Generative Artificial Intelligence: A New Era in the Delivery of Language Services

These revolutionary AI systems are taking automated translation and localization to new heights. Even as this technology evolves, put it to work for you now.

Lionbridge Embraces Generative AI Technology To Enable Customers To Augment Business Content


Position your enterprise to transition to this rapidly evolving technology with Lionbridge’s expert guidance. 

Although computer scientists have been following developments in generative Artificial Intelligence (GenAI) and Large Language Models (LLMs) for years, the technology grabbed the mainstream’s attention with the 2022 launch of ChatGPT, an LLM developed by U.S. tech company OpenAI. Reuters® reported that the novel app was the fastest-growing app in history, attracting an estimated 100 million users two months after its launch.

Generative AI technology is receiving attention for a good reason; its ability to generate text in pretty much any language will profoundly change how we work and conduct business.

Goldman Sachs estimates the tool could be responsible for an almost $7 trillion increase in global Gross Domestic Product (GDP) and raise productivity by 1.5 percent by 2033.

As this groundbreaking technology evolves and becomes scalable, it will disrupt the localization industry. This AI is already impacting the delivery of language services.

Lionbridge is an early adopter of generative AI technology and is poised to help you leverage all it offers.

Defining the Landscape

You’re bound to come across terminology associated with generative AI. Here’s what you need to know to get started.

What is Generative AI?

It’s an Artificial Intelligence (AI) system that can generate novel content, including text and images, based on prompts and extensive multimodal training. It determines the most plausible output that appears to have been produced by a human.

What is a Large Language Model?

It’s an AI system focused on languages. It can summarize, translate, predict, and generate text from knowledge gained from massive databases. Although it’s not specifically trained to translate text, it can do so with decent quality and is quickly improving.

What is GPT?

It’s a family of Large Language Models created by OpenAI. The GPT family includes various versions of the AI, such as GPT-3, GPT-3.5, GPT-4, and others.

GPT-3.5 and GPT-4 power ChatGPT, OpenAI’s freemium chatbot product. GPT-4 is recognized as the most capable of all Large Language models and produces, among other things, better linguistic results.  

Other LLM brands include Google’s Bard, PaLM and LaMDA, Meta AI’s LLaMA 2, and DeepMind’s Chinchilla. Many other models exist or are in development.   

Large Language Model and Generative AI Technology On-Demand Webinar

Find out how this Artificial Intelligence will affect localization workflows. Discover new possibilities for translation and content creation.

Generative AI Across the Content Lifecycle

There are opportunities to leverage generative AI, even in its early stages, but you must use caution when deploying it.

When Should Generative AI Be Used?

During Content Creation

The technology can create content when you have references and examples. For instance, generative AI can help you create a new marketing campaign based on a prior campaign and help you check your content for grammatical or stylistic changes.

During Initial Translation

If you are performing multilingual content generation, generative AI will be excellent at creating multilingual prompts because it will have source input and output to use as a reference.

During Post-editing and Content Review

The technology is excellent at comparing content across languages to determine whether it has the same meaning. It can edit the text for a better fit when necessary. When considering whether to use generative AI instead of linguists for some of these tasks, make sure the language pair and domain work well with generative AI and that it is more cost-effective than using the services of a linguist. Our initial research suggests that some use cases will be more suitable for generative AI, and others will be more suitable for linguists.

When Shouldn’t Generative AI Be Used?

During Content Creation

Don’t use generative AI to create content when you cannot provide context. The technology cannot determine whether something is true and could make incorrect assertions. For instance, it’s not a good option when producing technical documentation.

During Initial Translation

The technology is not a replacement for Machine Translation and should not be used as such for initial translations. Current generative AI models are not economically efficient.

During Post-editing and Content Review

The technology was largely built from an English corpus of publicly available content and is, therefore, less able to determine the context of the text in highly specialized domains or provide quality reviews in less common languages. We expect improvements to these shortcomings in the future, but until such time, we recommend using a blended model that incorporates both generative AI and linguists.

Put Generative AI To Work for You With These Services

While generative AI technology has yet to mature fully, we can leverage it for certain content creation, translation, and post-editing tasks.

We are continuously researching and developing ways to incorporate LLMs into professional translation and can help you get the most out of it as it rapidly evolves. Let us help you get started via the following offerings.

The Lionbridge Content Remix App for Impactful Multilingual Content Creation

Generate fresh, personalized content for broad audiences and multiple channels quickly and easily with the Lionbridge Content Remix App. This AI-powered multilingual content creation solution enables you to create original content in 70+ languages for your websites, e-commerce sites, social media platforms, customer support channels, and more.

Provide your requirements and data, and the app’s content generator simultaneously produces output in one or more languages within seconds. The app’s array of editing capabilities can help refine the output. You may choose an optional human-in-the-loop review by our vast pool of language specialists and Subject Matter Experts (SMEs) to bolster your trust in output accuracy, authenticity, and cultural relevance.

Achieve ultimate efficiency, reduced content creation times, and enhanced personalization and audience engagement.

AI Training for Optimal Output

Achieve reliable LLM performance with continuous evaluation, prompt engineering, and assistance curating training datasets. With our AI, cultural, and linguistic expertise, you’ll consistently get AI-generated content that connects with diverse global audiences and falls within sensitivity guidelines.

Data annotation, collection, and creation: Using the right training data is crucial to ensuring your LLM understands prompts and responds with the output users seek. Get assistance choosing impactful, high-quality, multilingual data sets to train your LLM. We’ll help you pull from a comprehensive variety of text, images, audio, and video.

Prompt Engineering: Develop and alter prompts to improve LLM’s translation and localization performance. We assist with prompt creation, translation, and transcreation.  Lionbridge LLM Training Services verify that every prompt is built with natural language mimicking typical AI tool users.  

Response Evaluation: Our experienced crowd helps evaluate output and prompts, verifying that your LLM will consistently deliver desired output more efficiently. This evaluation also reduces risk of both inaccurate responses and introducing bias from user feedback, source material, users, etc.

Affordable Multilingual Asset Optimization

Safeguard your translation results with our AI-powered service that updates your linguistic assets more affordably than was previously possible.

Our solution uses AI to identify and modify strings requiring Translation Memory (TM) updates. Human linguists then perform quality assurance on a representative sample to ensure the AI’s effectiveness.

Companies may overcome exorbitant costs and time-consuming processes previously associated with multilingual asset optimization by using AI to replace many manual tasks associated with these projects. Informalizing an entire TM or addressing your massive TM containing legacy content with outdated terminology is now within reach.

LLM Automated and Assisted Post-Editing for Enhanced Localization Workflows

Achieve faster and more cost-effective translation by leveraging LLMs during post-editing.

For fully automated post-editing, we run source content through your Translation Memory and then Machine Translation. Next, a customized LLM decides what output needs enhancement and makes appropriate edits. For assisted posted editing, there’s an extra step whereby a professional human linguist reviews the LLM’s work and makes any required corrections.

Automated post-editing via an LLM will be of better quality than unsupervised Machine Translation. It is appropriate for moderate translation quality. LLM-assisted post-editing that includes a human in the loop ensures superior quality while maintaining a fast turnaround time as the LLM significantly reduces the editor’s workload.

Review Some of the Most Engaging Content

Can AI Translation Replace Professional Translation Services?

Learn why AI translation can’t replace translation services and Language Services Providers (LSPs).

Video: A Primer on Generative AI and Large Language Models

View the video to discover what generative AI and Large Language Models are and what needs to happen for the technologies to handle the enormous demand.

Case Study: AI Content Creation for the Hospitality Industry

Discover how Lionbridge’s innovative AI content creation solution enabled a luxury resort chain to save time and costs during global content creation.

AI Case Study: Multilingual Retail Marketing

How can retail companies capitalize on AI? Read about how Lionbridge helped one customer use AI to build content in new languages faster and more cost-effectively, thus allowing them to reach new markets.

What ChatGPT Gets Right and Wrong: Lionbridge’s Take

Understand ChatGPT’s weaknesses and strengths and the future of generative AI.

ChatGPT’s Translation Performance and What It Tells Us About the Future of Machine Translation

Machine Translation is on the cusp of leaping forward thanks to Large Language Models like ChatGPT. Learn how ChatGPT compares to MT engines.

The Lionbridge 2023 Machine Translation Report

Lionbridge experts highlight the state of Machine Translation in 2023 and how the technology will evolve to catapult businesses’ growth further.

Video: Why AI Translation Can’t Replace Translation Services

Watch the video to discover four reasons why AI translation alone can’t replace professional translation services.

Webinar Recap: Generative AI for Global Content

Read our webinar recap to understand Generative AI’s disruptive impact on global content creation, including workflow changes and use cases.

A Generative AI Model Outperformed a Neural Machine Translation Engine in One Evaluation

Lionbridge found that GPT-4 can outperform Yandex in one language pair. Find out what that means for the Neural Machine Translation paradigm.

4 Key Steps to Increasing AI Trust for Your Content

Learn about the AI trust gap and the 4 steps to minimize risk of untrustworthy content creation, localization, and translation.

AI Content Optimization and Localization Case Study

Find out how 4 tools performed in an experiment focused on AI content creation, AI content optimization, and localization in German.

What to Know About AI Self-Correction

Find out what AI Self-correction is, how to implement it with or without an AI solutions provider, and its current limitations.

Meet Our Generative AI Experts

Vincent Henderson

As Lionbridge’s leader of the product and development teams, Vincent focuses on ways to use technology and AI to analyze, evaluate, process, and generate global content. He is especially attentive to the disruption of content products and services brought about by Large Language Models.

Share on LinkedIn

Rafa Moral

As Vice President, Innovation, Rafa oversees R&D activities related to language and translation. His responsibilities include initiatives involving Machine Translation, Content Profiling and Analysis, Terminology Mining, and Linguistic Quality Assurance and Control.

Share on LinkedIn

Let's Talk

Business Email Only