Translation and All the AI Panic

I won’t start with clichés like “AI is here to stay” or “I use AI to my advantage.” Of course, I use AI. I’ll probably even ask ChatGPT to check my grammar once I finish writing this. But that’s the point: AI is good for checking grammar, punctuation, or answering very specific queries. It’s not here to create.

The first time I was exposed to MTPE was back in 2017, when one of the biggest video game companies decided to build their own LLM using the translations I had been producing for years. I was shocked. It was… good. So good, in fact, that it sometimes mimicked my writing style so closely I couldn’t tell if the text came from my TM or from the machine. But then again, the content was extremely repetitive, mostly about loot boxes. And even then, it still made some awful mistakes.

Fast forward to now, and suddenly AI is everywhere. People treat it like magic. On LinkedIn, I see endless posts insisting that you have to embrace AI in everything you do or risk becoming “obsolete.” But nobody seems willing to admit the truth: AI cannot match the quality of a translator with just three years of experience. It hallucinates. Even ChatGPT itself says at the bottom of every window: “ChatGPT can make mistakes. Check important information.” Yet somehow, we’re too scared to criticize it.

Here’s the reality: AI is not an intellectual being. It’s not “thinking.” It’s just searching data and stitching it together in a way that feels coherent. That’s it. What makes it impressive is that it can maintain a conversation -up to a point- and self-correct when you give it more context. But it’s still just a glorified search engine. Apple nailed it in their paper “The Illusion of Thinking.” Yet that paper gets ignored, because it doesn’t fit the narrative of enterprises that are desperate to cut costs, especially our rates.

A few days ago, I received a proposal from a Chinese translation agency. It came with the usual sales pitch about how “the industry is changing” and how “AI is so advanced now.” And then came the offer: $0.01 for AIPE. That’s right—MTPE has now been rebranded as AIPE. And if I complain, I risk being labeled an “outdated service provider crying over spilled milk.” But let’s be honest: AI is terrible. If you treat it like it’s infallible, it will eventually get you into serious trouble.

Here’s an example. I was recently working on a cartoon where a character’s English name already had an official Arabic version which is quite different from the transliteration. I fed the AI the context, gave it the phrase, and it still produced a basic transliteration. Only after I pushed it further did it finally provide the correct official name. If I hadn’t known the right answer myself, I would have delivered a catastrophic error.

So, let’s stop pretending. AI is just data, and often it’s the most SEO-optimized data, not the best. Don’t be afraid to say what Apple said: your AI sucks, and so does your LLM. Sure, we’ll follow the trends, because we must! But not at the cost of quality, professionalism, or $0.01 rates!