Generative AI and Life Sciences

This blog post will be the first in a series about generative AI and life sciences, especially its implications for the field and life sciences translation services. We’ll examine generative AI technologies, such as ChatGPT and neural MT. We’ll also outline the current state of the art, and then preview some potential generative AI use cases in our industries.

The Current Status of Generative AI and Life Sciences

You probably know more about generative AI than you thought. You’ve almost certainly heard of ChatGPT, which has made headlines since late 2022. You’ve likely interacted with AI chatbots too, even if you weren’t aware of it. If you’ve “chatted” online with a customer service agent recently, there’s a good chance it wasn’t a human.

By some estimates, about 70% of customer service interactions are now fully handled by AI. Facebook Messenger has over 300,000 chatbots in active use. The trend has been especially pronounced in sectors like online retail. The reasons why are worth exploring. Many customer queries in these settings are:

• Simple
• Repetitive
• Relatively urgent

More complex inquiries still require human expertise and judgment. However, chatbots are good enough at conversation to execute a form of screening. Only customers who can’t get a fast, satisfactory response from AI will seek additional human assistance.

Business leaders and investors are paying close attention to these developments. Unfortunately, some are drawing incorrect conclusions. Early experiments seem to show that AI can be successfully deployed in limited roles, resulting in improved service availability and customer choice. Even so, customers with complex problems or strong preferences aren’t going away. Neither is the need for human intervention. You can allocate costs more efficiently with AI, but not eliminate them.

In life sciences, regulatory constraints and safety-focused culture traditionally delayed technology adoption. However, innovation is the industry’s lifeblood. For innovators and shareholders alike, technology as potentially transformative as AI seems impossible to ignore. Could AI be used to reach patients and markets faster? Could it be introduced safely?

Lionbridge, a language service provider to leading life sciences organizations, is actively exploring this question. As it turns out, we have an important head start. Tools like ChatGPT may have grabbed a lot of recent headlines. But the underlying technologies – large language models (LLMs) and generative AI – are not new. In fact, Lionbridge has been using them for many years. We’ve built considerable expertise about what these technologies can and can’t do. We’re also actively exploring what might soon be possible.

The Road Ahead: Generative AI and Life Sciences

We’ve learned many valuable lessons in the years we’ve spent integrating AI with quality-centric processes. We're now seeing familiar patterns emerge in other industries. Again, customer service chatbots are an instructive example. As organizations deploying those solutions are discovering, AI is most effective when it eliminates friction and accelerates critical decisions by human experts.

Does that mean AI tools can’t be deployed without close supervision? Not at all. It just means conscious choices must be made. These choices must be made based on the nature of the content and the needs of the intended audience. We naturally take a conservative approach if that audience includes patients or regulators. When outcomes are less critical, we offer them appropriate choices. This approach has served us well, and we believe it can be more widely replicated.

Here are some of the areas we’ll explore.

Analyzing and classifying large content sets: LLMs and ML systems are designed to find meaningful patterns in large datasets. Researchers are already examining some relatively obvious applications of these technologies. In the field of diagnostic imaging, there is growing evidence that machine learning may have an important role. Rather than replacing expert clinicians’ judgment in these settings, AI is more likely to aid it.

Making clinical data more accessible: Recently, regulators and patient advocacy groups have made significant progress ensuring patients and lay people can access and understand clinical data that informs both regulatory decisions and patient treatment choices. Notably, producing reliable information in appropriate language requires specialized skills in science communication and knowledge of relevant standards and best practices. Again, AI systems are unlikely to replace this expertise soon. They may help expand the range and type of accessible content.

Discovering what we don’t know: Performing meta-analyses of multiple scientific studies has long been recognized as a means of arriving at statistically reliable findings from multiple studies whose sample sizes, methods, or biases are often susceptible to dispute. One self-evident limitation of meta-analysis is that it can only find patterns it intentionally seeks. Machine learning systems have potential to identify meaningful patterns beyond those purposely measured or correlated.

Content discovery and classification: A challenge Lionbridge and our customers frequently encounter is that varying content types require different language services. This challenge is especially prevalent in life sciences. Many services performed for regulated content (e.g., expert clinician reviews) are irrelevant for online marketing copy or internal training manuals. Classifying and directing such content to appropriate service channels consumes significant time and skilled resources. Lionbridge is already working to eliminate these inefficiencies using AI technologies. We expect these developments to accelerate in the coming months.

Get in touch

Interested in learning more about generative AI and life sciences? Want to discuss potential generative AI use cases for your organization? Have life sciences translation needs? We have the deep experience, knowledge, and technology to help. Contact us today to find out more about Lionbridge’s Life Sciences translation services.