A multilevel garage next to my office building has a narrow glass panel on one side of the staircase. Once I observed a sparrow that kept hitting the glass as if it was stuck inside the shaft. Flying out in any other direction would have instantly set it free, yet it appeared determined to reach the sky through the shortest path it had once set its eye on, undeterred by the unexpected and unyielding resistance. This observation underscored a seemingly inherent design limitation of our biological minds.
Even those of us who are not mathematically inclined routinely face a problem of “global optimization”. Imagine trying to reach the highest point on Earth by foot while blindfolded. If you set out to increase your elevation as much a possible with every step, you would quickly find yourself on a nearby hilltop. Having climbed one would have accomplished the local optimization (after all, any step would now be down), but you would probably be a long way from the highest planetary peak. And while marching downhill may feel counterproductive, it is occasionally necessary for achieving your global objective. (In fact, a family of common mathematical modeling techniques is called “hill climbing algorithms”.)
In our daily life, we may never have sufficient computational power to discover the optimal path. But as we are enjoying the scenery from the height of our molehill, are we always prepared to pause and recognize that Mount Everest is yet to be conquered? Needless to say, such recognition is much harder when the ground is constantly shifting under our feet. And it is still more daunting when the goal itself is a moving target.
I have long been intrigued by the seemingly unlimited willingness of programmable devices to step back and start over. The invariable equanimity of my old GPS as it was “recalculating the route” made me revere its invisible inhabitant. Chess playing programs have long been able to swiftly abandon their strategies once the situation on the board changed. AlphaZero, a new chess champion, is “self-taught and therefore unconstrained by conventional wisdom about the game”, which gives it an “ability to learn each game afresh”.
The last several generations have had no trouble acknowledging the superior computational power of electronic minds. Yet for most homo sapiens, creativity has remained one of the few inherently human assets inaccessible to blind algorithms. Perhaps, it is time for us to swallow our sentient pride and learn a thing or two from our artificial creations that may help us become better humans?…
Exhausted, the sparrow finally landed, perched itself on the metal frame, calmed its ruffled feathers, closed its eyes, and became still. I felt like giving the fellow creature a high five. Or whatever one does to express their admiration and encouragement to a bird.