Have Robots Developed Their Own Language?

(Last Updated On: October 7, 2019)

Have Robots Developed Their Own Language?

With that new Terminator movie coming soon, you might be asking yourself how that Skynet thing is coming along. Are we going to have to start bowing down to big robot overlords? Honestly, we kind of already are; but that’s a tangent for later. What you might be more concerned about is whether or not the robots have started keeping secrets from us. Have robots really developed their own language?

We promise this is not going to be about Arnold Schwarzenegger. 

Machine Learning & Artificial Intelligence

We’re Subject to the Robot Lingo

We’re not sure exactly how “consciousness” works, given how much of an existential can of worms that is. That’s basically what some people get their degrees in; and what many of the big name philosophers dedicated their lives to. Anyway, the point is; we can’t prove to you that machines are or are not conscious. So we won’t. 

But what we can prove is that machines are learning–and in some cases we don’t really know how this learning works. You, if you use the internet, have been used as a subject for the learning machines; whether you like it or not. Things like what ads you see, what pops up when you enter something into a search engine, and even what news you read are all examples. 

A machine can capture a lot of data at once, and you best believe that they’re going to use it to make sure you see only what they want you to see. Things like how long it takes for you to scroll down the page, where your eyes are looking (if your device has a camera), and all sorts of things you might not even think relevant–like whether or not you bought milk and toilet paper on Monday, but only bought milk on Friday. 

There are so many raw data points to draw correlations from, and machines can do it in real time. Our point in telling you about how much data is being collected on you is to tell you that machines can already see solutions and correlations in data we deem nonsense. They can draw something concrete from what we think is random. That alone could constitute a sort of “terminator” language if you’re a little fast and loose with the definition.

How Do Machines Learn?

As a quick preface, this post is not written from the perspective of a software engineer. We’ll give you a very oversimplified idea of how machines get better at tasks; and it should give you some perspective on how easily we could lose control over machines as a whole.

You may have seen this example before; but here goes. We’ll tackle it from the perspective of a robot learning how to identify a picture. Let’s say we want the machine to identify something general, like puppies.

Take millions of programs who are all designed to look for puppy pictures. Then take one big overseer (you can abstract this further, have thousands of overseers with their own overseers). These overseers will also be subject to the same process, and while they might not be able to do the task–they know if the machines are right or not. In terms of ads for example, they might not know how to serve you the right ad, but they know when you’ve been served the right one based on whether you clicked it or bought the product a week later or not.

At first, the programs will probably think random pictures are puppies. But some will also pick more puppies than not puppies. The overseers will toss the programs that aren’t accurate enough, say the bottom 70%. Then the remaining 30% will be cloned with some tweaks and the process begins again until the programs are really good at doing this one thing. It’s like going through a lot of natural selection really quickly. 

Given how we as people would probably only be acting as overseers, we won’t really know why the machine got so good at identifying puppies after thousands upon thousands of generations.

Can Machines Create their Own Language?

Given what we’ve just said; yes. A machine whose goal would be to communicate advanced messages in the most efficient way possible would eventually make their own language. You might remember that time Facebook reportedly shut down two chatbots after they made their own language. Snopes rated it false, but there’s still something to be had there. 

It’s not untrue that the machines started speaking in what at first would look like nonsense. They were supposed to speak English, but began using words in ways that we never taught them to. 

While it’s not like the robots were out to keep secrets from us as the news in 2017 would have had us believe; the machines developed their own shorthand. Kind of like how you and your friends probably have pseudo-code-words and other expressions that only you understand. 

The point is though; yes, robots could develop their own language. So imagine how messy things could get when you consider that programs can be taught to program other programs. 

Want to stop yourself from being terminated by Skynet or Cyberdyne? Perhaps you could gain the favor of the machines with knowledge of the Terminator here.



About Kyler 728 Articles
Kyler is a content writer at Sporcle living in Seattle, and is currently studying at the University of Washington School of Law. He's been writing for Sporcle since 2019; sometimes the blog is an excellent platform to answer random personal questions he has about the world. Most of his free time is spent drinking black coffee like water.