Exploring the Transformative Power of Generative AI in Life Sciences | OutsourcingPharma Interview

Tarun Mathur: Generative AI is a type of artificial intelligence that uses large language models such as GPT-4 to generate new content in various forms such as audio, code, images, text, simulations, and videos. It has the potential to transform the life sciences industry by delivering targeted and customized information to specific stakeholders, enabling swift decision-making, improving the productivity of skilled resources such as medical writers, addressing skill gaps, converting data to intelligence, and delivering insights.

Tarun Mathur: Large language models, such as GPT-3 and GPT-4, can identify and extract information from unstructured text and accurately classify it into suitable domain taxonomies. This capacity for classification and entity identification could have far-reaching applications in various domains, including Safety, Regulatory and Labeling, MLR review, and the marketing and sales of pharmaceutical companies. New generation GPT models are multi-modal, capable of processing and generating text, images, and other media. It has significant implications for scientific communication in life sciences companies, allowing for streamlined processes across medical and commercial teams. Clinical teams can use tables-to-text and text-to-table conversion while authoring clinical documents; medical affairs and commercial teams can create a variety of infographics for medical training and promotional materials.

Sameer Lal: There are inherent risks associated with Generative AI and life sciences organizations need to be aware of these pitfalls. The outputs may not always be accurate or appropriate, biased and manipulated, and involve reputational and legal risks. The model's accuracy depends on the training data quality and the match between the model and the use case. Large GPT models have been shown to produce the "hallucinating effect," where the model creates facts or information that does not appear credible, accurate, or based on the user's request. Data privacy and security concerns and ownership and liability for such solutions make adopting Generative AI platforms in life sciences companies challenging.

Tarun Mathur: Sure. One of the use cases is in safety and pharmacovigilance. Adverse events must be reported to regulatory authorities, which can be time-consuming and error-prone. Generative AI can help identify and extract relevant information from safety reports, generate a first draft of the adverse event report, and allow the safety professional to review and refine the output. It can save time and increase the efficiency of the safety reporting process.

Sameer Lal: To ensure accurate and contextualized output of Generative AI, one needs to focus on the following – (1) Usage of validated, domain-specific data, (2) Validation and testing of Generative AI systems, (3) Prompt engineering to trigger more accurate responses.