Can you use AI to produce good translations? This is something a lot of our customers are wondering. And yes, a great deal has changed recently, with the advent of Large Language Models (LLMs), which generate text themselves using generative artificial intelligence (AI). What does this mean for translation? Does it mean we won’t even need translators any more in a few years? The simple answer is no. Ultimately, we’ll still always need linguists, just the technology will cause their role to change.
Is AI cheaper and faster?
The burning question all our customers have is whether AI makes translation faster and cheaper. Unfortunately, there’s no definitive answer to that question. AI can certainly make some translations faster and cheaper, but a different solution, without using AI, is sometimes better.
Translators have used advanced CAT (Computer-Aided Translation) tools for many years now. AI is simply the next step in refining those CAT tools. And yes, the job is sometimes cheaper and faster with AI, but not always. It all depends on what your ultimate goal is. Quality, price and speed are all considerations in your choice.
The most important advice we can give you is: talk to your translators! They know which specialist tools are available on the market and they know which solution is right for your project.
Translating with chatbots
Translation agencies use a specific AI application, which is different to chatbots like ChatGPT and Copilot. Chatbots weren’t designed for translation and the translations they produce require a lot of post-editing, which isn’t necessarily cheaper and faster than translation. The main problems with these chatbots are:
- inconsistency;
- incorrect terminology;
- wrong tone of voice;
- wrong style for the target audience.
What’s the difference between chatbots and translation tools that use AI?
Advanced translation tools combine three elements::
- a database of translations approved by the customer;
- Neural Machine Translation,
- a Large Language Model (AI).
It’s precisely this combination that makes translation easier. The first step involves translating text using the database. Text not found in the database is then translated using Neural Machine Translation (NMT). NMT uses ‘deep neural networks’, based on how the human brain works. The third step is using AI to improve the NMT output. This makes the text flow more naturally and AI also detects inconsistencies and terminology that deviates from the glossaries.
Post-editing
However advanced it may be, AI output is never good enough. Good translations pretty much always hinge on nuances, something which AI doesn’t pick up on. That’s why it’s crucial for a linguist to check the translation afterwards. The linguist painstakingly checks every last word, because even a minor error in the translation can have major consequences.
Linguists remain necessary
AI translation is best viewed as a significant addition to what the existing tools used by translation agencies can already do. But linguists remain necessary to check and improve the translations produced by AI. They are the only people capable of checking the text properly. They master the language at a professional level and understand better than anyone where errors can occur with AI and, more importantly, how to fix them.
AI can be cheaper and faster, but it’s sometimes better to use a human translation. Sometimes a different solution without AI is better.
See also: When is AI a good idea and when isn’t it?
See also: I don’t need a translation agency when I’ve got AI, right?
