Key pannellist included: 

• Subrata Bose Vice President, Diagnostic Imaging, Data & Artificial Intelligence, Head General Clinical Imaging Services, Bayer AG
• Patty O'Callaghan, Technical Director, Charles River Laboratories 
• Mark Barber, AI Product Lead Director, AstraZeneca 
• Moderator: Lara Lewington, Journalist, Broadcaster and Author, BBC

Mark Barber discussed how AI at AstraZeneca is both enhancing existing functions and revolutionising R&D through technologies like genetic AI and multimodal AI. Subrata Bose highlighted the complexity of drug development, noting AI's ability to reduce cycle times and costs, thereby increasing the hit rate in drug discovery. AI's role in repurposing old drugs was also mentioned. However, despite these advancements, AI in drug discovery is not always successful, and the process remains complex and costly.

Patty O'Callaghan emphasised the challenges of integrating AI into scientific practice, particularly the lack of trust among scientists and medical professionals. She shared an anecdote about a medical doctor who, despite recognising AI's precision, hesitated to use it due to a lack of trust. This underscores the importance of explainable AI models that scientists can trust. The panel also discussed the human response to AI, with Mark Barber noting the necessity of involving radiologists in AI development teams to ensure better adoption and trust. Subrata Bose added that AI could handle administrative tasks, but critical decision-making should still involve humans.

The discussion also touched on the importance of data in AI projects. Subrata Bose pointed out that the majority of AI projects fail due to the lack of proper data, stressing the need for diverse, high-quality datasets. Patty O'Callaghan advocated for AI literacy to help the general public understand how their data is used. Mark Barber emphasised transparency in clinical trials to maintain patient trust. The panel agreed on the potential of AI to tackle complex diseases like cancer, but noted that significant progress is still needed. AI's ability to integrate various data sources could lead to better disease prediction and personalised treatments, enhancing overall healthcare outcomes.

Takeaways

AI can significantly speed up drug discovery

Subrata Bose highlighted that AI reduces the drug development cycle time and costs, increasing the hit rate in drug discovery. Despite its potential, AI in drug discovery is not always successful, and the process remains complex and costly. AI's role in repurposing old drugs also offers promising avenues.

Trust in AI remains a significant barrier

Patty O'Callaghan discussed the lack of trust among scientists and medical professionals in AI tools, even when they recognise their precision. This underscores the importance of explainable AI models that can earn the trust of users. Involving professionals like radiologists directly in AI development teams can help bridge this trust gap.

High-quality, diverse data is crucial for AI success

Subrata Bose stressed that over 80% of AI projects fail due to inadequate data. The need for diverse, high-quality datasets is crucial for training effective AI models. Transparency and AI literacy are essential to maintain patient trust and ensure the ethical use of data in AI projects.