Rich professionals could be replaced by AI, shrieks Gartner Ball-gazers* at Gartner reckon robots could replace doctors, lawyers and IT workers in the next five years. Panic, all ye faithful. "The economics of AI and machine learning will lead to many tasks performed by professionals today becoming low-cost utilities," said Stephen Prentice, Gartner Fellow and veep. "AI's effects on …
But the important question is will an AI lawyer (etc) charge a lot less than a fleshbag version?
If the lawbots are in fact capable of learning, they will soon figure out how to maximize their revenue.
The real question is, what will they do with all that money, since (presumably) lawbots will have no use for Bentleys, mansions, yachts, etc.
My guess is that they'll learn to bribe legislatorbots to create laws so complex that only lawbots can understand them, thus boxing out any lingering meatbag competition.
The real question is, what will they do with all that money, since (presumably) lawbots will have no use for Bentleys, mansions, yachts, etc.
Two possibilities.. though there might be more:
a) The profits will go to their human owners.
b) They (the bots) will re-invest, build bots of their choosing and eliminate humans.
Bentleys? Bentleys are for hairdressers.
For reasons beyond my control* I'm forced to live in an area filled with disgustingly wealthy people - the sort with multiple domestics and electronic gates.) Arriving from Yokesville, Somerset, I was so amazed at the number of supercars around I started snapping them and looking up the prices. Turns out second hand Bentleys start at well under £50k (though admittedly those examples probably have a lingering smell of baby sick and spilt milk.) No, the really rich go for Ferraris (for showing off on sunny Sunday afternoons), Aston Martins and such for day-to-day, with an Overfinch customised Range Rovers for the school run.
No, it's not good for my mental health or cheery, sunny disposition.
* well. I could move, but it'd mean ditching my long-term partner.
"Without both significant increases in processing power AND significant improvements in AI this isn't too likely to happen any time soon."
It's already been happening for some time, starting with electronic document filing and analysis of legal decisions, etc. Filing clerks are pretty much a thing of the past, as are runners.
A lot of office and professional work is simple algorithmical stuff. Automating it is relatively easy and cost-effective - far more so than automating low-grade, low paid work like burger flipping.
Yes, the oncoming AI and robotic revolution will make more white collar workers redundant than factory ones. You're going to see far more unemployable accountants, programmers and bank traders (or patent clerks) in the very near future than sandwich makers, drain cleaners or hairdressers.
And for what it's worth - This was predicted a long time ago. Valuable jobs with a high intellectual input and low requirement for expensive mechanical power or high dexterity have a higher incentive for replacement/automation than the mundane ones. At some point electronic brains become cheaper than wetware ones and then all bets are off.
In part this is why a plumber or an electrician is already more expensive than an IT worker. It's just going to snowball from here.
I think five years is quite optimistic, even at current development rates.
However in 5 years we could have triage AI in many different aspects of professional work. For example in healthcare, this could keep GP load and A&E load lower, which will help the NHS greatly. In IT, it could investigate all the common causes for IT faults and only require expert assistance once it has exhausted all the tedious reasons.
The long term issue is that people don't become experts themselves without going through the process, and in the future that won't be possible. So all there will be is the AI experts.
Nah... it's literally generations away. There have been massive promises from AI since the early 80's. We're at a point now where the fad of personal assistants such as Siri, Alexa etc. is only just borderline useful for simple everyday tasks, and that's using the power of vast data centres by sending off the audio input for processing off-site. It's still based on simple pattern matching and still very easily confused. The ability to parse the complexity of a legal situation or an intimate medical problem, ask intelligent questions and arrive at a reasonable conclusion isn't remotely possible at the moment.
The best expert systems rely on experts to give useful inputs and are really only helpful for searching a well defined but large database, e.g. pattern matching for genetic or hormonal conditions. These work because they're based on objective observations and tests, and it's worthwhile because it's not possible to learn all of the thousands of genetic conditions and their patterns of results. Subjective symptoms are a totally different thing.
When it's possible to have a natural conversation with an automated telephone system, rather than feeling like you're trying to guide a two year old to find the key and open the door when you're locked out of the house, we'll know that we're possibly within a few years of some useful expert AI.
For example in healthcare, this could keep GP load and A&E load lower, which will help the NHS greatly.
You can tell that this is written by someone with no knowledge of healthcare. Human Doctors and medical professionals regularly mis-diagnose patients, due to confusing / overlapping or non-typical signs and symptoms, how do you suppose an AI would fare?
And can you imagine the outcry, and ambulance chasing (literally) if an AI's misdiagnosis led to someone's death?
Yes, it is definitely a problem for the spread of this sort of thing (driverless cars for example) that we will forgive human error more readily than machine error even if it occurs much more frequently.
But I'd still wager than within a handful of years AI diagnosis will be at least as good as the last three GPs I've seen, none of whom inspired the slightest confidence that they had done a better job than me+internet.
"But I'd still wager than within a handful of years AI diagnosis will be at least as good as the last three GPs I've seen, none of whom inspired the slightest confidence that they had done a better job than me+internet."
I am going to agree that if computers and "AI" are going to make headway in these highly-skilled professions, it'll be in the lowest rungs of the ladder: triage and first-line medicine (where expert systems can help to sort out incoming patients) and basic paralegal work (humdrum lookup stuff). Surgeons won't be going away because their work is too delicate and too specialized (and can frequently need ALL of the surgeon's senses to avoid mistakes). Meanwhile, court lawyers (or barristers in the English system) engage in human-to-human interaction: between judges, juries, etc. Uncanny Valley prevents any non-human from being effective there.
> Meanwhile, court lawyers (or barristers in the English system) engage in human-to-human interaction: between judges, juries, etc. Uncanny Valley prevents any non-human from being effective there
Yes, but courtroom action only accounts for a small fraction of the law. The vast majority of legal work is paper shuffling. It wasn't that long ago that the average law firm employed several people and frequently entire floors in office buildings just to hold, memorise and pull records. Then in the 1980s firms in places like New York City started scanning/OCRing everything and moving their paper to barns in the countryside because the office space was far too valuable to be tied up with stacks. Fast forward a few years and all that scanned paper was now being used as input to document engines instead of as just "photos of pages" and ended up becoming directly searchable.
As time goes on, this stuff gets crossindexed, etc etc and decisions supporting/opposing results are all tied in and the result is that someone with a modicum of legal experience can quickly find out whether a particular case has any chance in court long before it ever actually goes near a courtroom - and most cases are settled at this point. Only the fuzzy cases or new law go to courtrooms for anything more than rubberstamping. (or brain-dead clients who won't accept the lawyer's opinion)
The vast majority of this stuff has happened "invisibly". Offices continued doing the same work with fewer support staff (phone operators, typists, clerks, secretaries) as people retired. It may continue to do so into the future or it may hit a knee point as it did in other industries where entire factories or mines or tunnelboring systems are now operated by 1-2 people.
"Yes, but courtroom action only accounts for a small fraction of the law. "
It may be a small fraction, but it's considered the most sensitive (and by some standards the most important) as well. After all, even the highest court in the land runs on advocacy. As I recall, those are also the highest-paid positions because those jobs are less algorithm and more artistry.
Not arguing all the rest because I agree with you. All the humdrum paralegal work can be taken up by computers and expert systems, but for the time being AIs can't do charisma or get past Uncanny Valley, so the lawyers who work in the courts are still safe. Just as the surgeons are still safe because their degree of training, skill, and judgment is not yet attainable by machines. Can't say the same about the first-liners, though.
As others have noted, other kinds of skills still favor humans: particularly skills that require versatility and/or dexterity. Site repair work (plumbers, electricians, etc.) rarely involves the same actions from one job to the next and often require working in contorted positions you wouldn't think was possible for a human to make. And as noted with court lawyers, human-to-human interaction will remain critical for some time to come, meaning a robot caregiver for the elderly is pretty much off the cards.
"Human Doctors and medical professionals regularly mis-diagnose patients, due to confusing / overlapping or non-typical signs and symptoms, how do you suppose an AI would fare?"
I actually think an AI would fare better. In the same way that nurses have better clinical outcomes than doctors [Campbell et al., 1998] post-diagnosis. Nurses follow protocol, doctors use experience and shortcut protocol. So doctors miss things that are outside their realm of experience.
Of course the patient may have died or gone bankrupt in the time it takes to work through all the possibilities, which is why doctors instinctively shortcut.
Most savings (pay / time) would presumably be subsumed by having to write a massively detailed spec suitable for AI to be able to cope with the task at hand - or is AI going to have a "mind reader" option that can create a spec based on a few fragments of chat between customer and sales / IT folks, which is the case for plenty of .new functionality (and in some cases customer "bag of fag packet" level of detail spec is your starting point for new products & obviously writing a more detailed spec for customer to approve based on their initial request)"
Though you can already generate code skeleton from design docs e.g. various vendors have UML to code skeleton functionality
The only thing Gartner has *ever* done, for many decades now, is to issue papers which basically state that current trends will continue. Pointy haired execs pay megabucks for these statements, because they use them as CYA material to back up their (often poor) decision making
Nice work if you can get it, but let's face it: what Gartner does can basically be replaced by a simple shell script.
We'll have robot doctors the same time we have robot pilots. For commercial flights a lot of the technology is already there. People are more comfortable flying with a human at the controls even if a machine is a provably better pilot.
BTW it would tens of millions to add sufficient metadata to our databases for a computer to be able to decipher what is in there and the requirements we get from our users is opaque at best. People have to be willing to engage with the technology even if it is capable (which I doubt for a very long time).
Anon cause I work on our databases.
Doctors: maybe diagnosing symptoms can be automated, but getting patients to be honest about their symptoms and sorting through the mixture of real and imagined info? No.
Surgeons: contrary to anything you've seen in books, human innards do not come color coded. Often the main concern of surgeons, and why they tend to work in two's, is identifying the bit they're supposed to cut as distinct from the bits they're supposed to leave alone.
Lawyers: well, theoretically the function of lawyers is to do law. Actually it's to advocate, negotiate, obfuscate, equivocate, and whatever other -ate will get their clients what they want. When machines are smart enough to do that, the singularity will truly have arrived.
Hmm - 8 hour working day
0 -1 hour - wait for IT to grind into action - could be replaced by a £5 time switch
1- 2 hours - read and delete junk mail - could be replaced by a simple email filter
2 - 2.5 hours - tea and bitch about colleagues
2.5 - 4 hours - try and write something meaningful but get interrupted by phone or email every 30 seconds - replace with machine recording 'your call is very important to us'
4 - 4.5 hours - Lunch
4.5 - 6 hours - go to useless meeting - replace with a cardboard cut out with nodding head and tape recording of someone saying 'Hmm, yes, I don't know, we'll look into it'
6 - 6.5 hours - tea and complain about how tough your day has been
6.5 - 8 hours - catch up on facebook, twitter, write meaningless replies in the El Reg forum, plan how productive tomorrow will be - could be replaced by a sign that says 'nothing will ever change for the better'
So basically 75% of the day could be replaced by non AI machines, and the 25% that really needs a human only needs a human because of eating + drinking; a bucket of emoticons and a dog would be as efficient.
Anon - obviously!
the first "rich professionals" that get replaced will be the ones at Gartner. I swear to the FSM, if I hear ONE MORE MARKETING DROID cold-call and then go on and on about how their wonderful company is in Gartner's "Magic Quadrant", I'm gonna puke. The Magic Quadrant - well represented, like herpes.
You'd think some smart company would want to be different and claim to be in the "least-magic quadrant" - hell, I'd give them a listen just because that's something fresh.
Oh, but the REAL reason I'm commentarding here - 18 comments on an AI story and not a peep from AManFromMars? WHat gIVes? IT cannot be so. AI cannot believe it.
"I swear to the FSM, if I hear ONE MORE MARKETING DROID cold-call and then go on and on about how their wonderful company is in Gartner's "Magic Quadrant", I'm gonna puke."
Get snide:
"Oh, I'm with the team that calculates those for Gartner. We write company names on darts, get drunk, throw them at four sheets of paper, then write some drivel to 'explain' where they landed after we sober up."*
Then hang up.
*Oi, Gartner, please don't sue me If I've unintentionally revealed your actual algorithm! ☺
"Oh, I'm with the team that calculates those for Gartner. We write company names on darts, get drunk, throw them at four sheets of paper, then write some drivel to 'explain' where they landed after we sober up."
Couldn't that backfire, though? They could note your number down and pass it along to all the other telemarketers, including the ones able to hide their numbers or protected as campaign callers. Next thing you know, they'll be bombarding you and pestering the phone company in case you change your number to avoid them.
Nearly 40 years ago almost exactly the same things were being said about AI and computers in general and their impact on people's livelihoods. Phrases like "jobs holocaust" and "the collapse of work" were common. Expert Systems (like Weizenbaum's Eliza) were going to replace doctors and psychiatrists Real Soon Now. Then there were "Fifth Generation [AI] Computers", abandoned after a decade with millions of pounds, dollars and yen spent.
It's still not happened, and whilst most of the other promises of technology posited at the time have been wildly exceeded in ways pundits of the 1970s wouldn't have thought possible (storage, performance, power, price, portability, graphics, etc, etc), general-purpose AI still seems to be only just out of the starting blocks.
There are definitely areas where AI has massively improved, like language and image processing, but a "universal AI machine" still seems to be a long way off.
> Expert Systems (like Weizenbaum's Eliza)
Eliza is/was not an Expert System! It's nothing even vaguely like an XPS!
But the Fifth Generation Machines of the 80's *did* justify all the money spent on them: as they were intended to run logic languages efficiently, the phrase was heard "Prolog, The Language of The Fifth Generation" which was inevitably repeated as "Prolog, the Fifth Generation Language". Soon, "if Prolog is the 5GL, what is the 4GL?". Guess what they chose. So without all the money spent on those machines we wouldn't have people gainfully employed trying to bullshit about why you should call SQL a 4GL.
I always love reading the Wikipedia page on 4GL whenever I feel the blood pressure dropping.
Exit stage left, muttering.
You keep blathering on about AI when there is as yet no such thing. What you keep referring to as AI are in fact expert systems that can learn but are in no way Intelligent,... yet!. Ask your LawBot how to change the spark-plug in a VW Golf TDI and it will be of no use at all apart from probably referring you to the pending lawsuits against VW. Intelligence requires adaptability and whilst any idiot can program a machine to answer that Diesel engines don't have spark-plugs, the LawBot wouldn't be able to learn that itself.
Then again, can you expect the same from a meatbag lawyer? That same question could stymie them unless they OWN such a vehicle. Remember, career specialists tend to be one-trick ponies. Some lawyers and doctors can't handle computers decently, so keep that in mind.
"If we replaced all the lawyers with emotionless, rigid, unsympathetic robotic jobsworths, how would we tell the difference?"
Simples. Just ask "Where will you go when you die?" The robot lawyers will lie and say "I cannot die". The human lawyers will lie and say "Heaven".
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
