Despite the potential of AI, there are roadblocks when executing an AI strategy, particularly because today's publicly available generative AI tools are not conducive to the pharmaceutical industry. The language of pharmaceutical research has its intricacies and terminology that AI models often struggle to understand. Furthermore, hallucinations, which occur when AI produces results based on incorrect or incomplete data, can lead to missteps in clinical trial and drug development, wasted investments, and biased or misinterpreted clinical trial data. Katz concludes that once these issues are addressed, AI will be a powerful tool for clinical researchers.
Key takeaways:
- AI holds significant potential for transforming healthcare, particularly in clinical trials, by aiding in the identification of ideal trial sites, streamlining patient recruitment, and reducing costs associated with trial design.
- Three critical developments must take place before AI can be fully applied to the clinical trial process: improving the underlying data sources, training the tools on pharma-specific lingo, and overcoming hallucinations.
- AI models should be trained on data from comprehensive internal data sets and pharmaceutical-specific sources, such as patient records, claims, past trials, research publications, healthcare provider information, and diversity data.
- AI's success in clinical trials hinges on robust data sources, tailored language understanding, and mitigating hallucinations. Once these imperatives have been addressed, AI will be a powerful tool for clinical researchers.