Something I wrote for the Japan Association of Translators’ annual anthology, 2023 edition:
I hear a lot of people saying ChatGPT and other large language models are going to drive us all out of business. (Disclaimer: I have already retired, so that “us” is rhetorical.) After all, they can parse source text and come up with reasonable-looking translations. Sure, they make mistakes. Sure, they leave things out. But so do we. That’s why both they and we need checkers and editors. The big difference is that they can work sooo much faster and sooo much cheaper that there’s not going to be any room left for the rest of us.
Of course, this is pretty much the same thing I was hearing when machine translation was being hyped 30 or whatever years ago. Fast, cheap, consistent. The Yoshinoya model. Going to drive all the other restaurants out of business. Except it didn’t.
But this time it’s different, this time’s for real, they say. And yes, the results look good, look threatening. But that does not mean there is nothing left for the I-am-not-a-robot crowd. Rather, it should be prompting us to rethink what we are doing and to upgrade our positions, even if that ends up with us doing essentially the same core stuff packaged differently.
You’re a translator, right? Tell me what a translator does without using the word “translate.” You take a message written by a specific person/organization to be read by and impact another specific person/organization (often in the plural) in a specific way and you recast it to do the same thing (have the same impact) for a different target audience in a different language (where “language” can well include the non-verbal). Do we have to call that translation? No. It is much bigger than most people’s conception of translation. It is content packaging. It is messaging. If you get to the client before the source text is finalized―or if you assume considerable leeway on the output end―it can be communication design. It really doesn’t matter what you call it, so long as you get beyond word substitution.
Nor does it matter what you call the company you cloak yourself in. Call it Going Live. Call it Settoku Station. Advantedge. More Than. Call it whatever you like―aware that what you call it will influence how people see it, hint at what you actually want to do, and probably be sufficiently singular that the many AI programs do not claim to be doing it. Why restrict yourself to the “translator” label?
The question is not whether you can survive as a narrow-definition translator. The question is how you get people to keep paying you to do what you want to do. How do you package what you want to do so it looks like something they really should be asking/paying you to do because it is such an essential service and nobody else is even close to as good at it as you are? LLMs may have landscaped the playing field, but they have not ended the game.