AI Translation: Convenience vs. Risk

The Rise of AI...

The subject of artificial intelligence (AI) has been around for a while, but since late 2022, when ChatGPT was released, the technology has become nearly ubiquitous. Many of our everyday tools are now assisted or powered by AI, from social media platforms and productivity suites to entire operating systems, supercharging every user interface with the promise of more and better capabilities to automate tasks that would otherwise take up more of our time and resources. And translation is no exception, becoming one of the many tasks AI has set out to automate.

Now, AI solutions offer users what could only be described as “translations on tap,” where you can paste the content or upload a file to get an instant and seemingly “flawless” translation in the language of your choice. This has led many to ditch their professional translation providers in favor of this cheaper and faster option.

...but is AI translation really worth the savings?

A translation obtained through an AI solution (which in this context would be a large language learning model, or LLM) may be faster and cost less than a professional translation, but is it always worth the savings? 

At OGMA, we ask ourselves that very same question every day. How can we, as professional language service providers, leverage these new technologies to benefit our clients while also making responsible decisions regarding their content?

Not every translation involves the same stakes...

Maybe all you need to translate is a small bit of text—a post for social media, an email, a news article—but sometimes, the need is more considerable in terms of word count, complexity, and even confidentiality. And AI translation may have improved by leaps and bounds in the last few years, but some problems have persisted across iterations, and new ones have cropped up as well.

How AI translation actually works

To translate, AI solutions use an LLM to predict the most likely translation based on patterns learned from multilingual datasets. These datasets are usually massive collections of information and content scraped from public web pages, digital books, academic papers, user-generated content, etc. 

In short, your run-of-the-mill AI platform has been populated with a lot of information, and that includes the good, the bad, and the ugly.

The risks of relying solely on AI 

AI’s most common issues include hallucinations (fabricating information) and context blindness (missing nuance, tone, or local meanings), among other problems that could be of a greater or lesser impact, depending on your content and priorities. 

Keep in mind that significant confidentiality risks may also arise when using free and public AI models, as these often retain user input to train their algorithms.

Questions to ask before choosing AI

This is when most people start weighing the pros and cons of letting AI do it, vs. reaching out to a professional translator. The question of whether or not to use AI for your translation needs is ultimately not a cut-and-dried issue. As the owner or user of the content, you will need to ponder your needs and priorities:

Find the Right Translation Solution for Your Needs

Still unsure about how to proceed? At OGMA, we can help you figure out your options, with a variety of services, ranging from proofreading and revision to in-depth editing, post-editing, and professional translations, that can be tailored to your specific needs. Give us a call to get started!

Written by:
Dorian Gray-Sorgin, Lead Translator

Let us tackle your language needs!

Precision, experience, and expertise —linguistic excellence is just a click away.
Contact us today, we will gladly help you out.