Entry 2441, on 2026-05-06 at 20:50:14 (Rating 3, Computers)
One of the biggest problems in modern society is poor provision of health services. This is partly to do with the economics of health, including the cost of producing new drugs and medical devices, partly to do with government policies encouraging poor organisation of services, partly to do with incompetent and excessive management structures in health, and partly to do with an ageing population in many countries and the expectation for more conditions to be treated.
There are several approaches which have been taken to try to improve the problem, such as greater investment from health funders, attempts at reorganisations with the intent of gaining greater efficiency, and changing the rules to allow some health services to be delivered outside of traditional hospitals and medical practices.
But, despite all of this, there has never been an obvious and substantial improvement in outcomes, although occasional small successes do seem to occur.
Needless to say, I have the answer, as I'm sure you have probably already guessed. And considering my background, you won't be surprised to hear the answer comes from the world of IT, and especially AI.
One of my favourite podcasts is called "Freakonomics". It covers how economics, especially behavioural economics, affects the real world, and how real people behave in complex societal and economic environments. When I say "behavioural economics" and "real people" I am emphasising that there is an attempt to examine what happens in the real world where people don't necessarily act rationally or logically. In many traditional economic studies it is assumed people will react rationally and for their own greatest benefit. Of course, this is not true, and economics must take account of psychology.
Freakonomics did a series of episodes on AI recently, which increased my confidence in the technology. And yes, you might know based on previous posts, I am already a fan of it. There are various fields where AI might be particularly useful, and medicine is near the top of the list.
So moving on to how AI might be used in health. First, we need to acknowledge that it has only become a useful and widely used tool for most people in the last few years. This means we are really at the start of producing AI tools which are genuinely useful and reliable, and we can expect massive improvements over the next 5 to 10 years.
Second, we shouldn't dismiss AI because of any real or perceived defects without comparing it to the alternatives. For example, a fleet of self-driving cars were tested and were involved in 2 accidents over the time they were tested. These were reported widely and resulted in the program being suspended. But based on statstics, during the same period we should have expected about ten times as many accidents if the cars were driven by humans. Also, in both accidents the automated cars were not primarily to blame.
I'm a "car guy" and I often enjoy driving, but I can see a time (maybe 10 years away) where the idea of a human driving a car is seen as absurd. They will look back at the deaths and injuries we have on our roads and wonder "what were they thinking?"
So can this apparently major improvement be applied to medicine too? Well yes, of course it can. In fact medicine might be one place where AI offers the biggest potential technical improvement. I say technical because it is something which can be done if we act based on facts, but the real problem with be overcoming the bureaucracy and conservatism of government regulation, management, and the health workforce.
Many experiments and trials have shown how good AI is at various medical activities. It can analyse scans, x-rays and other tests better than humans, it can find information in hundreds of pages of medical records instantly and use it in context, it can diagnose some conditions based on simple data better than even the best specialist, in one test is even delivered bad news of a patient's death to relatives better than a doctor!
The tech companies creating the AI tools say they want to produce something that medical professionals can use to make their jobs easier, but I'm sure many of them really think those professionals can be replaced completely.
But what if something goes wrong? We know AIs can make mistakes, and even "hallucinate". What happens if an AI kills a patient after making an error? Who's to blame? Well, doctors already make a pile of mistakes and we cope with that. AIs would also make mistakes, but probably a lot less, so surely some mechanism for assigning responsibility and improvement could be developed.
I rarely go to a doctor myself, in fact I can't remember the last time I did, but I always knew the diagnosis and treatment before I went, thanks to simple web searches using "Doctor Google"! I can imagine I could do a lot better with today's AIs instead (and I will definitely be doing that in future). Do I not trust doctors? Well, to some extent I do, but I would prefer an AI any day!
There are no comments for this entry.
Thanks for reading this blog post. Please leave a message below.
You can leave comments about this entry using this form.
To add a comment: enter a name and email (optional), type the number shown, enter a comment, click Add. Note that you can leave the name blank if you want to remain anonymous. Enter your email address to receive notifications of replies and updates to this entry. The comment should appear immediately because the authorisation system is currently inactive.
I do podcasts too! You can listen to my latest podcast, here: OJB's Podcast 2026-04-25 Think For Yourself: In the end, everything is just an opinion. Be skeptical, and think for yourself! Or subscribe to my podcast RSS feed, on my RSS Feeds page.
Latest Site News and Notifications (Desktop): You are currently viewing OJB's web site, version 2.4 which has some major changes, and possibly errors! Please report any problems to ojb@mac.com.
BLOG PAGE