rabble_rouser

Brisbane. Why?

Beoz

Confirmed again. Brisbane is just a miserable place.

Kim67

Only if you're a misery guts. 😉

Beoz

There's a few of them in Brissy by the looks of it.

rabble_rouser

Brisbane is a tropical paradise.

Beoz

Is that the new slogan after -

"Beautiful one day perfect the next"

"The Smart State"

BadgeIsBack

You talk about deep learning : that has always existed for lawyers, doctors and computer programmers. Automation cannot diagnose everything or interpret case law.
To do so, you would need a grandiose points system and of course big data but many mistakes would be made : would you automate the appeals process?

My life coexists wth the sciences and the arts. I've noticed there are some very talented deep learners and scientists in the Arts who have been in the same industries that have been around for 50 years. But talk to a less rounded tech head or a classical geek about the arts and often they will look as if you are talking double-Dutch. It's as if all they know is Netflix, movies and games which is where they make sense of everything but it is not deep experience either. Being well rounded will continue to be an asset...

GarryP

OK, first off 'deep learning' is the application of deep neural network technologies to effectively 'pattern match' inputs with outputs. As I think I've mentioned before, it can do this at, or above, the level of a human. Couple that with other techniques and it can take over the jobs of a significant number of people. Basically if you can do a 10 week training course to do a job, the AI can learn the same rote stuff. Even the higher levels can often be automated.

As far as law is concerned, the expectation is that a significant percentage will get automated, partly because the rote stuff is so predictable, and partly because what the law is based on - defined cases and rules. You may have heard of the 'robot lawyer' that has helped overturn 200,000 parking tickets.

Doctors is another area where it is expected that large numbers of particularly GPs will lose jobs. Again, the case data is already available, as is structured knowledge, and already AIs are beating doctors in diagnosis.

And finally programmers, well there are lots of cloud based services coming out already to add AI smarts, just plug them together. And there are systems coming out to do the plugging together too.

Whilst 'well rounded' has a value - so does the sheer speed and ability to employ new ideas of those you decry. They break through "that's the way its done" and dream (and deliver) big. Throw in some face recognition, speech recognition, a spot of semantic analysis, and they have some working in an afternoon.

And if you look at where the world is getting better, rather than worse, it's generally those same people that are doing it.

Stinkypup

Im bored in Canada- so spotted this

I certainly think computers are in excellent aid to stratify risk. I was interested as to why you felt that GPs particularly would be at risk of losing their jobs. For example, that you gave was more cardiological rather than general practice. I am certainly not worried about my job. I spent my days, even without computers, trying to work out the motivation as to why some patients have come in to see me-it is quite often nothing to do with any medical condition!

Our work involves a huge amount of communication skills, I'm not blowing my own trumpet, and certainly use technology to help guide me. AI. I feel certainly currently is less able to either take a half decent history, make diagnosis, if necessary reassure the patient sufficiently.
In the future, progress, may be made that certainly I have no worries at the moment about job loss

BadgeIsBack

Good points, but certain case law cannot be distilled to points : you'd also have to capture the case correctly as well. Any lawyer will tell you case law can be complex where many are differentiated on small details of circumstance and intent. And again you can't just adjust the system for appeal. You could of course check the inputs once more but that is not an appeal against the decision itself.

GarryP

OK, well this is something I've done in the past - so I can talk on it.

First, and this was from a GP themselves, they said that 80% of patients coming to them had what they said they had or was obvious from symptoms. Came in with flu, had flu, was sent away with prescription. 16% needed a test undertaken to work out exactly was wrong (could be this, could be that), and 4% they had no idea and just tried stuff in the hope it would go away. That was borne out by stats, etc.

So, straight off, there's scope for automated diagnosis to cut 4/5 GPs. And actually it could probably manage a good part of that 16% too, so 9/10 maybe.

As for motivation and the worried well - that goes together with history and understanding the context. Both are within the scope of deep learning to recognise the patterns as well as the GP (and potentially do better, since you can have intercommunication between AI 'doctors').

In addition, GPs particularly have the problem that they are usually significantly out of date with current knowledge, and get corrupted by pharmaceutical companies. They also only know what they see, and so can't easily spot multiple instances across an area (which AIs can).

And as far as reassurance was concerned, there were two aspects. First was the authoritative voice, giving them confidence of what they had. The second was the compassionate voice, listening and telling them it was OK. And what we found was that the second tended to be best done by a nurse, and the first the doctor; but that it didn't actually rely on the human element, just tone, information, etc.

Now, looking at the types of patients, you can group them into those wanting someone to talk to and listen to them, those with something wrong wanting it fixed, and those that don't want to go anywhere near a doctor because they are bad news. And the thing that came out was that the second and third types were generally happier with a machine, and one probably in the home, than they were going off to a GP and waiting around for ages for the GP to give them 5-8mins total. The first were better off with a nurse.

Pull all that together and the upshot is that there is ample scope for automating 9/10 GP posts with an AI dealing with most of the rote stuff and referring on to a nurse for tests, etc., or a human GP for the difficult cases. The general level of care across the population would improve, and costs be cut. Automating data gathering was a separate area, but equally open to significant improvement.

I'd say the biggest impediment to it actually happening is not if it would work, but doctors protecting their empire and trying to make sure it doesn't get implemented. Certainly that's the experience that people have found (eg NHSDirect getting nobbled by the doctors). However, with costs rising as they are, eventually it will fall and there's a good chance it will switch quite fast.

GarryP

Point to bear in mind, it's not just being pushed from one end, there is also a desire to keep the costs of courts, etc. down by making sure things get sorted in a cheaper, more automated, fashion.

You can make significant inroads in the bread and butter stuff with automation (no paying for expensive lawyers, etc.) and you can deal with quite a lot of the criminal court as well. High court is more taxing, but even here the desire is to cut down on interminable grandstanding.

I wonder how long it will be before there is the first AI in court, but assuming it does well, it would open the flood gates. I'd guess 50+% of the total number of lawyers etc. could go. And indeed the researcher type role has already been decimated.

Here's an article on some of what is already being done, even before most of the deep learning advances hit :

Is artificial intelligence making lawyers a disappearing profession? | afr.com

PS Don't underestimate the ability of AIs to discern and make a case. They have been taking in the details of sports matches and churning out newspaper copy for a while now. Courts are no less formulaic.

Stinkypup

Please excuse my delay in replying-I was clearly carrying out a pointless and useless day of existence looking after my patients. My typing skills are clearly going to be limited as one hand has all day, been wrestled behind my back by "The Big Bad Pharma" and it is only now that the circulation is just returning…

Sadly, this confirms my suspicions. You may be clever in your field but this field certainly does not seem to incorporate health economics and clearly shows a significant lack of knowledge in what happens in general practice in the 21st century. You are clearly showing unnecessary symptoms of paranoia with your suggestion of doctors protecting their empire. I feel that my job is safe for a few more years yet- I welcome technology and use it every day in my practice.

NHS direct was the biggest waste of money aside from trying to centralise NHS clinical records, both a huge amount of money that should have been spent elsewhere which is sorely needed. I spent years having to receive information from hospitals because someone was acting on protocols at the NHS direct desk for example a child had a headache, this was therefore taken to be a likely sign of meningitis and they were subsequently transported by emergency ambulance to the hospital, at vast expense and totally unnecessarily. Telephone triage is quite a tricky thing to do and only I feel comes with years of experience and subtle nuances that one can determine when talking to patients/relatives on the phone give us a fairly good idea as to the urgency of the situation . Mindlessly going through lists of protocol questions to then subsequently request an unnecessary ambulance is I feel both expensive and pointless and it does no favours to the patient population.

ABD79

Having worked for the NHS myself, and with a husband who works in the healthcare profession, I have to say your example of a child with a headache being taken to hospital in an ambulance is a very far fetched scenario. I have used NHS Direct on a number of occasions, as you'd expect with three children over nearly 7 years, and they were extremely thorough in ensuring that all information was taken and the right information was collected in order to give the suggestion of me taking them in to see a doctor. I can quite categorically say from my own experiences phoning them that if I ever called with such a common thing as my child having a headache, they would not have called an ambulance out! In fact, my eldest was unwell at 8 weeks old. I called NHS Direct as he had been whimpering all night and by morning had a temperature of 40 degrees. They said they'd ask a doctor, which they did with me on the phone and the doctor came on the phone and asked us to take our little boy to the walk-in centre. We did so and it was there that the doctor called an ambulance as she had suspected meningitis. In hospital, after a lumber puncture and further tests, it was confirmed he had swine flu. He was very sick and in hospital for nearly two weeks. They'd just had a baby die that week on his ward from swine flu.
NHS is a fantastic service and from what my husband has said speaking to doctors, it has alleviated a lot of time being used up unnecessarily in medical centres.
This aside, I have to say this is way off topic of this thread. Unless I'm missing something about a doctor in Canada and his experiences there in the medical profession having anything to do with life in Australia?

Beoz

I wouldn't take Garry's synopsis of automation of GP's personally as you have. There is another thread on automation. You will see that Garry trips up in most places on this topic.

Automation has already done most of its job depletion to GP's. Dr Google.

And are GP jobs declining? No way. In fact every time I go to a GP, there's not an Australian GP in sight. All newly imported due to the GP shortage.

There's a few things will keep GP's alive.

1. Take your knee example. A GP can't fix a bung knee, but you need the GP for a referral to the specialist. The whole referral thing is a winner for GP's

2. Human and emotional interaction. You may have a physical ailment but humans need the human touch and reassurance from another human. On top of that humans need options and those options often revolve around ones personal and emotional circumstances. Can AI be taught that? Not in our lifetime. But if they did, and this is the scary part, you might be finding yourself emotionally attracted to bots.