The Digital Delphi – Artificial Intelligence & Medicine

If knowledge is power, then from 7th century BC to the 4th century AD, the most powerful women of the classical world were undoubtedly the Oracles of Delphi.

Supplicants would travel far to seek the Pythia’s wisdom, delivered in ecstatic frenzy after inhaling the spirit of Apollo. Such was the cryptic nature of their utterances that prophets would be employed to help travelers make sense of their revelations.

Important decisions in matters of war, trade, marriage, and business, all made with reference to the divine knowledge imparted by the Oracle.

In matters of medicine, doctors have been the gateway to knowledge. With that, doctors have significant power in the war against disease.

On the afternoon of 22nd February I went to Digital Health.London’s ‘Collaborate’ event. It was a celebration of the programme’s first year’s work, an award ceremony, and innovation event all rolled into one. They were kind enough to let me speak about Virtual Reality in patient care as well.

Molly Watt & Dr Keith Grimes
Molly Watt & Dr Keith Grimes

There was a fantastic, diverse audience in attendance too, from the cutting-edge 360 live-streaming surgeon Shafi Ahmed, to my colleague Sunil Bhudia (with whom I’m working on the PREVENT-ICU-Delirium project), and the inspirational Molly Watt, who took part in a panel discussion about accessibility and digital participation. It’s also brilliant to meet people I know from twitter face-to-face, like Victoria Betton, who led a session on hacking STPs), and Dr Robert Lloyd, who skillfully MC’d the proceedings

So why the Delphi reference? Of all the talks, the most fascinating for me was the panel discussion on Artificial Intelligence.

Patients should use Artificial Intelligence to reduce the amount of time they need to spend with healthcare professionals

The panel comprised of three heavyweights in the Digital Health field:

  • Dr Ameet Bakhai Consultant Cardiologist Royal Free Foundation NHS Trust
  • Professor Nicholas Peters – Professor of Cardiology & Electrophysiologist at Imperial College London

Entertainingly and insightfully chaired by Dr Jordan Schlain , a fellow US based GP and Founder of HealthLoop, the debate fell nicely into those for, against, and balanced on the fence.

The AI Panel @ DH:L Collaborate
The AI Panel @ DH:L Collaborate – notice the #PinkSocks!

Ali Parsa began with a blistering defense of clinical AI, delivering an impassioned argument for how it can meet the yawning gap in healthcare provision around the globe. His team at Babylon have seen an 80% reduction in conversion of clinical inquiries to video consultations since the introduction of their triage AI. As far as he is concerned, there is a moral duty to implement AI to meet the care divide, and augment the diagnostic capabilities of clinicians.

It’s difficult to come back against that kind of rhetoric, although I would posit that Babylon’s UK business deals with a small proportion of range of presentations seen in General Practice, and from a significantly healthier and wealthier cohort. It might be difficult to extrapolate the 80% reduction in demand. With an ongoing trial front-ending GP in North London though, I guess time will tell.

The progress in developing world is also laudable, but I wonder whether it can ever cover the totality of care needs? Perhaps 10% of something better than 100% of nothing.

Next up, and taking an opposing view, was Dr Bakhai. He apologised for playing ‘devil’s advocate’, perhaps a Freudian slip given the perception that some have of the 5 year deal Royal Free London made in sharing data with Google Deepmind. For the sake of clarity, I think it’s a wonderful thing. Ameet worried about the loss of empathy in healthcare, and that AI had no ‘base case’ – no ‘first law of robotics’ akin to the principle of  ‘First, do no harm’

The issue of AI treating humans well and not turning them into batteries comes up time and again, and from my own point of view I swing between optimism and existential dread. It’s not unique to patient care, but does have an interesting twist in that should AI be successful in helping the patient, it may harm the doctor.

As for empathy, this is stronger ground. I certainly believe that humans need contact with other humans, particularly when it comes to the ritual of the consultation. Who’s to say that empathy is a uniquely human property though? Any pet owner will attest to the ability of their loved companions to deliver comfort without words. In time, empathy may well be better delivered by machine, especially to those raised in the post-millennial world – Homo Digitalis .

Professor Peters was left to adopt the middle ground, adding nuance to the preceding statements. I was particularly taken by his observation that the addiction that doctors have to treating patients, felt as a need deeper than simply a method of paying the bills, coloured their opinion. AI and the fear of being replaced and made worthless affect us all.

The framing of the debate made this last point ironic – there were no patients on the panel discussion the proposition. I raised this point, which was countered with an anecdote about Henry Ford who was quoted as saying that if he’d given the public what they wanted, he’d have made a faster horse. Professor Peters also referenced Molly Watts’ support of Apple’s commitment to user interface excellence & accessibility, which comes despite their famous lack of user input. All good points, but I’d still want to hear a patient make them.

The concept of Artificial Intelligence has been around since the 1950s, and we’ve seen this level of enthusiasm before. We’ve also had our hopes dashed before. I believe we’re entering a time when some of that promise will be realised, especially in narrow specialist areas. For this reason, I see the generalist physician outliving the specialist when it comes to head-to-head performance against AI in terms of patient care. Before my GP colleagues get too comfortable, even that advantage will pass in time, leaving the nurses as the most valuable humans in the healthcare system.

It may be that we will co-exist as providers of healthcare, but surely there will come a time when AI will be superior to human in a number of areas. When this happens, if we truly respect tradition of medicine, and believe we must first do no harm, maybe doctors should stop diagnosing patients?

And with that, is the best that doctors can wish for is to become gatekeepers of the knowledge – prophets at Delphi helping the patient understand the superhuman wisdom of the AI oracle? Or will this knowledge by available to the patients directly, leaving doctors shorn of their power?

It’s no surprise that our excitement is tinged with fear.

A STORIFY of my tweets from the day is also available

2 Replies to “The Digital Delphi – Artificial Intelligence & Medicine”

Leave a Reply

Your email address will not be published. Required fields are marked *