Given my practice, it probably isn’t surprising that I hear data and technology related buzzwords all the time, and often they’re just strung together in an incoherent jumble. “Blockchain crypto machine learning AI alakazam.” Usually this jumble is followed by a statement like “it’s changing the world, man,” at which point I always know that someone is going to make a grab at my wallet, because using the word “blockchain” in a sentence immediately increases a company’s value. Don’t believe me? If I renamed my firm “Ward Blockchain LLC” I’d have investors lining up around the corner.
The problem is that we’re hearing these terms and not understanding what they really mean. It’s the opposite of what we expect machine learning to be, ironically. I’ll give you an example. I can vaguely recall the first time I heard the phrase “mayday.” It had to have been a Saturday afternoon showing of a movie about fighter pilots, maybe 30 Seconds Over Tokyo or something like that. I heard it, and without having any idea what it meant or where it came from, I understood it was what you said when things were really, no kidding, in bad shape. I had occasion to use it later that day, screaming “mayday” from the top branch of a tree that was a) forty feet above the ground, and b) not substantial enough to hold me.
I didn’t reflect on it in the moment (there were also bees, you see), but my ability to hear someone use the word “mayday” and immediately grasp its purpose was an instance of good data analysis, the kind that we expect machines to develop. Machine learning is, in many ways, a human impatiently waiting for an algorithm to understand context and nuance without the human having first incorporated the context or nuance into the algorithm.
The big problem with that approach is that it presumes a machine learns exactly like humans do. They don’t. Or rather, they don’t learn that way yet, because we haven’t taught them how to. You see, the trick with machine learning is understanding that it is primarily a question of inputs. To move to the next level, what we have to teach a machine to think like an adult human (they’re already good at thinking like teenagers.)
Consider my mayday story. If you taught a device like Alexa that if someone says “Alexa, Mayday!” they mean “Lock our Nest-locks and call the cops!” then that is a direct input that triggers a preordained response: nothing new. If you teach a device to be multilingual and hear “mayday” as it originally was, “m’aidez” (“help me” in French), that’s a step further, but still only an iterative step or two away from Google Translate. But to teach a device to observe human behavior in a film and, as I did, hear the word mayday and deduce that it means “I’m in trouble?” Now we’ve made the kind of quantum leap that makes futurists excited.
We’re not there yet. In fact, we’re not really all that close to being there yet, because we haven’t even defined what “there” looks like. You can play on neural networks and see for yourself that algorithms still return plenty of totally wrong answers, and if you ask Siri something she doesn’t understand, the answers are often literalistic or simply irrelevant. But recall that, just ten years ago, a voice activated digital assistant was the realm of science fiction, and that things move quickly.
My point (oh good, there is one) is that if we don’t want to make the same kind of mistakes that frustrate us when machines make them, we have to understand context. When you’re making plans for your business, you can safely assume that technology will change what you do. But you have to learn how it will change things and why. Being able to explain how machine learning, when paired with the right data sets, will help drive clients to your business because you’ve learned how to respond to the expectations of Google or Yelp’s algorithms so you’re not an outlier, and therefore, ignored? That has value, and power. Saying “machine learning is a powerful tool that will revolutionize our business” with nothing else added has no value at all, and likely means that you’re going to miss out on opportunities for growth. You might as well be saying “mayday.”