I’m very much looking forward to viewing them this evening.
In the meantime, I found an article that cites several studies from the past year or so and includes interviews with 3 people in related fields who have been studying and talking about this very subject: Experts Answer: Who Is Actually Going to Suffer From Automation?.
IMO this article is short on histrionics and long on info, but I couldn’t really find any one passage that typified.
I’d go with this part from the end:
Not me; IMO that guy really shouldn’t have been part of the article since he works for the Association for Advancing Automation, a group that could hardly be expected to offer an objective opinion on the subject.
Again, this ignores the fact that a huge chunk of the population is not intellectually suited for “insight-driven, decision-based, and creative processes”. It’s often easier to illustrate with a relatively extreme case, and I have a good one from real life.
Where I used to live in Missouri, the local Walmart employed a man with a mild intellectual disability to bring the carts back in from the parking lot. It has been a great situation for him, and presumably for the store. He appears to do a good job, doesn’t complain, and makes a little money, and does not require any special treatment. But I’m pretty sure he couldn’t do any other job in the store. I know he can’t read, and doesn’t talk much. If and when his job is automated, what job will be simple enough for him to learn it? We are going to hit a point when anything a low IQ person can do, a machine can do better and cheaper.
And then it will be the same for people of middling intellectual capacity. If we don’t genetically engineer everyone to have a high IQ, there is going to be a significant chunk of the population that is unemployable. Even if it’s not a majority, if this is true for a significant minority of the population, say 20% or 30%, that will be a huge crisis if we don’t develop mincome, or makework government jobs, or something.
There will be some who will insist the answer is education, retraining. These people resist the notion that some people are just intellectually limited (or they accept it only for the small fraction of the population that is medically diagnosed as intellectually disabled—what used to be called mental retardation). Their agenda will set us back, because it won’t work.
You think job losses are going to be determined by intelligence or education? That’s not the way it works. Whether or not a job can be automated really has nothing to do with intelligence, and everything to do with other intangibles.
For example, robots will not be framing houses or installing drywall. They won’t be tailoring suits or looking after children. On the other hand, industries which have been heavily impacted already include accountants, ‘calculators’ (people hired to do manual math calculations), draftspeople, some engineering jobs like quality inspectors, etc.
The change due to automation does not break cleanly along old class and income lines. This isn’t about the poor becoming unemployed and the rich getting all the gains. Automation could wind up devastating high income ipeople like financial advisors, middle managers, lawyers, stock brokers, and other ‘mind’ workers. Plumbers, not so much.
I did already give some examples; the stuff after “e.g.”
I don’t think that’s fair. I think a business or organization will have various short and medium-term goals, and jobs often involve an understanding of those goals, an understanding of the current situation and ability to make plans. And I mean even the typical admin job includes aspects of this; you can answer “why” questions and decide what actually needs to be done.
Now again, I’m not saying there aren’t jobs vulnerable to being automated out: clearly many are. But I’m saying the bulk of the problem is not “A computer can now do your job” it will be “Software and 2 guys can now do the job that 5 guys used to do”.
It’s still a big problem if a lot of jobs disappear this way at the same time, but it is the same kind of effect technology has had on human labor throughout our history.
Yes. Though it takes a much smaller training set to work a problem with a smaller delta. The more complex the relationship between input and out, the larger the set you need.
For example, training an AI to differentiate between “red and blue” takes a trivial training set. You’d get correct answers with just a few examples and a little amplification.
It takes an enormous training set to differentiate between “cat and dog in a picture”. But if you had a previous classifier break the data into “contains a mammal with these features highlighted (whiskers, ear locations, etc) and “background””, the training set needed is orders of magnitude smaller.
What I am getting at is by the time we want to develop AIs that “feed a chip fab and check a chip for defects”, the AI developers will not be starting from scratch. They’ll have dozens of algorithms and premade tools to start with, including probably some various hard to describe today “meta pattern finders” that will be able to “spot” irregularities in a chip like a human can.
First, huge thanks for posting the link. I’m all over the internet researching AI and machine learning for the past 3 months and I hadn’t found that yet AND it was excellent. I subscribed to AJ+'s channel before the 1st part was over, too.
Lots of good stats, lots of good info, lots of good interviews.
I think it’s very helpful for people interested in and taking part in this discussion to watch. I think it’s extremely relevant and very up-to-date. For instance, I’m fairly sure that a bunch of the footage from here in Las Vegas was shot during CES just a few weeks ago (I kept looking for me and/or a friend in all those shots too; no sightings tho).
This was like if CP Grey had done an update but he had a budget.
If anything, it makes me feel more justified in my attitude: this is going to cause problems because we as a society are simply not prepared to deal with people being unemployable in the kinds of numbers that are going to be affected.
I agree with Vivek Wadhwa that we have 5-7 years before we really see critical mass starting to build, but I’m apprehensive that by the time we notice, it will be far too late to avert. And to be perfectly honest, I don’t think that most of the people at the top of the economic food chain give a shit. I think most of them think their jobs are safe and their wealth is safe and so they’ll be fine; too bad so sad for the rest of us. I don’t think that’s an attitude that they’re going to find carries the day as well as they think once push comes to shove, tho.
To those who have not watched this yet, PLEASE WATCH ALL 4 PARTS. There is a ton of information in there.
Umm…really? The only task you mention there that I see even difficulty with is “looking after children”, and actually, there’s a perfect value-add. Building a robot that can change a diaper or grab a kid who is trying to escape wouldn’t be easy. But building robotic monitoring systems that can watch each kid 24/7, checking for and deterring child abuse, that’s totally viable. Even if the system can’t recognize all forms of abuse correctly, it would also be logging and transcribing and making searchable every minute of the kid’s life. Few people would dare harm a kid if it’s absolutely certain they will be caught, and nobody would get away with it for long.
They won’t? :dubious:
You can No True Scotsman that all you want, but the fact is that video shows a machine that installs drywall for modular homes, so that company isn’t paying people to do that job. There are only 2 people ever visible in the video.
Here’s a robot that does drywall taping, even on ceilings. There are 2 people there to set it up, but that thing is already displacing workers, even if it isn’t an exact machine analog of “a person hanging drywall”.
And ultimately, we all need to be thinking bigger: just because that’s how we used to do things, doesn’t mean it’s how we’ll always do things. Not when we can just print out a house, ffs.
I agree about plumbers, but I don’t consider that a low IQ occupation. Notice that I specifically talked about intelligence, not education. In fact, I said some people will wrongly think that more education can solve it.
And I have repeatedly said that many high paying professional occupations are vulnerable, particularly highlighting radiology. And just upthread I said:
But the specific argument I was responding to holds that workers who are pushed out of repetitive drudgery type jobs will end up working in future jobs that require creativity and decision-making. (Again, sounds like plumbing: but there is a limit to how many plumbers are needed.) And my point is that specifically regarding those types of jobs, many of the people working in them do not have the intelligence and creativity to do jobs machines cannot. So it’s not that current high IQ jobs are invulnerable to automation, but that whatever future jobs may arise will be more suitable for those intelligent people to shift to, rather than someone who has been mopping floors.
In fact, your point supports mine because there will be competitive pressure from a variety of people who have had their old jobs made obsolete, and the more intelligent job seekers will have a leg up.
As for framing houses, installing drywall, tailoring suits, and even looking after children: I do think those are much more vulnerable to automation than is plumbing (although plumbing installation can probably be automated). I’m frankly surprised you would use those as your examples.
Yes, this is it for sure. Or it may be software and 2 workers can do what 20 workers used to do, or 50.
Agreed, and as noted above this process may continue for higher IQ, white collar workers. Hardly anyone has to file any more (I assume), but there are still plenty of jobs for the kinds of people who used to do filing (who, although the task was routine and boring, had to have the smarts and attention to detail not to put something in the wrong place).
And as I also noted above, there were plenty of low-skill urban jobs for all the low-level agricultural workers displaced by agricultural mechanization. But do you think that dynamic can continue indefinitely? Politicians and policymakers usually acknowledge this–up to a point. They say American kids can’t compete in the 21st century without learning more skills than their parents or grandparents had to. But again, the dirty little secret most people either don’t believe or don’t want to face up to is that there is a large segment of the workforce (let’s say at least 20 percent, conservatively) made up of people who just don’t have the brainpower to acquire advanced skills. Unless we use genetic engineering to eliminate this group from the gene pool, how are such people going to be needed in the workforce of 2060, or maybe even 2040?
I would say the latter would be much easier than the former. But here again, it would likely be a case where there would be some humans still involved, but maybe you’d need half, or one-third, as many human workers per every 100 children as you’d need today. The “robots” would play educational games with the kids, watch for trouble signs, and sound an alarm when they think a human minder needs to come address an issue. Sometimes that would be a false alarm, and the minders would make snarky comments to each other about how stupid the machines are, but they’d be good enough to significantly reduce the payroll of a child care center (with the caveat that in this case, governments would have to agree to reduce the legal minimum adult/child ratio, and they might well not go along with that).
Really enjoyed the first and third videos, which both show stuff happening that I assume was not possible when this thread began (which, if true, really says something in and of itself about how fast things are moving). But I don’t understand what is happening in the second video. I didn’t see any difference between “before” and “after”. What am I missing?
The video is pretty poorly shot (it was posted in 2013), but that device is applying the joint compound then drying it and sanding it smooth AFAICT.
Children need human contact. Monitoring systems do not replace child care workers.
The point is that computers are not human brains, and robots are not people. So the Venn diagram between “simple for a human” and “simple for a machine or AI” does not necessarily overlap. A machine can be built that can throw three-pointers all day long with nearly 100% accuracy - a feat greater than a pro basketball player. But it’s hard for a computer to do many things that a human toddler can do.
We might find that AI’s replace a lot of high paying professions like lawyers, stock brokers, accountants, etc. Robotic drones might replace cameramen on movie sets. AI’s might become better at diagnosing human health problems than doctors, shrinking their profession. AI actors might even replace human actors. In fact, they already are. Many crowd scenes and battles are CGI now. I don’t think it will be long before an AI can generate a human face of almost any type that talks and expresses itself in ways indistinguishable from a human. Maybe the 2030 best actor Oscar will go to the programmers.
But house cleaners are probably safe. Plumbers are probably safe. Any profession that really needs extensive human contact and communication is probably safe. You’re not going to have a robotic social worker any time soon.
And it’s entirely possible than in a post-automation future there might be increased value assigned to anything humans can do that machines can’t. That will benefit artists, performers, sports stars, woodworkers, sculptors, jewelers and in general anyone who can use human skills to provide things machines can’t. That also does not necessarily overlap with current human class divisions.
I agree with most of what I quoted. It dovetails with what I said way earlier in this thread: in a post-scarcity world, when pretty much everyone can have a big house or apartment, all the latest high-tech toys, as much food as they want, etc., it will not quite be pointless to have a job and make more money than people who are on “mincome” will have–because while it may cost next to nothing to buy a wall-sized TV set, it will still be very expensive to get a table at the hippest restaurants, a ticket to the equivalent of “Hamilton”, or to have a home with a view of the Golden Gate bridge rather than a generic spot in a subdivision in “flyover country”.
Come on. Clearly I was not talking about factory assembly. I mean people coming into a framed house and drywalling it. And before you say that we’re going to start building modular houses instead - we’re not. Byzantine housing regulations make mass production of housing virtually impossible. House trailers skirt around all that by being ‘mobile’, so they can be exempt from homebuilding rules.
In any event, this is just one example of many kinds of jobs that are not ‘high intelligence’ and not the domain of the upper classes, and yet might be more resistant to the AI revolution than many white collar jobs. House cleaners are another good example. General handymen. Electricians. Nurses, mostly. Salespeople. Not retail, but the vast armies of technical and professional sales people, whose job is as much about learning the needs of a customer than making a big pitch.
That comes under the category of screw guns and hydraulic lifts - things that increase the productivity of drywallers. You still need the drywallers, but they can do more in the same amount of time. Sometimes that results in higher pay for the drywallers, or a drop in cost for drywalling, which frees up money to be spent elsewhere in the economy, creating jobs there. For that matter, by lowering the cost of housing it creates more demand for housing. The drywallers may not need more workers, but the increased demand for housing increases demand for all kinds of workers that automation can’t easily replace.
Absolutely. 3D printing could be as much a disruptive force as AI (or in combination with AI). But that’s another example of a disruption that doesn’t break along human class lines. 3D is a very democratising technology. It does for manufacturing what laser printers did for high quality printing - brought it way down in price and made it accessible to everyone. It killed the old publishing industry, typesetters became a thing of the past, but far more jobs were created because of what it enabled.
I broadly agreed with you apart from that point.
I think IQ in general is massively over-stressed in western culture, including the meme of “You must be at least X smart to do Y job”.
I am well-placed to comment on this: I joined Mensa and used to regularly attend the meetings and debates. I did a 2-year contract job where I had to work closely with neurosurgeons, making a custom software solution. And I now live in a country that de-emphasizes the role of intelligence and emphasizes hard work (China).
All my experience tells me China has it right. Within reason anyone can do just about any role. You don’t need to be particularly smart to be a neurosurgeon, and meanwhile many people who might struggle with a conventional IQ test are fantastic employees because they have other, more important traits.
I say “within reason” because yes someone with a learning disability might struggle with many jobs. And there are a small minority of roles, like theoretical physicist, where you can only do something useful in that role if you’re among the very best for mathematical reasoning and visualization skill.
(Finally though, I’m not suggesting there won’t any unemployable people. There always has been. Unemployment never gets below a certain baseline, because there are people who never got even a rudimentary education, or have mental or physical health issues we don’t know how to deal with, or various market failures like ageism)
Great videos–thanks for posting!
One item in the third video that I think deserves attention is the skyrocketing cost of housing in San Francisco. They said a family of four there who makes under $105K is officially low income. My family of four has an income of barely over half that here in Minnesota and I don’t think that describes us at all. We pay $1200 a month (heat included, which really means something in MN) for a modern (constructed two years ago) three bedroom, two bathroom apartment with stainless steel appliances, washer and dryer, balcony, huge walk-in closets. Included is a two-car garage with free electricity (we do have to walk outside to get to it), a fitness center, and a reservable indoor/outdoor community area with barbecue grills. I can’t even imagine how much the same thing would cost in SF!
Anyway, what occurred to me is that the high cost of housing in places like that will accelerate the development of robotics. It’s not a sustainable situation to have the people who work in the service industry living in RVs because they can’t afford to live in a house or apartment; but I doubt businesses are going to want to pay six figures for someone to wash dishes or deliver pizzas. So if they automate all that stuff, they can actually have a city where everyone is either affluent or rich. It’s kind of a nauseating idea, but it makes sense.
The video also touched on the conflict between those who balk at taking “handouts” and those who are like “nah, c’mon, it’s fine”. Ultimately I think people are going to have to learn to accept the latter position; but if they don’t, it could certainly fuel social unrest even if a mincome is implemented. Either way, I do foresee a weird and slightly unsettling future in which people with high paying jobs (or trust funds) inhabit coastal cities, while everyone else lives in tract homes inland, with (as I’ve described) all the basic creature comforts they need but no social capital to speak of and a distinct sense that they are an underclass. Studies have shown that people get more upset about relative wealth and disparities therein than about absolute poverty, so that could roil populist ressentiment indefinitely.
So, what you are saying here is that the reason they won’t be automated isn’t that you couldn’t automate, but because the current regulations make the environment unfavorable? I don’t know enough about the various regulations to speak to that, but as for in situ building, there are ways to do that instead of building everything at a factory. Or, a house can be built in pieces in a factory and sent out for final assembly on site (this is already done with trusses and other parts of the house, so it would just be an extension of that already existing technology). This is an older article (2016) but shows one concept for how you could build a house in situ using automated systems. Here is another, plus a video about a brick laying robot.
Regulations aside, construction is something that lends itself naturally to automation, since automation is all about performing repetitive tasks that can be modeled. Building a tract house today in the US is already optimized to a certain extent, which is why even today it uses fewer workers and automation than in the past. Assuming you could get around whatever regulations and such you are talking about I don’t see why you couldn’t fully automate the process and cut the number of construction people down to minimal numbers. Whether it’s worth going through the upfront costs to design and code such a system I have no idea.
With 8 people owning more wealth than 50% of the world population - 3.7 billion people (including 1 family, the Waltons, having 40% of America’s wealth), and wages having remained mostly flat for decades when inflation is factored in, some would argue that is exactly what happened. We haven’t reached the “out of work” phase yet, but we’ve already entered a race to the bottom with how little employers are willing to pay. Granted, automation isn’t solely to blame; outsourcing and the Great Recession played a role. But I don’t see the trend reversing until most consumers have effectively negligible buying power and the capitalist masters of the universe are forced to throw everyone a bone by implementing basic income so they’ll actually still have customers.