Babel fish, I wish!

“Summarize this research for a second-grade student” you see written on a computer screen.

The machine thinks for a second as it chews through page after page of scientific technobabble and promptly spits out a few short sentences, a perfectly understandable summary of the entire text.

This is already a reality. As millions of people around the world are starting to speak to massive language models, the ability to summarize complex ideas into simpler language is not only a useful tool but also an incredibly common desire. We specify “second-grade student” (or anything along these lines) to signal to the machine to explain in simpler language, words that anyone can understand. In essence, what does the average second-grader know, and how can I explain this concept to them?

This is really just a proxy in most cases. Contrary to what I might tell the language model, I am not summarizing papers and then literally walking into a class of second-graders to read them out. No. What I’m really saying is summarize this for me, please translate this complex article into simpler words that I can understand.

The Babel fish is small, yellow and leech-like, and probably the oddest thing in the Universe […] if you stick a Babel fish in your ear, you can instantly understand anything said to you in any form of language.

I was going to open this article by talking about universal translators, like this Babel fish from Hitchhiker’s Guide to the Galaxy. But, It doesn’t feel quite so compelling anymore, does it? universal translators are old news. And while services like Google Translate are not perfect, it feels like a societal superpower that anyone in the world is just a clicks (or taps) away from passable translations into (almost) any other language, or even to have entire conversations with people without sharing a common tongue.

Imagine instead, a device that translates not only between languages but between minds. In a conversation, you would hear not exactly what someone says, but what they mean. Complex jargon gets filtered in words you can understand, until, over time, you learn to communicate at that level. A professor might need only say a single word to a roomful of students, and they may spend hours or even days assimilating this concept into their own minds. Through a single word, a math teacher might not only the method but also a perspective of the beauty and interconnectedness of mathematics.

In such a world, the constraints holding languages together as large slowly evolving modes of communication would vanish. You could speak in your own made-up language, or invent new words for ever-more complex concepts, until everyone is speaking their own little micro-language. Existing languages will mix together, and themselves be swept away by the deluge of new vocabulary, concepts, and ideas being spread around.It is important to note that such a device would be as wondrous as it would be terrifying — for if it ceases to function after many decades (or centuries) we may find our languages so far diverged that we struggle to communicate basic concepts. Someday, perhaps, I’d like to write a story with this premise.

Whatever AI might be: a teacher-in-a-box, a personal therapist, assistant, or collaborator, it is above all, a translator. Almost every use of AI-as-a-tool involves translation of some kind. That device I describe, it already exists (in a primitive way at least, like the first stone tools of a new way of seeing.) Already we are starting to use it to translate complex topics into explanations that we can understand, or translate thoughts into code, or into new kinds of imagery. The walls between us, our mediums, and each other are falling apart in slow-motion.

But things are starting to speed up.