Think AI for Health, and there is a tendency to think of remote diagnostics, AI-powered walking sticks and tricky operations being carried out over 5G from the other side of the planet.
Sadly, much of this dream is still in the future, but the UN initiative, AI for Health (managed under the umbrella of the ITU), provides a solid base from which to build that dream. There are, of course, deep concerns about AI generally, and AI in Health in particular. These are mainly about whether AI will deepen the digital divide and whether the whole AI journey is too risky.
That said, as Frederic Werner, project manager at the ITU points out, “AI is far too important to leave to the experts.”
Oddly, the initiative, far from suffering from the restrictions of the pandemic, has benefitted. Where pre-COVID, there was an annual gathering called ‘AI for Good,’ this event is now available all year round, with the spin-off ‘visibility’ benefits to sponsors and partners alike. Attendance has doubled, the number of countries involved has doubled, and the impetus has accelerated.
The sobering news is that the progress is not meteoric and nowhere near where the hype suggests it should be. Instead, we are still grappling (successfully) with the practical issues of digital exclusion, data management and open data exchanges. This against a backdrop of security and governance issues and the practicalities of enabling real innovation in this environment.
Encouragingly, stories are emerging of practical benefits of AI for Health initiatives. Moez Draief of Cap Gemini described great progress in predicting the availability of beds, availability of necessary, non-COVID related care and how they can monitor these trends at the patient level, not just by ward or department.
Soo Jun Park of ETRI described how South Korea has managed to get a grip on the pandemic, not necessarily just using AI, but using AI within already tried and tested systems that were enhanced to accelerate testing and tracing, CT scans and X-rays, to the point where communication with patients was down to 10 minutes, rather than hours or days.
One of the main issues remains the digital divide and what can be done about the 50% of the unconnected population.
The answer, according to Ulla Jasper of the Botnar Foundation, is to make it simple. If AI for Health, when it comes to diagnosis and monitoring, were as simple as an app on a smartphone, it would succeed in places with little ‘modern’ infrastructure.
Werner agreed and likened the possibilities for AI in Health in emerging economies to the miracle of mobile payments across, for instance, the African continent, which saw the enormous success of MPesa and other payment platforms. It is entirely possible that these countries could leapfrog more ‘advanced’ economies and enjoy the benefits that innovation can bring.
Of course, data is the key to AI for Health (or AI for anything else) and remains a major issue. Jaspar believes that there is not enough of a concerted effort in this area, and part of the problem is that there is so much data missing.
How and who manages the data is, of course, going to be an on-going (and intense) debate and Werner likened a solution to working with diamonds. You leave the diamond itself in a secure glass case and, using gloves attached to the sides, you work with the diamond, but you cannot take it away.
There is much to do and many issues to discuss. The good news is that the ITU initiative, AI for Health, with its numerous stakeholders and partners, is the perfect platform for building on the successes so far and delivering AI for Good.
What is clear, though, is that we must quickly move past the hype that is getting in the way of practical, solid progress.