the assumption that only good and memorable information is posted. In most cases, that may be true, but it wasn’t a particularly good assumption.

There seems to be a question surrounding whether the current focus of artificial intelligence (AI) is bad for humanity. AI is, and will continue to be, based on mathematical algorithms that assign probability and make determinations that guide decisions.

Certainly, any abstract intellectual, physical, or emotional responses in robots (however you might define that term) would result in an increasingly boring world. Knowing the algorithm means knowing the response in advance. While Asimov’s logic puzzles toyed with a number of logical holes in these laws, for the most part real life experiences are unlikely to be so unpredictable. His were edge cases.

In reality, it will be a long, long time, if ever, before we see anything resembling Asimov’s positronic brains encoded with the Three Laws of Robotics. Robots, and AI in general, will continue to become more complex and capable, and will continue to nibble around the edges of jobs that can be designed and executed according to rules.

But we won’t be replicating judgment in algorithms, because judgment is an individual characteristic. We make judgments in different ways, based on our backgrounds, experiences, age, and a myriad of other factors. The judgments and results will be different. That doesn’t make some judgments wrong, but it does make them human.