Hacker News Reader

Comment by Razengan | Hacker News Reader

Doctors in that sense, and similarly lawyers, seem like the best candidates to be replaced by ML/AI.

Why aren't we on this yet? (or are we?) That would be one of the best applications of AI, with wide benefits for civilians the world over but also causing social upheaval and thus facing resistance from the humans it attempts to supplant.

View on HN

In my experience, we don't need ML/AI to replace doctors. A good flowchart would do it. When your condition is on their standard list of diseases, the medical system (at least here in the UK) is a marvel of bureaucracy. Everybody works together with perfect coordination, to deliver the standard treatment at the lowest cost possible. Patient time or comfort isn't a priority, but it gets results. But as soon as you stray from their flowchart, the system is completely useless. Nobody is allowed to think for themselves, because any deviation from procedure would make the bureaucracy machine break down. As far as I can tell, the main purpose of the extensive doctor training is to raise barriers to entry and keep doctor wages up.

1. Because the health industry is about as regulated as it gets so even if ML/AI is possible there is zero chance of it being allowed to replace the advice of a doctor.

2. Because most people like seeing their doctor and talking to someone about their issues. There isn't exactly an appetite for disruption here.

To add to your second point: some conditions, such as mental health illnesses, require personal contact. Also, many people need to feel reassured when they're sick. An algorithm is not going to provide that, no matter how you dress it up.

I make reasonable use of: https://en.wikipedia.org/wiki/Dr_Fox_Pharmacy

When I want a medicine that requires a Dr's prescription, I can go and fill out a survey on their site. A real Dr (presumably) makes the actual prescription and my drugs show up. I am much happier for not seeing a doctor or talking to someone about my issues.

The survey I take isn't traditional ML/AI, but it's definitely AI, as it boots me out of the process if I answer some of the questions incorrectly.

Anyway, my point is: there's already disruption here, and it's awesome.

That's not awesome that you hacked the system to prescribe yourself anything you want. At best you're going to get the least efficient path to treatment. At worst you could miss a serious condition and kill yourself.

I am using the system directly as it is intended to be used, exactly as the UK government is intending it is to be used. I'm not sure what you think the hack is here. How the process works is detailed here:

https://www.doctorfox.co.uk/how-it-works.html

And you can see how they're regulated in the UK here:

https://www.doctorfox.co.uk/regulation.html

> 1. Because the health industry is about as regulated as it gets so even if ML/AI is possible there is zero chance of it being allowed to replace the advice of a doctor.

i.e. the necessity of an identifiable scapegoat if things turn out badly

> 2. Because most people like seeing their doctor and talking to someone about their issues. There isn't exactly an appetite for disruption here.

Maybe a certain type of person who craves an authority figure of some sort to soothe their fears, but I'd have to see evidence of this. Most people I know hate going to the doctor, and dislike and distrust the doctor they end up seeing as often as not. Having a friendly doctor that you've been seeing most of your life, that other members of your family see, and that you get Christmas cards from is a class thing (at least in the US) that maybe 10-15% of people get.

> There isn't exactly an appetite for disruption here.

Apple has already begun doing it in baby-steps, doing it well so far and is lauded for it.

> The first expert systems were created in the 1970s and then proliferated in the 1980s. Expert systems were among the first truly successful forms of artificial intelligence (AI) software.

~https://en.wikipedia.org/wiki/Expert_system

Applications include diagnosis and some systems are listed: CADUCEUS, MYCIN, PUFF, Mistral, Eydenet, Kaleidos.

Facebook announced that they are using ML to fix compilation errors just yesterday. They said they were contracting out companies to build ASICS for the AI. We are a long way from AI making good medical decisions.

Reply on news.ycombinator.com