Translators' order cautions public about risks of using artificial intelligence for translation

With the use of AI in translation exploding, Quebec's professional order for translators, terminologists and interpreters is sounding the alarm. Although the technology has great potential, it also poses great risks.

MONTRÉAL, Oct. 16, 2024 /CNW/ - Faced with the rapid rise of AI-based tools and their growing use by the general public, particularly for translation, the Ordre des traducteurs, terminologues et interprètes agréés du Québec (OTTIAQ) is alerting the public about the risks of failing to work with a qualified professional when using these tools.

Despite the innovative opportunities created by automated translation tools and chatbots such as ChatGPT, these technologies have their limitations. As OTTIAQ points out, errors in meaning are frequent—particularly in complex or specialized documents—and unacceptable cultural biases are often embedded in the output. In translation, such mistakes can have serious legal, financial or reputational consequences.

Case in point: a pharmacy customer received an email that wished her dead. The culprit? An automated translation. Obviously this shocked the customer, but it was the pharmacy's reputation that truly suffered.

Privacy and data security are also at risk, since most free translation tools don't provide any guarantee that the text fed into them will be kept confidential. Moreover, they actually use that data to produce other content, which goes against Canadian and Quebec privacy and data protection laws.

"Having a rigorous, human-led process is the only way to ensure reliable, secure, quality language services," says OTTIAQ president Betty Cohen. "We strongly recommend working with certified translators to avoid potentially costly errors."

As a professional order, OTTIAQ has a legal mandate to protect the public. It does so by informing the public, setting high professional standards, and promoting the expertise of its over 2,800 certified members.

OTTIAQ's position on artificial intelligence in translation

The following position on the use of artificial intelligence in translation has been approved by the Board of Directors of the Ordre des traducteurs, terminologues et interprètes agréés du Québec (OTTIAQ). Given its mission to protect the public, OTTIAQ believes it is important for everyone to understand the potential uses and limitations of artificial intelligence in translation, interpretation and terminology. OTTIAQ encourages its members and the public to share this information and stay abreast of the latest technological developments in order to use these tools safely and effectively.

To fulfill its mandate and provide better guidelines for the work of translators, interpreters and terminologists, OTTIAQ must keep close tabs on technological advances that have an impact on the language professions. The rapid pace of these advances can be seen in the advent of generative artificial intelligence and tools such as ChatGPT. The general public now has access to automated translation tools that are fast and easy to use, but come with a high degree of risk. As a professional order governed by Quebec's Professional Code, OTTIAQ believes it has a duty to inform the public of the benefits and drawbacks of the automated translation and interpretation tools currently available.

What you need to know

Translation is one of the main applications of artificial intelligence. Neural machine translation (NMT), which is based on the neural networks used in AI, appeared on the market in 2016, well before generative artificial intelligence.

Professional translators were already using other digital technologies and quickly added NMT to their toolboxes, but not without putting processes in place to detect and correct its mistakes. They were thus early adopters of this revolutionary technology.

The advent of generative artificial intelligence hasn't actually changed those processes much. What is has done is provide language professionals with new tools that can improve their workflow, particularly when researching terminology. Yet academic and corporate studies have shown that dedicated automated translation tools produce more specific and accurate translations than ones based on generative AI. While translations produced by generative AI are based on whole documents and can be stylistically superior, they often contain numerous inaccuracies—the product of AI's tendency to make things up, called a "hallucination." In such situations, the involvement of a language professional becomes even more crucial.

Automated translation can be very useful for content that is straightforward or general in nature. It is far less so for more technical or sensitive documents, where a poor translation can have serious physical, mental, financial, legal or reputational consequences. In the same way, automated interpretation apps are helpful for simple conversations—when travelling, for example—but cannot replace the services of a professional medical or court interpreter.

In the interest of public safety, automated translation and interpretation tools must be used wisely. That means that whenever there are real-world consequences, a language professional should be involved.

Further considerations

Being easily accessible to the general public, automated translation tools and generative AI are often presented as miracle solutions for translation and interpretation. But along with the risk of sheer mistakes, there are other reasons to proceed with caution:

• Confidentiality – Free automated translation tools don't provide any assurance that the data you provide will be kept confidential. Worse still, the data is used to train the AI or produce other translations. The terms and conditions of some of the most popular apps even state outright that any content fed into the system will be reused.
• Data security – The translation tools available to the general public operate from servers located outside Canada. Using them to process confidential data or personal information goes against Canadian and Quebec privacy and data protection laws.
• Quality – Large language models like ChatGPT are trained on billions of gigabytes of online content. The machine retransmits that content, and in doing so, can misinterpret information, reproduce cognitive and cultural biases, fail to be inclusive, and contribute to declining writing quality by reusing machine output. It's important to remember that AI does not think. Only humans can work consciously and from a multidimensional perspective.

OTTIAQ's position

OTTIAQ welcomes new technology and encourages members to use it to provide the best possible language services. But it also calls on them to integrate these tools into an OTTIAQ-approved process, which must always include at least one round of quality control by a professional. In OTTIAQ's view, this is the only way for language professionals to uphold their code of ethics and properly protect the public.

By the same token, OTTIAQ calls on the public not to use free automated translation tools without due consideration for the risks outlined above and strongly advises anyone in need of quality language services to work with a conscientious, certified professional.

Key takeaways

• AI-based translation tools have been available since 2016. The technology isn't new for translators; it's already part of their workflows.
• Translation tools for the general public can be useful for simple texts of a general nature that don't contain personal or confidential information.
• In any other situation, failure to involve a language professional could have serious consequences

SOURCE Ordre des traducteurs, terminologues et interprètes agréés du Québec

For more information: Soumaya Boumazza, Communications Officer, [email protected], 514-845-4411, ext. 1222