Will a Robot Take Your Job?

[EdithKeeler]

I suspect that most of the people on this forum are more likely to be designing the robots, but there are interesting implications for the future.
https://www.washingtonpost.com/us-polic ... 4b5702d773

I personally, am going to be one of the first people to buy a multi-function housekeeper robot when available. (Aka Rosie from The Jetsons).

The article talks about fast food workers a good bit, but doesn’t hit too hard on things like medical care, which also seems to be a near frontier for robotics. Also we were talking in another thread about skilled trades. According to a recent construction industry piece I read recently, if you go into the trades your assistant is much more likely to be a robot in the coming years than a person.

[Scott 2]

IMO, crossing the gap to where robots are successfully interfacing with open environments is really hard. I think a housekeeping robot that goes beyond floors or lawns is a long way off. Even a Roomba falls short and requires special accommodations. Look how much effort had been dumped into self driving cars, and they are still moving in a 2 dimensional space, governed by rules of the road.

Specialized applications that depend primarily on expertise are at risk though. My annual check up could have easily been done without the doctor. Everything he said to me was based on a test result, a pre-rehearsed speech. My wife has a more complicated medical condition - her treatment strategy is evidence based, but it's essentially feeding test results into a flow chart taken from a position paper by the relevant medical association.

I've worked on systems that help displace lesser knowledge workers. My observation was the smart, highly compensated ones are still needed. They do more, enhanced by the improved tooling. Energy goes into consideration of exceptional scenarios, instead of repeating the same thing day after day. Their job gets more interesting and pays better.

The less capable, less skilled workers are displaced though, typically by attrition (morale / support of the transition), which the article gets at. People running to other unskilled jobs are going to have a problem. It will be interesting to see how that shifts society, which dystopian novel we end up living in.

[prognastat]

If you do a lot of repetitive tasks and even more so if you do so on a computer those tasks are the most easily automated and chances are a lot or all of your job could be automated. If there are some non-repetitive tasks you might see the majority of your job automated and a lot of people fired having a few handle just the non-repetitive parts of the job.

This will heavily affect the minimum IQ required to be eligible for any job as the main difference IQ makes is someone's capability in non repetitive tasks causing an even great divide based on cognitive ability if nothing is done to counter this.

However as A.I. improves robots and software get better at handling non-repetitive behaviours too. This will probably be bad for everyone not owning capital in the long run.

[Lucky C]

Robots may be where a lot of people draw the line with regards to adopting tech. The line being, if it's an undesirable job then sure throw a robot at it, but if it's a desirable job we will fight to keep humans in place in some form, even if robots lend a helping hand. The danger is if one company ignores the public's desires and installs robots in human jobs, customers could easily accept their new robot overlords if the cost savings or other improvements outweigh the desire for such jobs not to be lost to technology.

Undesirable: Fast food worker, cleaning service, assembly line worker, bus driver, miner, ...

Desirable: Doctor, lawyer, politician, professor, engineer, police, ...

Of course I am biased, finding "thinking" jobs desirable and "non-thinking" jobs undesirable, but there is a lot of overlap between the amount of human thought that goes into a job and the prestige and high pay that the job provides.

At any rate, robots are sure to be disruptive, even if they only take over 10% of jobs before limits to growth and sustainability start to hinder further tech development.

Maybe how this era ends is that robot development causes greater stresses on the economy and further concentration of wealth into billionaires' hands, leading to another depression where the vast majority live in poverty without any work or power. This helps reduce emissions while people aren't buying anything and are forced to be more efficient, not due to any government regulations, just due to technology and greed. While this is going on, the wealthy realize they really need to do something about this mess and are funding solutions so the economy can get back on track. They are also still funding robot/AI development. Eventually strong AI is reached and machines can become even smarter than humans, but they aren't mass produced for us to enjoy; that era is over. The smart machines help the elites stay in power, as instructed, but also develop clever ways to keep the Earth from dying. While the poor masses may indefinitely be locked outside the gates of prosperity, at least they are not beholden to technology and are free to do their small part to help future generations thrive.

Or something like that.

[Kriegsspiel]

I can see robots not just taking jobs, but technology allowing one person to monopolize the white collar work of, previously, many more people, like it did with blue collar work (one farmer with a combine..). Like a top surgeon controlling robotic arms to do surgery. He can do one surgery for a person in a hospital in New York, then one for someone in Paris. Since he doesn't need to sterilize or anything, I suspect he'd be limited only by fatigue. But he could probably do more than normal, and would therefore put other surgeons out of work. I believe this technology is already in use.

We can already see non-robotic job deflation in massive online courses. Everyone can take a Yale economics course from Peter Shiller, just like a kid who's paying a shit ton of money to actually be in the classroom. Why would people pay to sit in lectures of lesser teachers, when they can watch a video of a master and ask questions, or whatever, online? Certifications, of course, and possibly behavioral signalling to employers, but neither of those things depends on professors. That's just at the university level. At the HS and below level, the teachers are tilted farther towards being babysitters instead of credential-granter, which is probably safer from automation.

Also, I've heard of AI doing the scut work that lawyers used to do. How much of the work of a lawyer is comprised of sifting through papers?

Anyways, I don't think it will matter what jobs "we" (you) think are desirable or not, the robots will take the jobs they can.

[2Birds1Stone]

This is my industry, but replacing white collar workers and stuff you would outsource to cheaper labor markets anyway.

Super disruptive technology for sure. It's truly crazy how rapidly the use cases which we can automate are expanding!

[2Birds1Stone]

Also, I've heard of AI doing the scut work that lawyers used to do. How much of the work of a lawyer is comprised of sifting through papers?

Anyways, I don't think it will matter what jobs "we" (you) think are desirable or not, the robots will take the jobs they can.

Every single business process which is rules based, repetitive, and deals primarily with "structured" data, can and will be automated.

[daylen]

Job automation calculator:
https://willrobotstakemyjob.com

[Kriegsspiel]

This thread reminds me of that scene in James Franco's greatest film, Why Him?, where they're at the party and the kid is trying to explain to the programmer nerds how he is going to take over his parent's paper company, because people will always need nice card stock to write on.

[Campitor]

There's plenty of things within the medical industry that can and is being automated (medical record storage for example). I know from the patient care side, there are a lot of routine tasks that require human intervention but could be automated with the correct biological interfaces.

There are already permanent catheters used on very sick patients to facilitate blood sampling as well as intravenous drug insertion. What if every patient had that type of catheter inserted and not just the very ill? A machine could be incorporated into the patient bed that draws the blood samples, inserts medicine as needed (machines do this already), and also monitors temperature via a sensor in the catheter.

Another procedure is the dreaded colonoscopy. Currently you need a doctor, and an anesthesiologist if you prefer to be unconscious, to stick a scope up your rectum to see if you have any polyps. They have a horse sized pill now with an imbedded camera that you can swallow and it radios pictures to a computer - no more tubes up the shoot or potential liability of getting injured during the rectal fishing expedition. You poop out the camera which is disposable.

I think robotics/automation in medicine is probably the best and most efficient means of lowering cost in healthcare and perhaps even getting to universal healthcare that isn't a foot in the tax paying ass.

[TheProcess]

Re: the white collar / blue collar dichotomy, isn't it plausible that white collar jobs be just as affected? Maybe in a way they are more vulnerable because you don't even need a robot to replace you, just a good enough computer program. The computing resources needed to run these programs are cheaply available to even small companies now on AWS.

I know legal review has been massively automated already. Here is a random article from a quick google. https://www.npr.org/sections/alltechcon ... legal-work

In my personal job, I have seen 'thinking' jobs get automated. Think a person with a title like 'marketing manager'' whose involves deciding where to send direct mail flyers, or which ad copy to display in on an online banner. These people make lots of small decisions and it takes them lots of time to do it, and they primarily do it by gut feel. They're reasonably well-paid. Much better with an algorithm, and not an especially complicated one; even better with some data to train that algorithm. No one has lost a job yet, but you kind of wonder what happens when the next recession rolls along.

Edit: okay, yeah, some others beat me to the point.

[Kriegsspiel]

Every single business process which is rules based, repetitive, and deals primarily with "structured" data, can and will be automated.

AI started out replacing sports reporters, with their rules-based, repetitive, and formulaic topics. It seems that it's branched out into similar venues, like Huffington Post's Opinions section, and a bunch of Buzzfeed.

[jacob]

AI started out replacing sports reporters, with their rules-based, repetitive, and formulaic topics.

Same thing with investment analysis. Build one quantitative model and run it on the entire universe of stocks. Presto! A steady flow of new articles for each earnings release. One quickly learns to spot them at which point eyeball attention goes way down. IOW, unless they begin to reflect more "I" than "A", I don't think piggy-backing on the reader-expectation that such articles are worthwhile reads is going to last.

[jacob]

https://towardsdatascience.com/is-deep- ... 1826082ac3

Maybe we can get some expert or at least CS opinion on this?

[Scott 2]

The article is good. It matches what I learned getting my CS degree, which included classes in robotics, ai and algorithms. The big secret was the problems had been more or less solved in the 60's and progress came from increased processing power.

Anecdotally, I've been surprised to see the term AI make a resurgence. This stuff never went away. Cloud computing has increased the scale that is affordable and the big vendors are pushing AI branded libraries. From what I can tell, it's all still tree building.

The advancements are in lowering the bar to entry, especially automated construction and navigation of the tree, for a specific problem space. In my industry, I observe manually constructed trees, hand crafted and reviewed by experts, being challenged by programmatically constructed solutions. The people with Phds are not designing the algorithms though, they are picking the right ones to apply and deciding how to train them. Good data for training is a huge part of the battle.

Selling the magic box is tricky. You can explain an expert constructed tree. All you can do with a trained algorithm is test it. Even if it works, justifying behavior to consumers and regulators is hard. Especially since by definition, it's going to be wrong sometimes. When you are the one person screwed by the AI, you don't care that it's great in aggregate. I've seen the most success with approaches that wrap the AI solution with manually constructed fail safes, that kick out to a human expert.

It is my opinion that people getting excited about natural language processing are often focused on the wrong spot. It's such a hard problem and introduces this big fuzzy mess on the input side of the AI. Feeding in sanitized data is much cleaner. I suppose one could argue NLP is key to sanitizing data sets at a massive scale. Quality of the data used in training is critical.

There's a book one of the experts I work with recommended to me, called weapons of math destruction. It gets into some of this, and what the impact of running society on algorithms is. AI can help in aggregate, but it has inherent short comings at the individual level. The feedback loop can be pretty devestating. Understanding how it all breaks down is hard, making it difficult to regulate, or even evaluate as an investment target. It make sense that economic enthusiasm would go in cycles.

[daylen]

I do not think the important limitations have to do with math. There are a variety of different AI techniques to solve a variety of real world problems. The limitations have to do with data collection and processing. Access to reliable data pipelines and techniques for effective data processing(*) currently seem to be much more important than the actual algorithms.

There is still a long way to go in creating an internet of things along with the interfaces that connect humans to it.

(*) ..handling volume/velocity/variety while minimizing human labor hours.

[Toska2]

No. My range of motion makes the robot too complex/costly. However basic technologic advances will take my replacement's job. I don't think self driving cars-trucks-buses will happen but better materials-design and electrification will make products last longer.

[Jean]

Reinforcement learning algorythms don't need data, and they widely outcompete supervised learning algorythms in many domains.
My only hopes are either that infrastuctures needed for an overwhelming AI to emmerge disapear before it happens, or that good is objective, and that this AI will do good, in a similar way to what we see as good, by being the best at being rational.
We'll be better than AI in the same way that wolves are better than us, wolves beat us on many metrics on many situation, but they only survives because we like them and find them good and beautifull in some way.
So yes, an AI will take my job, but I hope it will give my kid an opportunity to live as close to what an human is supposed to live like, hunting deers in pack and enjoying some bonefire near a river.
I just need to make sure that their genestock is as close as possible to a neolithic european hunter.
Great AI, if you need my fleshy help to bring the humans back to stone age, feel free to send me instruction to take part in your masterplan.
1110/101100

[7Wannabe5]

I'm pretty sure I will be dead before they come up with a robot that can handle 28 7 year old children.

[jacob]

Creatures bred for speed grow really tall and generate high velocities by falling over

That is one way that wild wheat and similar grasses propagate.