Johnny Cabs - the AI watch

[ChewingGrass]

 Being AI'd or automated out of a job tell us more

[workingpoor]

A Japanese insurance company sacked their 34 employees last week and went over to full A.I 

This is the future, Legal firms and Solicitors next? 

[long time lurking]

34 minutes ago, workingpoor said:

A Japanese insurance company sacked their 34 employees last week and went over to full A.I 

This is the future, Legal firms and Solicitors next? 

I think the data gatherers and pen pushers are going to face the same demise as the manufacturing industries already have ,if it`s worth automating it has already been done when it comes to manufacturing 

[EnglishinWales]

It was 1815. During the cold winter. The Cloth Man said he would no longer be buying my wares as he could now get cloth for half the price because of a great warehouse nearby full of people working all day at machines spinning bales of cloth that the likes of me would have taken a week to produce!

I said 'Kind sir, that cloth may be cheap but it has no traditional embroidered edges like mine', like how my mother taught me to do, and her mother taught her before that. But he said 'Nobody cares about decoration anymore. Only what's functional. The cheaper the better'. But I could not match the price. With a sigh he bid me farewell and strode off in to the town, without another word. He never spoke to me again. 

[davidg]

The problem with AI lawyers is they'll have to do better than their opponents. Who wants a cheap AI handling your case if you lose?

There are a lot of areas where just good enough works - IT development in anything non mission critical being one of them.

[Ill_handle_it]

The Thai rub and tug parlours must be getting worried.

[Frank Hovis]

14 hours ago, workingpoor said:

A Japanese insurance company sacked their 34 employees last week and went over to full A.I 

This is the future, Legal firms and Solicitors next? 

I don't see how that can possibly work. I assume it's not literally "all" but, for example, you need to structure your reinsurance programme each year, set rates, set limits on the business you will (can) accept by category, handle claims, project future claims to come up with your year end liability etc.

I would alternatively take it to mean that they have dispensed with call handlers who type in things that callers tell them and explain what is the information required.

If this was all that this company was doing then this company could itself be dispensed with entirely as a bigger company could do this equally well but for much higher volumes.

Automation is probably coming for those things that, once you've paid somebody to do them, you say "Is that all it is?".  A lot of the routine tasks of solicitors would fall unto this category IME.  I work in finance and the great majority of tasks that could be automated were rapidly automated thirty years' ago when computers came in with their ability to instantly perform big calculations.

[stormymonday_2011]

3 hours ago, Frank Hovis said:

I don't see how that can possibly work. I assume it's not literally "all" but, for example, you need to structure your reinsurance programme each year, set rates, set limits on the business you will (can) accept by category, handle claims, project future claims to come up with your year end liability etc.

I would alternatively take it to mean that they have dispensed with call handlers who type in things that callers tell them and explain what is the information required.

If this was all that this company was doing then this company could itself be dispensed with entirely as a bigger company could do this equally well but for much higher volumes.

Automation is probably coming for those things that, once you've paid somebody to do them, you say "Is that all it is?".  A lot of the routine tasks of solicitors would fall unto this category IME.  I work in finance and the great majority of tasks that could be automated were rapidly automated thirty years' ago when computers came in with their ability to instantly perform big calculations.

Absolutely.

Organisations such as the UK government, banks, etc like to put about the Digital Revolution but in fact much of what has gone on is actually not the radical application of new technology but simply the transfer of in activities performed by waged employees on in house computer  systems to web based IT applications where the same task get performed by the public for nothing. This is great for the company as they are essentially able to fire the low paid staff who do data entry into systems and replace them with people who do it for free while pocketing the difference as pure profit. The same thing applies with technology such as supermarket checkouts. When people use online banking, fill in a digital tax form, run their shopping through an APOS etc they are simply doing what some grunt got paid to to do years ago. Some people may find this way of organising such things 'convenient' but they should not fool themselves into thinking it is anything but themselves providing the free labour not the computer program. And needless to say automation does not necessarily mean that customer service improves because skinflint companies who dont back up the applications with suitably trained first line support on a help desk are essentially leaving their customers to flounder when the tech f*cks up or breaks down as it inevitably does sooner or later. In fact the problem of IT since its inception is that while it can perform certain tasks far more rapidly than its human equivalent it can also screw things up on a far bigger scale since manual processes normally limit the 'productivity of incompetence' to what a single individual can do in a day.

 

[onlyme2]

3 minutes ago, stormymonday_2011 said:

Absolutely.

Organisations such as the UK government, banks, etc like to put about the Digital Revolution but in fact much of what has gone on is actually not the radical application of new technology but simply the transfer of in activities performed by waged employees on in house computer  systems to web based IT applications where the same task get performed by the public for nothing. 

Basically putting some of the business rules and data collection into the hands of the computer rather than at a employee level, which is displacing staff. Just depends how much of the business rules and data collection can be automated. Likely the process will be front end first, then admin functions like reporting, tracking exposures, each touch point with the end customer will be looked at and one by one the function replaced by automation. On the insurance side all that would really be left is a few tech admin to look after the services, a few on the ground maybe to sort out losses and a management team to work out what insurance to write and the terms/conditions under which that insurance is written and funded and a sales arm (which in itself might be largely automated or run through ad agencies or direct with online suppliers.

 

[stormymonday_2011]

3 hours ago, onlyme2 said:

Basically putting some of the business rules and data collection into the hands of the computer rather than at a employee level, which is displacing staff. Just depends how much of the business rules and data collection can be automated. Likely the process will be front end first, then admin functions like reporting, tracking exposures, each touch point with the end customer will be looked at and one by one the function replaced by automation. On the insurance side all that would really be left is a few tech admin to look after the services, a few on the ground maybe to sort out losses and a management team to work out what insurance to write and the terms/conditions under which that insurance is written and funded and a sales arm (which in itself might be largely automated or run through ad agencies or direct with online suppliers.

 

There is nothing new about this type of expert application

I cut my teeth in IT writing these sorts of programs for mainframe systems in the early 1990s. In fact one of the applications I helped build eliminated far more posts at second level support (ie people who had to apply business rules as part of their job) than it did at the basic data entry level. The most commonly applied set of expert business rules are things like  double entry bookkeeping, purchasing and payroll management  which were already being automated back in the days when computer programs were stored on paper tape and punch cards. These IT systems generated all the things that you now get by email from online systems such as purchase orders, invoices etc. The main difference then was that the output was normally produced in paper format in large data centres and sent out through the post. However, even in the past the process of producing and enveloping documents for delivery was automated using mailing machines to handle mass print runs . So what we have seen recently is not so much a change in the 'intelligence' of the systems more  change in which the basic inputs and output processes are handled largely enabled by the enhanced 'connectivity of IT systems through modern communication networks such as the internet Now this is not to say that there is not any genuinely innovative  work being done on expert IT systems but these are far less widely distributed in the business world than people imagine and the people who write these things to gain first mover advantage are not going to give them away free to their competitors if they can avoid it. Moreover as I keep pointing out till I am blue in the face the underlying model of the modern digital computer set out by people such as Von Neumann in the 1940s still holds sway in the data centre with all its in built limitations

As an addendum it is worth noting that business rules are not fixed for ever. Legislative, business, economic  and societal changes etc mean that applications  have to be constantly updated and retested to continue to function. This is not a trivial activity as the government is finding out as it tries to replace Tax Credits with Universal Credit

 

[Frank Hovis]

Excellent posts stormymonday and only me. I entirely agree.

The last point, about rules not being fixed forever, is the one the advocates of automation miss.  The more senior my role the more I find myself dealing with "incoming"; the unusual, the exceptional, the new that has to be dealt with.

Tax manuals, for example, have roughly tripled in size since Gordon Brown took over and the era of tinkering began. My tax return changes in a small way each year.  I know clever people who now pay accountants to do it for them because, as with their cars, it has become " too complicated". And, to be fair it, has.

At work the endless changes of legislation from the Osborne era, now continuing, means that something that was a routine annual exercise is now so complicated that it takes weeks, requires recourse to expert legal advice, and whilst not quite needing to be done on an individual basis has taken the categories that need to be looked at individually from about three to over a dozen.  

I recall John Major boasting, with justification, about simplifying the tax system. That has ceased to be an aim.

[stormymonday_2011]

2 hours ago, Frank Hovis said:

Excellent posts stormymonday and only me. I entirely agree.

The last point, about rules not being fixed forever, is the one the advocates of automation miss.  The more senior my role the more I find myself dealing with "incoming"; the unusual, the exceptional, the new that has to be dealt with.

Tax manuals, for example, have roughly tripled in size since Gordon Brown took over and the era of tinkering began. My tax return changes in a small way each year.  I know clever people who now pay accountants to do it for them because, as with their cars, it has become " too complicated". And, to be fair it, has.

At work the endless changes of legislation from the Osborne era, now continuing, means that something that was a routine annual exercise is now so complicated that it takes weeks, requires recourse to expert legal advice, and whilst not quite needing to be done on an individual basis has taken the categories that need to be looked at individually from about three to over a dozen.  

I recall John Major boasting, with justification, about simplifying the tax system. That has ceased to be an aim.

The more complicated the business rules being computerised and the greater the range of data conditions the application has to test the more likely the unexpected, unusual, the exceptional, the new is going to cause it to malfunction. Sometimes the the failures are trivial but occasionally things can go spectacularly wrong and unpicking the damage can take months and sometimes cost millions. While people laugh about the automated EPOS machines issuing the tiresome 'unexpected item in the bagging area' warnings these are just a sign of how computers struggle to cope with things that do not meet the predefined sets of rules. I dont suppose this is a huge issue at the checkout but it potentally could have far more serious implications elsewhere. The human mind for all its failings is actually pretty good at dealing with certain types of changes as it is capable of making logical leaps and inventing new paradigms to cope with altered circumstances. This ability is not without its draw backs  as this process  works by a kind of natural selection that weeds out failures. There is no doubt that computers can replicate this type of fuzzy logic using enhanced modelling techniques such as those deployed in areas such as weather forecasting. The problems is that  by mimicking some of the probablistic aspects of human intelligence you also adopt many of its shortcomings. In addition there are 'unknown unknowns' out there (to quote the loathsome Donald Rumsfeld) and chaotic systems like the weather which are mathematically unpredictable beyond a certain point no matter how much technical grunt you throw at them. As long as people realised IT systems are always inherently fallible and cannot be anything else  no matter how they are produced and tested it would not be an issue. Sadly far too many people particulary in senior positions place far too much faith in them and normally dont appreciate what the consequence of them failing catastrophically would be. In fact it my time in IT I have noticed that disaster recovery planning has got lower and lower priority. Sooner or later we are going to 'dine on the consequences' of that neglect 

[The XYY Man]

46 minutes ago, stormymonday_2011 said:

The more complicated the business rules being computerised and the greater the range of data conditions the application has to test the more likely the unexpected, unusual, the exceptional, the new is going to cause it to malfunction. Sometimes the the failures are trivial but occasionally things can go spectacularly wrong and unpicking the damage can take months and sometimes cost millions. While people laugh about the automated EPOS machines issuing the tiresome 'unexpected item in the bagging area' warnings these are just a sign of how computers struggle to cope with things that do not meet the predefined sets of rules. I dont suppose this is a huge issue at the checkout but it potentally could have far more serious implications elsewhere. The human mind for all its failings is actually pretty good at dealing with certain types of changes as it is capable of making logical leaps and inventing new paradigms to cope with altered circumstances. This ability is not without its draw backs  as this process  works by a kind of natural selection that weeds out failures. There is no doubt that computers can replicate this type of fuzzy logic using enhanced modelling techniques such as those deployed in areas such as weather forecasting. The problems is that  by mimicking some of the probablistic aspects of human intelligence you also adopt many of its shortcomings. In addition there are 'unknown unknowns' out there (to quote the loathsome Donald Rumsfeld) and chaotic systems like the weather which are mathematically unpredictable beyond a certain point no matter how much technical grunt you throw at them. As long as people realised IT systems are always inherently fallible and cannot be anything else  no matter how they are produced and tested it would not be an issue. Sadly far too many people particulary in senior positions place far too much faith in them and normally dont appreciate what the consequence of them failing catastrophically would be. In fact it my time in IT I have noticed that disaster recovery planning has got lower and lower priority. Sooner or later we are going to 'dine on the consequences' of that neglect 

You talk a lot of sense stormy.

Please stop it at once - it's very distracting...!

 

XYY

                                                                                                               

The dog's kennel is not the place to keep a sausage - Danish proverb

 

[Big Orange]

We need a single point of total failure for that to happen, but seems implausible when even pocket devices are vastly, vastly superior to office space sized devices from less than a quarter of a century ago and masse of servers are like cells in a regenerating body.

But the IT infrastructure is becoming complex and bloated enough that ignorance and neglect will make it slowly decay and grow harder to fix over the time..

[onlyme2]

59 minutes ago, Big Orange said:

We need a single point of total failure for that to happen, but seems implausible when even pocket devices are vastly, vastly superior to office space sized devices from less than a quarter of a century ago and masse of servers are like cells in a regenerating body.

But the IT infrastructure is becoming complex and bloated enough that ignorance and neglect will make it slowly decay and grow harder to fix over the time..

But at the same time the hardware and communications infrastructure that connects them is getting more normalised - the server, the message passing, the underlying tech. Once you had disparate systems with varied communications and connection methods. What the internet/web has done is standardised many of the protocols and systems that are in use today. What that has done is made it very much easier to make individual functions into components that can easily talk to each other. This is what is making large amounts of embedded AI achievable. Even the hardware/systems side is being virtualised. You can set up a whole system configuration into a script and throw it at a bunch of servers and within a few minutes you have a whole custom IT infrastructure which in the past would take months to put together. This also leads to scaleability, redundancy and reliability. 

[workingpoor]

AI takeover: Japanese insurance firm replaces 34 workers with IBM Watson

http://www.thedrum.com/news/2017/01/05/ai-takeover-japanese-insurance-firm-replaces-34-workers-with-ibm-watson

Fukoku Mutual Life Insurance believes the move will improve operating efficiency and increase productivity by 30%. The implementation of the AI system is expensive, costing 200m yen ($1.7m) to install plus a yearly running cost of 15m yen ($130,000). However, the company believes this will lead to savings of over 140m yen ($1.2m) and it will see a return on its investment in less than two years.

The system is based on IBM’s Watson Explorer, a cognitive search and content analysis platform which claims “cognitive technology that can think like a human” and can analyse and interpret all data, including unstructured text, images, audio and video, according to the tech company.

The technology will replace members of the company's payment assessment-related department, which as of March 2015 had 131 employees.

It will have the ability to read medical certificates written by doctors and other documents, and factor in the length of hospital stays, medical histories and any surgical procedures before calculating payouts, according to Japanese newspaper Mainichi Shimbun. However, the sums will not be paid until they have been approved by a member of staff.

The aim is to reduce the time needed to calculate Fukoku Mutual’s payouts, which in fiscal year 2015 reportedly totalled 132,000 cases. However, it's a concerning move from a firm in a country that already has a reputation for an unhealthy work-life balance, brought to life by a tragic suicide case in Japan last year at Dentsu, with overwork being suggested as a the main reason.

[stormymonday_2011]

3 hours ago, onlyme2 said:

But at the same time the hardware and communications infrastructure that connects them is getting more normalised - the server, the message passing, the underlying tech. Once you had disparate systems with varied communications and connection methods. What the internet/web has done is standardised many of the protocols and systems that are in use today. What that has done is made it very much easier to make individual functions into components that can easily talk to each other. This is what is making large amounts of embedded AI achievable. Even the hardware/systems side is being virtualised. You can set up a whole system configuration into a script and throw it at a bunch of servers and within a few minutes you have a whole custom IT infrastructure which in the past would take months to put together. This also leads to scaleability, redundancy and reliability. 

The problem is once you move to genuine Cloud computing you don't really know how much redundancy is built into the system. Underlying all the virtual machines are hard physical devices sitting in data centres. Users are often sharing these devices with other customers as disks, memory, cpus etc get subdivided. Most resources in Cloud systems are overcommitted by design (ie the overall resources allocated to all virtual machines in a Cloud is greater than the sum of the physical resources that underly them). Standardised communications and protocols do not  guarantee that systems will not fail and indeed sometimes could make the impact of cascading failures worse. The temptation for  suppliers to cut corners because the underlying hardware is no longer visible to the end users is very great and the underlying redundancy that you are assuming exists may very well not be there when the SHTF. Moreover if your current Cloud supplier goes bust, as some have done, then while you might easily be able to rebuild your virtualised systems quickly with another Cloud supplier using pre-prepared scripts  you wont have any data to put on them unless you are in control of the backup/recovery process and you physically own the backup media

http://diginomica.com/2015/01/06/cios-worst-nightmare-cloud-provider-goes-bankrupt/

 

[newbonic]

On 1/7/2017 at 11:05 PM, Ill_handle_it said:

The Thai rub and tug parlours must be getting worried.

The Chinese looking synth in Humans would get my business every time...

[ccc]

The Thai parlours are doing just fine I can assure you. 

[onlyme2]

On 08/01/2017 at 8:00 PM, stormymonday_2011 said:

The problem is once you move to genuine Cloud computing you don't really know how much redundancy is built into the system. Underlying all the virtual machines are hard physical devices sitting in data centres. Users are often sharing these devices with other customers as disks, memory, cpus etc get subdivided. Most resources in Cloud systems are overcommitted by design (ie the overall resources allocated to all virtual machines in a Cloud is greater than the sum of the physical resources that underly them). Standardised communications and protocols do not  guarantee that systems will not fail and indeed sometimes could make the impact of cascading failures worse. The temptation for  suppliers to cut corners because the underlying hardware is no longer visible to the end users is very great and the underlying redundancy that you are assuming exists may very well not be there when the SHTF. Moreover if your current Cloud supplier goes bust, as some have done, then while you might easily be able to rebuild your virtualised systems quickly with another Cloud supplier using pre-prepared scripts  you wont have any data to put on them unless you are in control of the backup/recovery process and you physically own the backup media

http://diginomica.com/2015/01/06/cios-worst-nightmare-cloud-provider-goes-bankrupt/

 

Yes, but that is where in a lot of cases you run your own cloud infrastructure, monitor it and put sufficient services in place with overhead to grow as required. Big companies spend time and money assessing the technical and financial metrics of the data centres they use for this purpose and then do not rely on that data centre alone and distribute across the world too. Backups can be distributed too  so even in the unlikely event a data centre is completely taken out companies can come back from really catastrophic failures and they are more robust (or can afford to be more robust) than setting up all the infrastructure themselves in their own premises. 

[Rave]

"Ohhh-kay, ****ing Gretsky over here" 

[Ill_handle_it]

On 09/01/2017 at 1:54 PM, ccc said:

The Thai parlours are doing just fine I can assure you. 

Any field reports ?

[ccc]

9 hours ago, Ill_handle_it said:

Any field reports ?

Currently got a barmaid here. No money mentioned so all good. Well unless she hits me with a 3 day invoice day I leave 

Buy her drinks and bit of food appears to be the deal. I'm fine with that. 

[200p]

The Dystopian AI future

There will be mass suicides from people having little worth. Citizens Income will pay for the bare necessities, but it is never enough. People live out their meagre existence as low level life forms compared to AI. There are fortunes to be made from the underground Black market if people are prepared to become pirates or outlaws. The streets are empty either because they are too dangerous from crime, or from AI police in the central areas that are efficient in picking up crime - there is no prison as it is more efficient for AI to be judge, jury and executioner. At night, little lights can be seen from windows as people live out their fantasies online.

Governments lurch from so called "debt crisis" to the next, as they always kick the can down the road. There will be one big "Weyland Corp" that scoops up the profits from government contracts and the profits are offshored. The business owners are also the politicians, so nothing changes. The people left which are worth anything, will be the business owners that monopolise AI, asset owners and land owners.