Generative AI and Life Sciences Webinar

Interested in current generative AI use cases for life sciences translation? Wondering how AI tools might be effectively used in the future? Curious about the potential risks and ethical challenges? Read below for the answer to these frequently asked questions about generative AI and Life Sciences.

How do we marry generative AI and Life Sciences for content production and optimization?

Remember that large language models (LLMs) are essentially text-completion machines. They produce the most likely output for a given input or prompt. In practice, this means they’re useful in any situation where “information work,” time spent on digesting and transmitting information, gets in the way of decision-making, creativity, or other valuable human efforts.

What types of content are suitable for generative AI use cases? What types of content are not?

In general, LLMs can read and understand anything a human can. However—at least in their current form—LLMs have limitations restricting their applications in some situations. For example, LLMs have a limited “window” of context. As a result, they may perform unreliably when faced with very long instruction sequences. Additionally, LLMs are unreliable for validating factual statements and have a limited ability to perform calculations or logical reasoning.

Finally, cybersecurity must be considered depending on the type of content. Using commercially available LLMs involves transmitting information to third-party systems. It’s critical to exercise caution and judgement for any content that is:

• Confidential
• Proprietary
• Subject to privacy regulations

What are some current examples of generative AI use cases in Life Sciences?

This is an area of rapid innovation, but some clear trends are emerging. At Lionbridge, we’re developing solutions for generating or “remixing” new content for specific markets or audiences. For example, given a suitable input (like a product information sheet), we instruct an LLM to generate a range of content types, from blog posts to social media snippets. We can adjust the style of these outputs as needed, based on audience-specific requirements. Similarly, instructional content can be generated, modified, and adapted without traditional source documents.

We’re also exploring how AI can accelerate translation and review workflows to eliminate “information friction.” This application will allow human experts to focus on decisions that ultimately shape quality.

Lionbridge’s Life Sciences language services team currently has customers interested in applying AI to use cases, from clinical translation, to marketing content, to plain language summaries.

What are some potential future use cases for generative AI and Life Sciences?

Future innovations will likely be in areas where “information work” impedes objectives human experts need to reach. This applies to many areas of language services for life sciences. Lionbridge is exploring how AI could accelerate these activities:

• Authoring and editing plain language summaries
• Writing comparative reviews in clinical workflows
• International harmonization for clinical outcome assessments

What are the risks of combining generative AI and Life Sciences?

LLMs are rapidly evolving, with different models carrying different strengths and weaknesses. These are some general current risks.

• Factual errors: LLMs are designed to produce responses. They can’t evaluate the veracity of the information they’re trained on.

• Computation: Current LLMs are notably bad at arithmetic.

• Context window limitations: Available computational resources limit the size of the “context window” LLMs can maintain during interactions.

• Data privacy: AI tools are like any other third-party system unless you’re hosting and training your own LLM. Exercise caution and judgment about the information you transmit with them.

How can you address or mitigate risks of using AI tools for Life Sciences workflows and content?

Mitigate risks of using AI tools by ensuring all users are well-informed. Formulate a clear policy on usage of AI, and provide users with access to reliable, updated learning resources. Policies and training resources should account for existing compliance obligations. Authorities in the EU and elsewhere are already considering AI regulations.

How can AI optimize clinical research?

Exciting developments continue to emerge in this area. For example, AI seems likely to play an increasing role in selection of candidate molecules for new therapies. In the wider clinical sector, LLMs may play a positive role in:

• Assimilating large and/or poorly structured data sets

• Managing and monitoring safety surveillance data

• Reducing documentation tasks and accelerating time-to-decision in clinical language workflows

• Assisting with authoring plain language content and promoting accessibility of content

• Enabling more rapid deployment of training and learning resources

What are some ethical concerns surrounding generative AI and Life Sciences?

Generative AI doesn’t necessarily pose any new ethical challenges. However, its applications should be subject to scrutiny. Areas of particular concern may include:

• Authorship and ownership of intellectual property used in training or other AI interactions

• Adherence to ALCOA principles in deployments and integrations involving LLMs

• Rigorous protections for patient data and other data subject to privacy regulations

• Elevated caution in AI workflows involving clinical data

• Elevated caution in AI workflows involving patient-facing content

What are limitations of generative AI in Life Sciences?

Most of AI’s current and potential shortcomings are familiar to providers of life sciences translation services and clinical trial translation services. They’re the same kinds of mistakes humans might make. Lionbridge has developed a sophisticated system of checks and balances to counteract and prevent mistakes, whether AI or human-originated. No matter how advanced AI tools become, they’ll still make errors. Our systems will address (or prevent) them.

Get in touch

Interested in implementing AI with your content production and optimization, multi-lingual clinical trials, and clinical trial translation? Lionbridge’s life sciences translation services are incorporating innovative and safe applications of AI to assist our clients through the translation process, content generation, plain language summaries, and more. Reach out to learn how our team can meet your language needs.