Translating languages is hard. If you’re fluent in more than you, already know that there’s never any direct translation for a lot of phrases. Hell, I’ll still use spanglish or other random blends to make sure I get the exact tone I’m looking for (And if my rather obvious humblebrag was a little off-putting, lo siento).
But this is a big problem that computer scientists have been working on for years. And it’s one that’s almost impossible to solve without one of humanity’s most recent (and terrifying) inventions: machine learning.
The thinking goes that machines are dumb, and for them to do cool shit, like translating, driving a car, or winning Go, we need to make them smart. And the only way we know how to do that efficiently is to teach them how to learn — just like humans. If you had to sit and explain everything you ever learned to a computer, it would probably take hundreds of years. Yeah, *YOU* learned it in less, but you had a couple things going for you. Your brain, for example, has a somewhat intuitive understanding of physics. It can watch a ball being thrown a few times and then use past experience to guess at where the ball will be at any given point so you can catch it. Without machine learning, you’d need a much more complex system to train a computer how to do even the most basic tasks.
Once you’ve started teaching computers how to learn though, their abilities explode. Computers, rather unlike humans, can be in hundreds or thousands of places at once. They can be almost anywhere. And they can communicate directly through the internet. Every one of Google’s self-driving cars now has all the information that the entire fleet has collected. Collectively, a few vehicles now has more on-road experience than hundreds of thousands of human lifetimes. THAT’S the real power of machine learning. Machines can and already are learning better or faster than we are. And that shows no signs of slowing any time soon.
This November, Google began applying the power of machine learning to one of its useful, though most often maligned products — Google Translate. Almost overnight the tool was spitting out text that was almost indistinguishable from professional translations, much to the surprise of real-world experts in human-machine interaction. Jun Reikimoto, a professor at the University of Tokyo was preparing a lecture when he heard about it. Reikimoto ran some sample text from a translated version of the Great Gatsby to compare with the original (so it’s worth noting that Google translate’s version went from English to Japanese and then back for this comparison) and then came up with two, practically indistinguishable clips.
A New York Times Magazine article, dubbed “The Great AI Awakening,” noted that aside from a missing article, the clips were basically indistinguishable. And if you’ve ever used google translate, that should come as a surprise. But, shortly thereafter, Google held an event in London to showcase the new-and-improved feature. There were fortune cookies, printed in any of several languages, and trays of food and drinks with names and flavors listed in everything from German to Portugese. There, Google announced that after just a few scant months of work, they’d created an artificial neural network of computers that could read about as well as any human.
This is the latest step in the techno-capitalist arms race between companies like Google, Facebook, and Apple for true, general Artificial Intelligence. That is, AI that can reasonably interpret and act with some level of agency — as any human might. A general AI could adapt, understanding the implicit and blurring the lines between people and their machines.
The progress of computing has been exponential for as long as anyone can remember, thanks to a rule of thumb known as Moore’s law, we’ve largely seen computer processing power double every few years. It’s what’s given us phones that are hundreds of millions of times more powerful than the computers that put Neil Armstrong and Buzz Aldrin on the Moon. But it’s also, somewhat famously, stalled in recent years. And instead, the growth of computing power has come from another source — the cloud. While it was certainly a bit of an overblown marketing tool just a few years ago, it’s not transformed how we use technology.
Amazon’s Alexa and Google’s Home are prime examples of the first steps. Google Translate and Deepmind are a couple more. But as more and more of these systems leverage the tremendous power of millions of computers working and learning together, all at once, we’ve crossed one of the biggest hurdles to AI. We don’t really need to match the processing might of the human brain in a PC or a phone. Not when we can tap the power of all of Google to function and learn in the same way.
And to be clear, we’re still a ways off from genuine AI, but with every major step we take, we accelerate the rate at which we’re approaching what may well be humanity’s final goal post. We keep seeing new barriers we thought would never be broken — like a computer beating a human player in Go — fall. And, as Google applies its tech to more and more, it’s never been clearer that our electronic children are growing up quite quickly indeed.