Can AI Translation Replace Professional Translation Services?

Generative Artificial Intelligence (AI) is radically altering the translation landscape. It has some more sophisticated capabilities than easily accessible Neural Machine Translation (NMT) tools, such as DeepL or Google Translate. While generative AI technology is not explicitly designed as a translation tool, it has surprisingly powerful translation capabilities. Lionbridge is embracing generative AI and recently explored its applications in a whitepaper and a webinar.

Unlike the major Machine Translation (MT) engines, 2023 generative AI technology can:

With access to such sophisticated technology, organizations may be wondering if they will still need professional translation. The short answer to this question is yes. Even with the aforementioned abilities, ChatGPT and other generative AI still can’t replace the translation services that most businesses need. Instead, generative AI can (and should) be incorporated into a hybrid workflow for a more efficient translation process. However, delivering the best translations (and customer service) will still require human work, experience, and expertise.

One of the fundamental reasons ChatGPT or other AI translation isn’t sufficient without human accompaniment is that it’s sometimes wrong. Accuracy is often crucial for effective translation. This reason alone renders current generative AI tools untrustworthy when used without a human reviewer and/or prompt engineer. Like Machine Translation, generative AI requires a human to review translations and enter suitable prompts to ensure accuracy and adherence to project guidelines. This weakness is just one of the reasons generative AI should be embraced only as a tool for optimized translation services.

Here are four other key reasons why AI translation should enhance, rather than replace, language services from a professional translator.

#1 Translation Companies Have the Technology to Streamline and Keep Translation Projects Organized

Completing a translation project requires multiple steps, including transferring content from the organization to the translator and back again. There are more steps in between, including various reviews and back translations. All these steps require a manual process when done via generative AI. The organization must complete the transference and ensure content and translations stay organized — an onerous task with high volumes of content to translate. A translation company offers a two-pronged solution to this challenge:

• Technology: Translation companies offer a Translation Management System (TMS) to organize a translation process and ensure they do everything efficiently and accurately. A TMS is especially crucial with high-volume translation projects, where keeping track of a lot of text is necessary. Custom solutions may even include a connector, which enables the translation company to connect directly to the customer’s Content Management System (CMS). This arrangement speeds up how translations are performed and eliminates administrative burden for the customer.

  Lionbridge’s Language Cloud™ uses a combination of AI and professional translators to help automate the entire customer experience, freeing up employee bandwidth for more impactful work. Customers can directly connect their Content Management Systems to Lionbridge’s translation systems. They’ll also receive customized workflows for the most efficient and highest-quality content translation.

• Project Managers: Translation companies provide experienced, efficient, and proactive project managers. These people are trained in retrieving content, shepherding it through the translation process, and ensuring everything is done securely and on time.

#2 AI Translation Can’t Protect Your Brand Voice

Brand voice is a crucial element of every organization’s identity. It helps customers recognize and connect emotionally with a business or nonprofit. Trusting the brand voice to a translation performed solely by AI, with no human reviewers or project managers, could cost significant ROI and have a lasting, negative impact on your brand's reputation. A translation company usually has multiple solutions to ensure retention of brand voice, regardless of the number of languages or countries for which translations are needed. (These may soon include training and incorporating generative AI.) These solutions include:

• Assistance building a centralized, corporate dictionary and glossary. These documents will include essential terms for the brand voice. Often, a Language Service Provider (LSP) can also adapt this glossary to the linguistic and cultural norms of the locations where translations will be utilized. When translations don’t follow the corporate dictionary, it creates significant confusion and hurts consumer confidence. This effect is most common for companies with very technical products and services.

• Enforcement of your language style guide in translations. These guides help define the voice and tone of the company. They’re full of subtle nuance. A human reviewer and translator can use AI tools to help them ensure a translation is faithful to the language style guide throughout the text. Language Service Providers can train generative AI technology to assist with implementing specific style guide standards. (Generative AI is distinctly better at this task than its predecessor, Machine Translation.) However, including a human reviewer is crucial to both ensure the style guide is followed and set the right prompts to help the generative AI understand the style guide. As noted throughout this article, generative AI by itself still makes mistakes or falls short of customers’ expectations. These are some examples of items a human should review for and help train generative AI to assist with:

#3 AI Translation Doesn’t Offer Deep, Industry-Specific Experience

Translation projects within specific industries necessitate deep, updated expertise. Consider these three examples:

Legal: Legal translation must be done accurately, down to every word. Otherwise, there may be dire consequences. A contract could be rendered meaningless if just a word or two are mistranslated. Lawsuits are won and lost based on a comprehensive understanding of relevant documents. Lawyers would never trust AI without human accompaniment to complete translations. The stakes are too high. (This New York Times article shares the disastrous results when a lawyer tried to rely solely on ChatGPT for his legal proceedings.) A legal translation requires a translator with knowledge of all relevant languages and a legal degree. They must understand the most recent versions of the law, which change over the years, sometimes on complex timetables. In addition to the country-specific legal system and law, a legal translator should also understand the socio-cultural context of the translated documents. An AI translation app is unlikely to “know” without human guidance how to utilize all this information to deliver the most accurate, updated legal document translation.

Life Sciences: As mentioned above, life sciences translation requires familiarity with applicable, ever-changing industry regulations. Life sciences translations also require an understanding of the latest advances in medicine and clinical research. Generative AI may have access to some of this information, but it can’t consistently choose the current, most widely-accepted language and terminology. In addition to offering the expertise of current industry information, a life sciences translation company helps customers develop a glossary. Glossaries ensure specific terms are always translated correctly, regardless of the language used. Note that a Language Service Provider may use generative AI to help apply a glossary to their translation process for more efficient translations. They will, of course, need humans to review the results and ensure the customer’s solution is tailored to their needs.

Financial Services: Financial services translation requires industry knowledge and experience. Mistakes within these translations could create catastrophic financial or legal problems, potentially damaging a company’s reputation. Financial services translations are full of complex terminology that may change from country to country. Additionally, the correct formats must be utilized, and financial data must be considered and correctly interpreted. Sometimes, the translator must understand country-specific tax regulations and legal business requirements. Generative AI can’t meet these vast, nuanced knowledge requirements when translating. Nor would most financial services companies trust their translations to generative AI without the oversight of a credible translation company.

#4 AI Translation Doesn’t Meet Legal or Regulatory Requirements on Its Own

Depending on the industry, translations may need to meet specific regulatory requirements. These are a few examples of regulatory requirements an AI translation could only satisfy with assistance from humans and professional translation services.

• Financial services: Financial services companies have stringent requirements for their translation services providers. Their client data must be safe from hackers and cyberattacks. Additionally, their own internal data also needs to be kept secure. Open-source AI software, the kind that’s usually free and utilized by non-professional translators, will always be inherently unsecured. This makes it unsafe for a financial company’s sensitive data. A translation solution using solely AI translation, done with open-source tools and no professional translation experts who know how to ensure security, will threaten a financial company’s:

  • Consumer confidence
  • Performance
  • Reputation

  Alternatively, a financial company can trust a language service provider using its own secure AI tools (or perhaps MT tools, depending on the customer’s needs) and its own experienced human project managers. This combination of human and generative AI can deliver translations and ensure data safety.

• Legal: Some countries require specific certifications for translations used in court or a legal setting. For example, Spain, Germany, France, Poland, and Romania’s legal systems all include variations of a “Sworn Translation.” This means that accuracy, confidentiality, and other legal formalities are certified with a seal from translators authorized by relevant government authorities. Sometimes a government-certified notary’s signature is required on a translation. For various reasons, AI does not, and likely will never, have a legal translator certification. Perhaps the most important reason is that AI cannot be held liable for a faulty translation like a person can. Thus, AI bears no consequences and cannot be penalized or held liable for incorrect work, as a legal translation services provider can. As in this legal case, chronicled in the New York Times, users will bear the consequences (and punishment) for relying solely on generative AI when they shouldn’t.

• Clinical research: Clinical trial sponsors must meet complex, ever-changing linguistic requirements to gain approval from industry regulators, like the U.S. Food and Drug Administration (FDA) or the International Council for Harmonisation for Technical Requirements for Pharmaceuticals for Human Use (ICH). Sometimes, these requirements differ from continent to continent or even country to country. Requirements typically include tight deadlines and necessitate a deep familiarity and understanding of regulations, including the European Union Clinical Trial Regulations (EU CTR), the European Union Medical Device Regulation (EU MDR), and other laws and regulations. Using an experienced life sciences translation provider is crucial to meeting all applicable regulations. ChatGPT or other generative AI simply can’t be trusted to handle projects like this by itself.

• ISO certification: ISO certification ensures a translation is accurate and classified documents are safe and secure. Many businesses, especially in the life sciences industry (including clinical research and medical services), must work with a translation company that holds either an ISO 9001 Certification or ISO 27001 Certification. These certifications help protect a business’s clients and patients from cyberattacks. As of now, and for the foreseeable future, generative AI will not be able to secure ISO certifications.

• Certified translation: A certified translation has a certification statement. It can be completed by a translator or somebody else involved in the process (i.e., a project manager or employee at a translation company). The certification statement typically includes the following:

Get in touch

Want to find out more about how Lionbridge’s translation services can help your organization? Need assistance with industry-specific translation needs? We utilize human reviewers and strategically implement generative AI technology to develop customized translation solutions. We have decades of experience and innovative tools to tailor language services solutions to your unique needs. Get in touch today.