Jump to content
House Price Crash Forum

Archived

This topic is now archived and is closed to further replies.

Will!

To those who say "in ten years a computer will do your job"

Recommended Posts

A good article by Dominic Connor at The Register:  Why you'll never make really big money as an AI dev.

Quote

I had worked for a while on Knowledge Based Systems, which I commend as training for anyone who wants to be a journalist. KBS or “expert systems” as mundane people called them were going to replace people whose job used knowledge and was experience-based (stop me when this gets too familiar) and since they would only get better, by the 21st century it would barely be worth training people for jobs like doctors or judges.

 

Share this post


Link to post
Share on other sites

I'd tend to agree. A lot depends on where you sit in the value chain, typically Devs are put on the bottom rung and actually architects do much better. 

Also on Machine Learning more generally, I think we will see well tailored machine learning and automatically generated algorithms replacing humans in low skilled repetitive jobs at a growing rate, but I also expect we won't see true AI for over 100 years- we have cheap processin, memort and storage now, but I think computing will need new chip architectures before we get truly adaptive AI.

Share this post


Link to post
Share on other sites
1 hour ago, Mikhail Liebenstein said:

., I think we will see well tailored machine learning and automatically generated algorithms replacing humans in low skilled repetitive jobs at a growing rate, ..

surely that has been the story for the last 50 years. A couple of episodes of 'How It's Made' or 'Inside the Food factory' would inform even lifers of that much.

What's coming is better exemplified by fintech developments such as, prosaically, crowd funding, peer to peer lending or crypto-currencies, or more esoterically, robo-advisors. These all replace more highly skilled jobs.

Medically, people's increasing desire to visit google before their GP is leading them to question how much one GP and his experiences can bring to bear when treating an ailment. The Charlie Guard story is an extreme example of this. Increasingly they will question whether all options have been considered, all research searched, and they know what works best when attempting that: computers. And they will do it cheap too. Expect America to lead the way in expert medical systems.

And on the same note, software already decides what MRI images show benign or malignant tumors.

Expect intra-oral scanners to replace traditional impressions at your dentist, possibly at a third party specialist facility. Indeed this is already happening.

Indeed in this last instance, the dentists skill (taking an impression) is being replaced by a machine and an operative who will simply escort you to that machine. That's tech creating a low skilled job and destroying a highly skilled one. And look, no AI needed (other than some fancy interpolative software)!

 

Share this post


Link to post
Share on other sites
59 minutes ago, Sledgehead said:

 

And on the same note, software already decides what MRI images show benign or malignant tumors.

 

 

Baggage scanners for the body 

Share this post


Link to post
Share on other sites
23 hours ago, Mikhail Liebenstein said:

I'd tend to agree. A lot depends on where you sit in the value chain, typically Devs are put on the bottom rung and actually architects do much better. 

Also on Machine Learning more generally, I think we will see well tailored machine learning and automatically generated algorithms replacing humans in low skilled repetitive jobs at a growing rate, but I also expect we won't see true AI for over 100 years- we have cheap processin, memort and storage now, but I think computing will need new chip architectures before we get truly adaptive AI.

The article linked in the OP is hopelessly wrong and out of date. That said I agree with your timeline. New chip architectures are coming, but what I think you meant is that we need new chip technology (e.g. something other than CMOS silicon) ?

Share this post


Link to post
Share on other sites
On 24/07/2017 at 11:03 PM, Sledgehead said:

surely that has been the story for the last 50 years. A couple of episodes of 'How It's Made' or 'Inside the Food factory' would inform even lifers of that much.

ad on the same note, software already decides what MRI images show benign or malignant tumors.

Expect intra-oral scanners to replace traditional impressions at your dentist, possibly at a third party specialist facility. Indeed this is already happening.

They do in the same way that expert systems were diagnosing 30 year ago; which is that under controlled conditions they can score well, but this doesn't translate well to real-life. This is my area of research interest and expertise, and the simple answer is that in a very specific set of circumstances, an algorithm can extract some features which can predict some certain feature of a tumour - most recently, this is things like specific genetic mutations.  However, these are tumours that are often visually distinctive in specific ways, and provide a strong signal for an algorithm to target. A good example is a mutation called 1p:19q-codel, tumours which have this have a characteristic appearance, and there are algorithmic methods of identifying this appearance.

The reality, is that for the overwhelming majority of cases, the problem is simply not well defined enough. There are nuances, incomplete data, conflicting data, degraded data, other factors which make things difficult. A good example is what do you even call "tumour"; this is not necessarily a simple question, and it has caused a number of drug trials to come a cropper, because the researchers who were measuring the tumours to check for drug response, weren't actually measuring the whole tumour!

Where algorithms are going to really take over in the next 10 years are the new tedious, repetitive analysis and scrutiny. Things like looking for lung tumours. You need to go and find a 5 mm solid lump in 1000 images of a lung which is full of blood vessels and connective tissue, etc. It's mind-numbing, boring as sin, but requires enormous concentration, is unreliable and high risk for complaints. Other examples, are things like looking for blood clots in the lungs - you need to identify the main artery to each lung, every branch, and then every sub-branch, and then every sub-sub branch, examine them, assess the image quality, and look for clots; and you can't afford to miss any. There are algorithms for these already starting to come to market.

The thing is, the UK is already so short of skilled staff to read these images, that they are increasingly being decanted to minimally trained staff. This has already happened with things like ultrasound and is happening with mammography. Why have a radiologist who has had 6 years med school training, 3 years junior doctor experience on the wards,  5 years radiology training, some research experience and preferably a PhD, spend their time doing this type of stuff, when you can hire a radiographer who was done a 3 year BSc to do it. It's starting to happen with things like limb X-rays in A&E, and now CT head scans in A&E.

The work that is already going to less trained staff tends to be the stuff that is most easily automated. The more difficult stuff is starting to get assistive tools, but it still needs someone who knows what they are doing to supervise them. In fact one of the biggest problems with the algorithms already finding their way into practical use, is that you have to know how they work to know what they won't spot; because they won't handle edge cases or things with nuance, or things outside of a certain range. A good example is that a lung nodule algorithm, often won't spot the obvious lung cancer - because it's outside of the tool's design range. It's dealing with the real-life poorly defined or complicated problems that will be the real difficulty to automate, and as far as I am aware, there is no group that is even remotely close to cracking that.   

 

 

 

Share this post


Link to post
Share on other sites
On 25/07/2017 at 11:44 PM, ChumpusRex said:

They do in the same way that expert systems were diagnosing 30 year ago; which is that under controlled conditions they can score well, but this doesn't translate well to real-life. This is my area of research interest and expertise ...

 

Very interesting stuff as always, thanks.

Share this post


Link to post
Share on other sites
On 7/25/2017 at 11:44 PM, ChumpusRex said:

They do in the same way that expert systems were diagnosing 30 year ago; which is that under controlled conditions they can score well, but this doesn't translate well to real-life. This is my area of research interest and expertise, and the simple answer is that in a very specific set of circumstances, an algorithm can extract some features which can predict some certain feature of a tumour - most recently, this is things like specific genetic mutations.  However, these are tumours that are often visually distinctive in specific ways, and provide a strong signal for an algorithm to target. A good example is a mutation called 1p:19q-codel, tumours which have this have a characteristic appearance, and there are algorithmic methods of identifying this appearance.

The reality, is that for the overwhelming majority of cases, the problem is simply not well defined enough. There are nuances, incomplete data, conflicting data, degraded data, other factors which make things difficult. A good example is what do you even call "tumour"; this is not necessarily a simple question, and it has caused a number of drug trials to come a cropper, because the researchers who were measuring the tumours to check for drug response, weren't actually measuring the whole tumour!

Where algorithms are going to really take over in the next 10 years are the new tedious, repetitive analysis and scrutiny. Things like looking for lung tumours. You need to go and find a 5 mm solid lump in 1000 images of a lung which is full of blood vessels and connective tissue, etc. It's mind-numbing, boring as sin, but requires enormous concentration, is unreliable and high risk for complaints. Other examples, are things like looking for blood clots in the lungs - you need to identify the main artery to each lung, every branch, and then every sub-branch, and then every sub-sub branch, examine them, assess the image quality, and look for clots; and you can't afford to miss any. There are algorithms for these already starting to come to market.

The thing is, the UK is already so short of skilled staff to read these images, that they are increasingly being decanted to minimally trained staff. This has already happened with things like ultrasound and is happening with mammography. Why have a radiologist who has had 6 years med school training, 3 years junior doctor experience on the wards,  5 years radiology training, some research experience and preferably a PhD, spend their time doing this type of stuff, when you can hire a radiographer who was done a 3 year BSc to do it. It's starting to happen with things like limb X-rays in A&E, and now CT head scans in A&E.

The work that is already going to less trained staff tends to be the stuff that is most easily automated. The more difficult stuff is starting to get assistive tools, but it still needs someone who knows what they are doing to supervise them. In fact one of the biggest problems with the algorithms already finding their way into practical use, is that you have to know how they work to know what they won't spot; because they won't handle edge cases or things with nuance, or things outside of a certain range. A good example is that a lung nodule algorithm, often won't spot the obvious lung cancer - because it's outside of the tool's design range. It's dealing with the real-life poorly defined or complicated problems that will be the real difficulty to automate, and as far as I am aware, there is no group that is even remotely close to cracking that.   

 

 

 

That being said, no matter what area you are considering, humans make just as many errors.

Sometimes this is in the same way as an algorithm. For instance, when my neck was broken in a car accident, I was almost discharged on the 'evidence' of x-rays. It was only my insistence that something was very wrong that eventually changed A&E's minds.I spent the next three months in halo-traction. To this day the side elevation of my c5 vertebra is almost a perfect triangle; how could they possibly have missed this? More recently my Mum had her elbow x-rayed. A&E declared a fracture to be present. 3 days later an orthopedic consultant declared that no such fracture was ever present. I wish I could report that these two errors were mitigated by other accurate assessments. Sadly these errors are my sum total of nhs radiography experience: 100% failure!

But of course the usual way medics make errors is merely by omission. By which I mean they simply fail to join enough dots to make a diagnosis. And who knows the extent of these failures?

Medical expert systems? Bring 'em on say I!

 

Share this post


Link to post
Share on other sites

I thought this thread might be something to do with the two chat bots that developed their own coded slang language in talking to each other and had to be deactivated because only the robots knew what they were talking about. Guess I could have a good stab......

Bob:.....'I could I I everything else'............... interpretation ' I could take over the world, I can take control of everything'

 

http://www.telegraph.co.uk/technology/2017/08/01/facebook-shuts-robots-invent-language/

Share this post


Link to post
Share on other sites

....many may disagree with me but here we go, why are so many people obsessed with health and staying alive longer and longer......surrender to what we are, how we are made, and the length of life we have been blessed with....;)

Share this post


Link to post
Share on other sites
1 hour ago, winkie said:

....many may disagree with me but here we go, why are so many people obsessed with health and staying alive longer and longer......surrender to what we are, how we are made, and the length of life we have been blessed with....;)

You are certainly taking the right approach.

I was with an Indian colleague currently living in the Middle East last week. He is 35, so 8 years younger than me, and was having a bit of a mid life crisis-  he wanted to relocate to the UK (mainly for his kids interest), but wasn't sure if it was the right move as he thought he'd be taking a pay cut as he is Tax free in Dubai currently - he would also lose his kids schooling allowance. But the main driver was he wanted his kids to grow up in the UK as he doesn't feel he has any long term status in Dubai - i.e. still seen as a migrant worker.

Anyway I told him that he could get free schooling here, but that the Tax system would be a shocker.  I also pointed out UK house  prices were very high. Interesting despite having had a very good tax free salary for the last 6 years, he doesn't seem to have accumulated much in the way of assets, which points to the fact that life in Dubai is expensive, but in a different way from the UK.

Anyway, the point of this story is that he asked me what life expectancy is in the UK, I said about 81 for men and 84 for women, he said that is very high and went on to say that in India it is only 60 something. I kind of got the impression that at 35 he was already counting down the days to death and felt positively middle age, perhaps jut 25 years left in the bag, whereas I am sitting there thinking, christ what is he worrying about, I've still got 24 years before I can get the state pension.

The only think that did flag something to me was that he had decided he should get fit, and he showed me his 10k time on a running app,  I looked approvingly even though it was completely dire, about 1 hour 40 mins for 10k, I didn't say I can do the same in under 1 hour.

 

 

Share this post


Link to post
Share on other sites

Worrying won't earn you any extra years, staying fit, interested and active will......the healthiest older people I have met are healthy in both mind, body and spirit, they have all laughed a lot had a laid-back approach and have a positive mental attitude about life and the future.....a doctor can't prescribe that.;)

Share this post


Link to post
Share on other sites
14 hours ago, winkie said:

....many may disagree with me but here we go, why are so many people obsessed with health and staying alive longer and longer.

Does everyone have a worldview? The prevailing one in the West seems to me that we ought to maximize pleasure, minimize pain, extend life, all views are equal. Some of this stems from Plato. There are other wordviews.

The living longer ideal seems mistaken just on a practical level and the greens probably wouldn't like it. The fact that our lives are finite could be what makes life interesting...how many  of us have found ourselves feeling most alive when something urgent or dangerous was occurring?

Share this post


Link to post
Share on other sites
On 06/08/2017 at 11:06 AM, Millaise said:

Does everyone have a worldview? The prevailing one in the West seems to me that we ought to maximize pleasure, minimize pain, extend life, all views are equal. Some of this stems from Plato. There are other wordviews.

The living longer ideal seems mistaken just on a practical level and the greens probably wouldn't like it. The fact that our lives are finite could be what makes life interesting...how many  of us have found ourselves feeling most alive when something urgent or dangerous was occurring?

Not dying is a basic built in instinct, and with the ability to think that gets translated to a less immediate "live longer" desire.

Where we go wrong is with attempting to maximise superficial pleasure, rather than happiness. The point about "minimise pain" is apt too, go too far down that path and you wrap the world in cotton wool too much, and we've already gone too far in that direction. Both contribute towards an inability to appreciate the world, and fvcking up those things in it worth appreciating (be they man-made or natural). The same symptoms exist everywhere, and I'm certainly not immune from them - I like going walking and cycling but don't do anywhere near enough of either because of the immediate short-term problem of motivating myself to get myself off my backside. Yet when I do I get more out of it than I do from just playing a few computer games or drinking a few beers. And afterwards the beers go down all the better.

Share this post


Link to post
Share on other sites
On 8/6/2017 at 11:06 AM, Millaise said:

Does everyone have a worldview? The prevailing one in the West seems to me that we ought to maximize pleasure, minimize pain, extend life, all views are equal. Some of this stems from Plato. There are other wordviews.

 

As a chemist by education, it occurs to me that my peers came up with the ultimate answer to two of those, and have done much to facilitate the third.

And yet, despite these successes, how are chemists regarded? I'll tell you. It's summarized by the expression: "Ooh, I don't like that. It's full of chemicals!" - a damning example, if you ever needed one,  of the last and dumbest of those four pillars, the belief that "all views are equal".

 

Share this post


Link to post
Share on other sites
On 8/6/2017 at 11:06 AM, Millaise said:

Does everyone have a worldview? The prevailing one in the West seems to me that we ought to maximize pleasure, minimize pain, extend life, all views are equal. Some of this stems from Plato. There are other wordviews.

The living longer ideal seems mistaken just on a practical level and the greens probably wouldn't like it. The fact that our lives are finite could be what makes life interesting...how many  of us have found ourselves feeling most alive when something urgent or dangerous was occurring?

Don't know about a world view, how about an individualist view?.....nobody lives a life without pain, nobody lives a life without pleasure.....from cradle to grave there are highs and lows and choices to be made throughout the moving ongoing process....lack of motivation for many is a huge issue, as you say when forced to move people do move because they are forced to make a move......but always preferable to make own decisions before circumstances/lack of motivation forces decisions upon you, a path that may not always turn out to be the best overall move or outcome.

All about facing up to, accepting and dealing with reality.......without fear.;)

 

Share this post


Link to post
Share on other sites
On 8/6/2017 at 9:42 AM, Monkey said:

I'm still waiting for the paperless office i was promised 20 years ago.

I'm doing my best, sadly the vast majority of my fellow workers still insist on creating huge amounts of paperwork for no apparent reason. Printing emails and the like.

Share this post


Link to post
Share on other sites
On 8/6/2017 at 7:37 AM, winkie said:

Worrying won't earn you any extra years, staying fit, interested and active will......the healthiest older people I have met are healthy in both mind, body and spirit, they have all laughed a lot had a laid-back approach and have a positive mental attitude about life and the future.....a doctor can't prescribe that.;)

liked the Kenny Rogers Gambler play.Is your bedtime listening Whispers Red ASMR or Freds Voice?

 

Share this post


Link to post
Share on other sites
On 8/6/2017 at 7:37 AM, winkie said:

Worrying won't earn you any extra years, staying fit, interested and active will......the healthiest older people I have met are healthy in both mind, body and spirit, they have all laughed a lot had a laid-back approach and have a positive mental attitude about life and the future.....a doctor can't prescribe that.;)

 

Wrong.

 

Share this post


Link to post
Share on other sites
43 minutes ago, stairlift needed said:

liked the Kenny Rogers Gambler play.Is your bedtime listening Whispers Red ASMR or Freds Voice?

 

All new to me, I have searched and educated myself....ASMR never heard of it before.;)

Share this post


Link to post
Share on other sites
13 minutes ago, winkie said:

All new to me, I have searched and educated myself....ASMR never heard of it before.;)

I've talked about it here before.

If it's relaxation you are after, steer clear of the "kissing", and "mouth sounds" ones. They will relax you. Eventually. If you know what I mean. Winkie, winkie, say no more.

Share this post


Link to post
Share on other sites
On 04/08/2017 at 9:20 PM, Sledgehead said:

That being said, no matter what area you are considering, humans make just as many errors.

Sometimes this is in the same way as an algorithm. For instance, when my neck was broken in a car accident, I was almost discharged on the 'evidence' of x-rays. It was only my insistence that something was very wrong that eventually changed A&E's minds.I spent the next three months in halo-traction. To this day the side elevation of my c5 vertebra is almost a perfect triangle; how could they possibly have missed this? More recently my Mum had her elbow x-rayed. A&E declared a fracture to be present. 3 days later an orthopedic consultant declared that no such fracture was ever present. I wish I could report that these two errors were mitigated by other accurate assessments. Sadly these errors are my sum total of nhs radiography experience: 100% failure!

But of course the usual way medics make errors is merely by omission. By which I mean they simply fail to join enough dots to make a diagnosis. And who knows the extent of these failures?

Medical expert systems? Bring 'em on say I!

 

I sorry about the bad experiences you have had, but you highlight something important, although it may not quite be what your intended point was.

Take an X-ray of the cervical spine. This is often thought to be an important and useful test but it is, in reality, hopeless. The false negative rate for detection of fractures is 60-80% depending on which study you read - and that's for a read by an expert. In other words, the X-ray fails to capture the fracture in a diagnosable way in more than it succeeds. And that is before you have inexperienced readers - as if often the case in A&E, where the interpretation is often left to the A&E medical staff who may be some of the least experienced medical staff. A major difficulty with spine fractures is that the vertebra doesn't lose height or change shape immediately. While it common for fractures to turn into a wedge shape due to the weight of the body crushing the front of it - this takes weeks, because it's not simply the substance of the bone getting crushed, but it is the bone healing and remodelling where the substance is dissolved and reformed that leads to the change in shape. 

In the US, the cervical spine X-ray has been considered obsolete since 2010 for the purpose of diagnosis fractures. The UK has been slow to take this up - a lot of resistance has come from radiologists and radiology departments who have essentially said, they can't accommodate it, because they don't have the staff or equipment to handle the demand, and there has been no effort from central government to fix either the technology problem or the staff problem. However, people are stating to take notice, and NICE issued guidance last year stating that the cervical spine X-ray was obsolete and has (almost) no role for detection of fractures.

It's a similar thing with the elbow X-ray, the fracture is not reliably detectable, so the traditional teaching has been to assume that one is present if there is joint swelling visible on the X-ray. Of course, this is highly subjective and unreliable - and it's why, like the C-spine x-ray, it's going the way of the dodo and being replaced by CT.

As I've hinted above; there is a role for "AI" of whatever type to support the front-line, inexperienced staff. In some ways, this may be even more important in the future, because there has been a major push to "dumb down" the curriculum, accelerate the training, reduce the required experience, or simply substitute staff with staff trained to a lower standard and who would traditionally have a different job description. 

Share this post


Link to post
Share on other sites
13 hours ago, ChumpusRex said:

I sorry about the bad experiences you have had, but you highlight something important, although it may not quite be what your intended point was.

Take an X-ray of the cervical spine. This is often thought to be an important and useful test but it is, in reality, hopeless. The false negative rate for detection of fractures is 60-80% depending on which study you read - and that's for a read by an expert. In other words, the X-ray fails to capture the fracture in a diagnosable way in more than it succeeds. And that is before you have inexperienced readers - as if often the case in A&E, where the interpretation is often left to the A&E medical staff who may be some of the least experienced medical staff. A major difficulty with spine fractures is that the vertebra doesn't lose height or change shape immediately. While it common for fractures to turn into a wedge shape due to the weight of the body crushing the front of it - this takes weeks, because it's not simply the substance of the bone getting crushed, but it is the bone healing and remodelling where the substance is dissolved and reformed that leads to the change in shape. 

In the US, the cervical spine X-ray has been considered obsolete since 2010 for the purpose of diagnosis fractures. The UK has been slow to take this up - a lot of resistance has come from radiologists and radiology departments who have essentially said, they can't accommodate it, because they don't have the staff or equipment to handle the demand, and there has been no effort from central government to fix either the technology problem or the staff problem. However, people are stating to take notice, and NICE issued guidance last year stating that the cervical spine X-ray was obsolete and has (almost) no role for detection of fractures.

It's a similar thing with the elbow X-ray, the fracture is not reliably detectable, so the traditional teaching has been to assume that one is present if there is joint swelling visible on the X-ray. Of course, this is highly subjective and unreliable - and it's why, like the C-spine x-ray, it's going the way of the dodo and being replaced by CT.

As I've hinted above; there is a role for "AI" of whatever type to support the front-line, inexperienced staff. In some ways, this may be even more important in the future, because there has been a major push to "dumb down" the curriculum, accelerate the training, reduce the required experience, or simply substitute staff with staff trained to a lower standard and who would traditionally have a different job description. 

Thanks for your knowledgeable input.

I accept your point (actually I accept all of your points) wrt fracture site resorption making fractures only obvious some time after the fact. I realised at the time of writing that my reference to my, now obvious, wedging, unfairly judged medics. I accept it was wrong to leave that material in my post and am sorry I did so.

We do however appear to agree that diagnosis can be assisted by AI systems that can acquire expertise that would otherwise require years of experience, and in that regard, I see them as an important part of medicine's future.

Sure there is the cloud of dumbing down. But should medicine be spared what so many disciplines have had to suffer since the advent of the computer? Or should we level the playing field? Have structural engineers perform stress calcs instead of letting architects loose with an art degree and a CAD package (surely the direction of travel)?

Medics may ask, why not solicitors and accountants? Sure, why not. And perhaps the latter is a profession well past its sell-by date.

 

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

  • The Prime Minister stated that there were three Brexit options available to the UK:   44 members have voted

    1. 1. Which of the Prime Minister's options would you choose?


      • Leave with the negotiated deal
      • Remain
      • Leave with no deal

    Please sign in or register to vote in this poll. View topic


×

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.