The language of the future won’t be one we understand. We’re building communications structures beyond our comprehension, and soon the very language used will be, too.
The Language We Have
The languages we speak are hugely inefficient. You’ll be familiar with this if you’ve ever tried to learn a foreign language – there are so many aspects that are counter-intuitive. Chinese characters are not as efficient as an alphabet. Some languages have verb tenses are absolutely baffling. Languages have many features that don’t make sense at all.
Look at this classic from the Finnish language – needing this much contextual understanding is massively inefficient:
Nine meanings from a simple two words! Give me a break. We mostly stick with the languages we have due to path dependence – we’re locked in, at this point.
Memetics And Language
The internet and free information are changing our communication. You’ll be familiar with various examples of online slang popping up, crossing language barriers. “lol” is a global shorthand at this point, used from Sao Paolo to Osaka. Image macros, animated GIFs, and Vine reactions don’t even require the sender to type anything out – and anyone can understand the message.
No words required.
The internet has created a memetic primordial soup. Messages are tested in real time on social media, and given instant feedback through likes, favorites, comments, shares, and retweets. Computers are communicating directly with humans. Twitter is filled with bot accounts that are doing real-time message testing, finding out which memes are most likely to resonate with other users. Microsoft’s “Tay” AI Twitter bot interacted with Twitter users, and became a racist within 24 hours:
Language changes online, and computers are actively involved in message-testing on humans. What happens when… they’re not talking with humans any more?
Neural Networks & The Language Of The Future
We are developing systems that will have no need to talk to humans. Will they bother to make it understandable for us?
We already have computer-to-computer communication. That’s how apps run on your phone or computer, and how you’re able to read this post right now. Computer languages are ways in which we communicate instructions to computers about how they are to operate, and communicate with other computers. An API, for example, allows one application to talk to another.
As we’ve developed computer languages and APIs, we’ve kept documentation that allows us to understand the code. After all, humans wrote it, so at least one human being will be able to understand it.
Already, we have competing, self-learning computer systems. Neural networks engage in ‘machine learning’, picking things up from inference rather than direct instruction. These have proven to be the most effective computer systems at doing things we’ve considered uniquely human, like perceiving different objects in a still image, or understanding the grammar of human language.
Eventually, we will see widespread adoption of neural networks that only communicate with one another. The cutting edge of AI research at Google is dueling neural networks – separate machine learning systems that compete and communicate with one another, becoming more advanced along the way.
When computers communicate with computers, they’ll look to do so in the most efficient way. They’re not path-dependent, like we are. Why keep human language around? As we’ve seen, it’s full of inefficiencies. If the neural networks have been given a task, coded only to find a result – does it matter how they get to it? They’re not designed to “explain your work” – just to spit out a result. So we won’t necessarily understand what’s going on between them.
Already, you have a situation where computers will do away with human language because it isn’t needed. And we haven’t even begun to consider smart-than-human artificial intelligence.
The language of the future is computer-to-computer communication. It will be selected for efficiency, not for human intelligibility. And we won’t be able to speak it.