AI in Healthcare: Panacea or Poison?

This session of Reuben College’s flagship Dining with Dinosaurs seminar series was hosted by Reuben’s own David Clifton, discussing the ever-more urgent prospect of Artificial Intelligence in Healthcare. The solution to all our problems, or a data leak waiting to happen? If there’s one person who should be able to help answer, it should be the Royal Academy of Engineering Chair of Clinical Machine Learning (David’s day job when not making excellent illustrations for his talks).

Some of the key criticisms of expanding the role of AI in healthcare relate to worries around misdiagnosis, malpractice, and compassion. If we replace doctors with computers, how can we ask for a second opinion, and who’s to blame if the machine is wrong? And who wants to hear bad news about their health from an uncaring machine? Fortunately, David was quick to dispel these fears, emphasizing that nobody wants AI to replace humans in the healthcare field. Rather, we should be aiming to use AI as a complementary tool to help improve the productivity of skilled - but often overstretched - medical workers.

My favourite line of the talk was his description of AI tools as a “co-pilot” that can help process vast amounts of data and identify key features for human doctors. David’s research group uses a combination of bespoke Natural Language Processing and knowledge distillation tools to create AI systems that can describe intricate relationships between diagnoses, medications, and medical procedures that would be impossible for a single person to identify. This allows doctors to compare each new patient against a vast historical dataset, which can summarise potential diagnoses and treatments and suggest the medications which have worked in similar cases, which would otherwise be impossible to identify.

But the AI isn’t just a co-pilot! David explained how AI can also be integrated into consumer wearables to increase the accessibility of healthcare tracking in low or middle income countries, providing the basis for ongoing research in collaboration with colleagues in Ho Chi Minh. Or it can be used in real time to evaluate complex healthcare challenges, such as the CURIAL AI which was developed during the COVID-19 pandemic to identify whether a new patient arriving at A&E had the disease within an hour of admission, based just on routine blood samples.

The picture that emerged from David’s talk was a reassuring one: not of a future in which robot doctors would make life and death decisions about patients based on data without discussion, but one in which AI tools provide a key complement to human expertise, allowing better decision making and faster, more reliable treatment based on limited resources that could revolutionise healthcare worldwide. Most importantly, Ai could help use eliminate problems caused by interfacing between medical professionals with different specialities - with the 3rd biggest killer in the USA being “Medical Error”, it’s clear that any AI that can help smooth the rough edges of the healthcare system is one that can help improve a lot of lives.