I think we need to think more systematically about healthcare in general.

We operate under the illusion (or delusion!) that there's a system where there isn't. Healthcare is a collection of lots of bits and pieces, under the control of different entities great and small, loosely strung together by funding and bureaucracy, but mostly disconnected where it hurts patients most. The reason it's flawed is it was never designed as a system, never built as a system, never even conceived as a system. If we don't have a system to start with, thinking about AI systematically would be like putting a posh frock on an orangutan - a lot of hard work to achieve nothing.

Actually, if I ruled the world (I can dream!) - the first task I'd give AI would be identifying how best to start connecting all those parts we currently have into a functioning system. Give it the arduous part of mapping and streamlining processes and pathways that make sense, but it's programming would be populated with the criteria for what humans value and need. And let us - patients, caregivers, families, physicians, nurses, policy makers etc. - concentrate on the relational part that will allow it to happen.

Now let me grab my wand and I'll make sure that orangutan can dance in glass slippers, and make it the belle of the ball

https://www.bbc.com/news/articles/cy8gy7lv448o

'I can't drink the water' - life next to a US data centre

The sound of a data center:

https://www.youtube.com/watch?v=gc5XZJfF0kQ

https://www.eesi.org/articles/view/communities-are-raising-noise-pollution-concernsabout-data-centers

I think one positive point I take from his analysis is that we need to think more systemically ie around workflows, about AI or indeed an digital application in order to implement it with noticeable benefit.

I have to take this with a pinch of salt, well probably more salt than is good for me I think framing AI in this way is problematic.

We tend to think of infrastructure as benign, a solid investment. We don't tend to think of infrastructure making decisions for us, making mistakes, needing constant monitoring, tweaking, retraining. We certainly don't think of infrastructure as compromising safety or eroding our rights. By reframing AI as infrastructure, we mask a lot of concerns and risk.

We also expect infrastructure to cost a lot, infrastructure projects are usually big, and take a long time to come to fruition, but then we expect the money we've ploughed into it to give us an asset that lasts for decades. By reframing, we make a bigger spend seem natural without making it clear that we won't necessarily get a long-term asset.

Electricity is a fantastic reframe - most of us can't even imagine life without electricity anymore - it ranks right up their with air and water. By reframing AI as electricity it becomes essential, the kind of thing we demand and lobby for as a right, something we fight for access to. But AI has yet to demonstrate the kind of consistent or essential value that's worth fighting for yet.

Perhaps - and this is a big perhaps - AI will eventually live up to these framings. But we're not there yet, we're a long way back from being even close.

Let's look at a different frame: From my perspective, AI is still - for all its seemingly magical properties, and all the bells and whistles - AI is still a toddler, not even out of nappies. It needs coddling and regular diaper changing. It needs careful parenting and proper schooling.
It's a precocious toddler, some see it as a prodigy, some as a savant - but despite the ability to parrot language and perform impressive maths, it has no understanding of the world it exists in, and no understanding of human beings or how we behave. It can't tell fact from fiction, because it doesn't understand either of those concepts. That improbable mix of precocious intelligence and unadulterated innocence is unpredictable at best.

So we reframe the unpredictable as something solid and reliable, as something essential and necessary - and as something that most people will feel confident they understand - infrastructure and electricity. The approach is glossy and compelling - dressed up with slick infographics and slicker podcasts - but also dangerous because it seeks to distract us from looking at the basic facts. Investing in AI on a large scale may be necessary, but we need to be transparent about what we can expect, and what it will cost us. And by cost, I don't just mean in terms of dollars, I mean in terms of what other priorities may need to be shelved or abandoned to get there. These will be hard discussions to have and harder decisions to make, and glossy reframes don't help because they add unnecessary confusion to a world already swimming in misinformation.