Can AI algorithms used for advertising deliver benefits to the health sector?

There has been a lot of talk in the health community recently about the burgeoning role of artificial intelligence in understanding needs and supporting diagnoses at patient level and service planning at system level. As with most new technology, the lead-in time is long, but the benefits increment and grow exponentially, usually in an inverse proportion to cost, supporting technology efficiency and overall access. I personally think we are beyond the slow walk and are well and truly moving at a trot that's rapidly increasing to a gallop.

So we've talked a lot about big data, the role of AI in diagnosis, telehealth, radiology and the like but a TED talk I watched recently has expanded my view of the role of AI and its potential benefit to the population in the health context.

The talk, by Zeynep Tufekci, a Turkish 'techno-sociologist' (thanks, Wikipedia) can be found at https://youtu.be/iFTWM7HV2UI and is actually more directed at the risk of the AI algorithms in Facebook, Google and YouTube, where the underlying technology follows your on-line behaviour and learns what you may be interested in. This is the well-known reason why on-line ads seem to follow you everywhere.

Tufekci's main concern is the role of that AI in social media is now firmly influencing population and group-think relative to election and socio-political issues, and she presents a truly alarming argument. Her commentary around the notion that even those that wrote the algorithms now no longer know how they do what they do, suggests that the 'system' is now running on a learning auto-pilot, where the human creators can no longer keep up.

In amongst all this, it is now possible for the Facebook, Google and YouTube algorithms to predict behaviour at a person level. For example, as Tufekci comments, it is conceivable that the 'system' is now able to identify elements in a person's on-line behaviour that indicate that they are about to enter a manic phase, and then respond by offering ads that suggest great prices for that holiday in Vegas.

In this context, the clickbait mechanism may actually shortly be able to deliver significant diagnostic and population health information (not just data), because it is, actually, watching all the time and noticing things that a human could not. It's not self-aware (nod to Terminator fans), but it is now so complex that we are unable to look under the 'hood' and see what does and how it does it.

So my question is; how can this new, almost accidental, tool be wielded for good, and not just to assist you to watch the latest video, buy the boots, or book the holiday?

There is much food for thought in this (and more than enough for concern in the socio-political sense), but it may very well be that the greatest benefit in the big data context will come, not from health service data warehouses, but from the proprietary algorithms set up to deliver the targeted advertising to us that we all hate.