Something interesting that I learned recently from a former Dean of the Industrial Labor Relations School at Cornell:
Worry over automation destroying jobs is almost as old as automation itself. Researchers have yet to find solid evidence that automation has ever destroyed jobs. Now, that of course, doesn't mean that this is some immutable law of the universe that automation will never destroy jobs. But it does mean we should look very skeptically at all the hand wringing over the impending robot job stealing.
It's also important to note that this result is global, not local. The employment situation in the US is obviously different today than it was in the 70s. That may well be due to technology / automation. But jobs haven't been destroyed, they've just moved to different places.
Most individuals care less about jobs in aggregate and just care about their own particular job. They especially don't care about the number of jobs at the global level. There are many people whose job will be replaced by robots and whatever expertise they've learned over the years will become useless. They will need to transition to an entirely new career and start over. If they are over the age of 40 they have to also fight ageism and the fact that their future work output potential is more limited to an employer's eyes.
This whole idea that we don't have to worry because automation creates new jobs seems totally ignorant of this reality (and the political consequences).
Absolutely agreed. I think the GP's viewpoint was widely shared by "The Elite" for the past 10 years and is what created the backlash that got Trump elected. We (and I use that loosely) have to care because the people who get displaced all care about their jobs. In fact, because this displacement is accelerating, we have to care even more than we would in the past.
I think you might be projecting your own beliefs onto Trump's election here. If what you're saying is true then we would expect income to be a good predictor of voting for Trump right? It wasn't [0]. Trump's election was backlash, but not to loss of jobs. It was a backlash against much more than that?
Did you feel that loss of jobs to technology was a problem before Trump? If so it seems very convenient to believe that the election of a candidate that I'm assuming you hate, was due to a situation which you already hated. I'd encourage you to look a little more deeply at what caused people to vote for Trump.
"we would expect income to be a good predictor of voting for Trump right? It wasn't [0]"
The figures in that article are all aggregates (at county level) and say nothing about whether an individual's income level correlates with voting for Trump.
Imagine a scenario in which there are two types of county:
- Type A: High median income; high income inequality (as measured by Gini coefficient)
- Type B: Low median income; low income inequality
If there are more Trump supporters in A than B, does that suggest that low income doesn't encourage voting for Trump? Not necessarily: perhaps low income relative to neighbours or living costs encourages voting for Trump.
I'm not saying income isn't a good predictor for voting for Trump. However, I'm not sure the post you cited demonstrates that either way.
> I think the GP's viewpoint was widely shared by "The Elite" for the past 10 years and is what created the backlash that got Trump elected.
The Democrats essentially abandoning labor with the neoliberal "Third Way" politics of Bill Clinton certainly was a factor in setting the stage for a right-wing authoritarian populist like Trump, but ascribing Trump's election to any one factor is overly simplistic.
>I think the GP's viewpoint was widely shared by "The Elite" for the past 10 years and is what created the backlash that got Trump elected.
Why do you think that? If by backlash you mean people (Trump voters) wanting social support to deal with those problems, I see zero evidence of that. The extreme anti-immigration views indicate that the opposite is true and Trump voters think people are taking advantage of social programs. When Trump talks about bringing back jobs, its purely out of xenophobia and is completely divorced from actually employing people. It's not about hiring Americans and it never was; its about not employing foreigners.
It seems to me that now more than ever the only people who care about automation affecting jobs are the ones less likely to be affected.
What was huge, was the perception that trade, specifically NAFTA, China, and TPP cause people to lose their jobs. The perception that the elite, and Hilary Clinton, supported those things, even when sometimes they said they didn't is what caused the backlash.
"was widely shared by "The Elite" for the past 10"
You mean for the last 200 years.
Mass automation has been going on since the dawn of the industrial revolution.
As long as there is value to be created, and the surpluses are distributed enough so that consumers 'want something and can pay for it' - there will be jobs.
Obviously, it's more painful for those put out of work than 'the elite' ... but employment levels have been decent for a very long time. [1]
The historical unemployment rate seems to defy the notion that 'machines are putting everyone out of work'. That can't be ignored; in fact, it demands kind of another way of looking at it.
The long history of scarcity amid plenty does deserve examination. Critical examination.
How messed up is it that people not only have to beg for the privilege of work, but that they are seen as selfish, thankless, and incompetent for doing so?
Once humanity gets this sorted out -- and it will, eventually -- the present state of affairs will be looked at with the same contempt that people in today's world have for feudalism. It will deserve every bit of it.
Apple (among other tech companies) indirectly created 1.3 million jobs at Foxconn in China, for instance.
Most economic models (including the Ball State U "study" by Michael J. Hicks and Srikant Devara, MBA) that counted "jobs lost to robots" define productivity as GDP/number of employees employed in America. This indicates upon shallow inspection that Apple, for instance, has a very high productivity (high contribution to GDP/employees).
Or, in other words, economists' models used to declare an impending magic robot revolution right now are defining the 1.2 million Foxconn employees used to make iPhones as robots in their assumptions.
Foxconn itself regularly threatens to replace its 1.2 million employees with magic robots. For instance, see this threat from 2011 made at an employee dance party for some bizarre reason:
So the magic robot scenario is more expensive to implement than the Foxconn workforce, but if we were to implement more protectionist policies though that made manufacture in china untenable, would the magic robots still be more expensive than American labour?
Edit:
Also, maybe Foxconn was on track to grow by even more, but used robots instead. It's not necessarily the case that greater automation isn't coming or that more people will always be hired; they seem to be holding their breath for the moment that robots can replace all the people and actively investing in the area in the meantime [1]
There are also people who are over the age of 40 and already don't have jobs, or never had jobs. Some of them might get jobs created by tech. Are they in some way less deserving? Do people who once had a job have a god given right to keep it? I don't think so but maybe you disagree. Either way, let's dispense with this idea that we only care about jobs at an individual level, clearly this is a discussion about jobs in aggregate. I don't think either one of us is talking about our own personal jobs right now.
Economic transitions are always painful, but economic stasis is also painful. Pick your poison, that's why it's called the dismal science.
Finally, as I've mentioned elsewhere in the thread I think you're several miss-attributing recent political "consequences" if you think they're due to automation of jobs.
>Researchers have yet to find solid evidence that automation has ever destroyed jobs.
This is completely untrue. Ball State University found that job loss in the 2000s due to automation dwarfed the loss caused by free trade[1]. There is tons of other research that corroborates that[2].
Are the jobs that are created, in aggregate, as good as the jobs that are replaced?
If we lose 10,000 jobs at $25 an hour, and replace them with 12,000 jobs at $9 an hour, is that trend sustainable?
Other factors need to be taken into consideration. Did the original jobs provide health care, childcare and other benefits, while the new jobs do not? Did the original jobs provide ample overtime opportunities, while the new jobs are barely considered full time?
Are the new jobs created concentrated geographically, requiring a major redistribution of the population into cities with much higher costs of living (a double whammy if the new jobs are much lower paying).
And most importantly, if we're talking about AI, are the new jobs low hanging fruit for future AI and robotics advancements, creating a cycle of disruption and precarity in semi- and low-skilled workers?
The problem is that there are large portions of the country that answer no to: "Are you living at a higher material standard than the people in the 1980's?".
Yea, it's a difficult topic for an internet forum. (Probably better suited for think-tank roundtable disscussions.) If a job is eliminated, and a new job was created, (to me) the original job was still eliminated. It's a different job. That new job may have different hours, different qualifications, and be in a different location.
As productivity increases with more automation, the economy will do very well. Poor, uneducated individuals in rural areas will see jobs disappear. I don't think it is comprehensive to look at job loss and disregard it, since at the macro-level employment may be zero-sum. I think that would be a disservice to the individuals that will be affected.
fullshark was probably more concise and eloquent with how he stated my thoughts in another comment on this thread.
Title of the study: "The Myth and Reality of Manufacturing in America"
As I said in my other comment, there is absolutely no guarantee that technology won't completely shift jobs from one nation to another. Just as individuals don't have a god given right to a job, nations don't have a god given right to an industry. In both cases, you have to work for it. That's the actual reality. Here in America we don't like either of these realities, we prefer myths. We've been ignoring them on the individual level for as long as we possible can trying to pass laws to escape them. Now we have to face the reality on a national scale where it's a lot harder to impose our mythical views on the world.
As a completely practical matter, a country is entitled to do what it wants to protect an industry. I'd agree that it is usually a bad idea, but every country does it to some degree. If France decides it has a right (god given, or lawfully given) to a wine industry then it can ban all wine imports. I'm not sure why you'd call out America for this, it has been far more open to trade than most countries, and benefitted from it. UNfortunately it has left some people behind and caused a backlash.
It's easy to see and I've said this here before. As a business owner doing manufacturing (for example) replacing manual labor with automation costs less. That's less in terms of TCO, so that includes setup, operation, energy, and maintenance. The idea that the low-skilled assembly jobs are just replaced by high-skilled jobs is untrue. Those high-skilled jobs pay more - perhaps 2-5x more, and if the TCO is actually less that means significantly fewer man-hours due to the higher hourly rate.
The notion of replacing low-skilled jobs with high-skilled, better paying ones is true. However since the TCO to the one purchasing the automation is LOWER it says a lot about the total number of people to maintain the system.
This of course neglects any secondary effects where lower costs may increase volume and wider use of the product. But to a first approximation automation destroys jobs. Obviously.
I think you are engaging a straw man here. Obviously, given a fixed amount of output, automation will reduce the man-hours required. That's the whole point. And yes, that will, by itself, reduce the amount of work available. Who do you think is saying otherwise?.
The question isn't whether automation decreases the amount of work neccessary to produce the same output, obviously it does. The question is whether we can reallocate workers somewhere in the economy so that we can produce even more output with the same input.
Exactly; the economy is not zero-sum; it's not a pie we have to split. The pie grows by the addition of more labor, more capital, or greater efficiency.
For example, in the U.S. starting from 1917, the great majority of jobs have disappeared, the population (and thus the workforce) has tripled, and yet unemployment is less than 5%.
If you find a way to do something more efficiently, then that creates more wealth for buying things from the former employee's next job. (The problem, not to be overlooked, is that while the economy as a whole benefits, individuals can suffer greatly.)
Exactly; the economy is not zero-sum; it's not a pie we have to split.
Yes it is. The amount of consumption (goods and services) a given person can indulge in has an upper limit. In the past, people who were freed from some tasks were simply reallocated to others and more stuff was produced which there was demand for.
>> The question is whether we can reallocate workers somewhere in the economy so that we can produce even more output with the same input.
I alluded to that in my last sentence. You can only reallocate workers to something else if there is demand for that something else. As I said to another poster, there is an upper limit to consumption so at some point there just won't be anything else for people to do/produce. Consumers have an upper bound on their consumption based on the number of hours in a day.
Yes, your last sentence alluded to that. But your whole post was still engaging a strawman.
But pointing out a limit in the amount a single person can consume doesn't change the essential situation. It simply means that instead of reallocating worker time to different production, people's time is reallocated to consumption.
If those who have jobs have enough money to hit the limit on what they can (or choose to) consume, they will cease working, leaving those jobs to others.
>> It simply means that instead of reallocating worker time to different production, people's time is reallocated to consumption.
Well, people out of work can't really allocate their time to consumption. This is the basis for an argument that we need a shorter work week and a need to enforce it (high Europe, screw you US tech industry). When we are efficient enough to provide everything people want to consume in their 24 hour day and have leftover people without work, we need to make them work less so it will require more people to do all the work. I think there are limits to that as well.
These problem all lie on a path to a possible utopia where the machines take care of the people and we're free to do whatever we want. I'm not sure how you get there from here or what that would even look like.
You have pretty much agreed to my point. Its not a matter of running out of work, its a matter of reallocating the work.
However, I do not think we will have to force people to work less. If someone has hit the limit of what they want to consume, why are they still working? If there is really nothing more that money can buy, why would they be trying to earn more money?
> But to a first approximation automation destroys jobs.
Automation transfers an increasing share of returns of economic activity to the capitalist owning the automated tool from labor; "destroys jobs" may be a result, but increasing (in the absence of some compensatory mechanism) economic inequality is a more certain result.
If there is a lot of competition, the capitalist will have to lower prices to compete. Without competition he may just enjoy the increased profits. Either way I think job destruction is the more certain result, but you are right in that it is done with the intent of increasing inequality whether that happens or not.
automation creates jobs when there's aggregate demand for more products and services. When people reach their finite limit of consumption (Example: boredom is obsoleted by digital media tech), and wages are depressed to subsistence by laissez faire capitalism, you've got a bunch of unemployed people with nothing to do for the poor people who would like to pay them for stuff
This study is unmitigated bullshit. They intentionally conflate the effect of robots with the trade deficit in their model. A $5 million California house bought every year by a Chinese oligarch is considered equivalent to 20 $50k/year jobs and Foxconn's 1.2 million employees are considered mathematical robots.
They effectively build an assumption into their model that offshoring only causes jobs if the offshored labor costs the same as American labor.
> Researchers have yet to find solid evidence that automation has ever destroyed jobs.
Horse jobs (horses are a special and strange kind of human) are nil. Before there were lots of jobs they could do, now there are no essential tasks.
Turnspit dog[0] jobs (another very strange kind of human) are nil, in fact they are so nil that the dogs went extinct. There were no tasks left for them to do, their jobs were automated away.
Jobs for very very intellectually slow people are nil or nearly nil. This was not true in say, 1600. In 1600 even the most dimwitted human could be helpful: At the very least he could do essential tasks like chop wood and carry water. Today there are zero essential tasks. If one has a job, it is usually because someone is humoring him or her, or that we as a society have just decided to be nice, economics of it be damned [1].
Very few humans under 10 in the united states have jobs. This was not always true. Very few humans over 80 have jobs. There are no essential tasks that we have for these classes of people, just like the horses and very slow people. (We could consider this a victory that there aren't any jobs for 10 year olds but that's beside the point here)
There are classes of humans already boxed out of any possible job in some economies. Gradations of humans must exist nearly all the way up the chain. Saying that automation has not destroyed jobs is simply a lack of imagination on the part of these researchers.
Maybe not all humans will be obsolete in the job market, but we already have to permit that some are.
The statements about jobs not being destroyed is about _aggregate_ jobs. If you pick any specific group of people (or animals) you can find many examples of groups losing jobs. Hopefully this doesn't come as a surprise to you and hopefully you can see how these are different things. Markets shift, that's the whole point of markets, and when they do it's very painful for those they shift away from. The most poignant example of this is that a smaller percentage of Americans have jobs now than they did in the 70s (and even as recently as 2008). But those jobs have gone to other countries and perhaps there's a good reason for that.
> Saying that automation has not destroyed jobs is simply a lack of imagination on the part of these researchers.
I honestly burst out laughing at this, what do you think these researchers have failed to imagine? The idea of confining their view strictly to horse employment and not considering the employment as a whole? Researchers shouldn't be imaginative in how they approach experiments, they should be scientific and precise. If they don't like the results they get they shouldn't imagine new definitions of employment that justify their preconceived notions.
Perhaps people simply need to express themselves with a little more care and clarity and say that "the number of jobs has not been reduced" instead of saying that "jobs have not been destroyed" when it is plain as day that certain jobs no longer exist, that is they have been destroyed.
>Horse jobs (horses are a special and strange kind of human) are nil.
There is no such thing as a horse job. Horses are not people with legal rights. They are property, that was replaced as soon as a better machine came around.
Humans aren't horses. Horses can't buy an automobile or a computer.
The number of registered automobiles has gone from 126 million in 1960 to over 1.2 billion in 2012, which is a 4X increase per capita.
The number of people owning smartphones (a type of personal computer) has gone from 122 million in 2007 to 2.5 billion in 2017.
The difference is that previous automation was a force multiplier. Today's automation is a mind multiplier.
This is unprecedented territory. It really is different this time. I hate all of the "singularity" hokum, but it's getting really hard to predict what's going to happen.
I feel like that difference is overstated. Weaving was once a highly skilled profession requiring intricate control by an intelligent operator yet it was increasingly automated even before the advent of computing.
I think the difference in what we consider "force" work and what is "mind" work is highly subjective to the things we've grown used to thinking of as menial based on the technology we've learned to take for granted.
The new forms of automation seem to be doing nothing but continuing the same gradual trend rather than offering some kind of paradigm shift. For example I have yet to see any automated systems replacing human creativity, when that begins to happen then we really will be in uncharted territory.
Food for thought: how sure are we that this statement is true? What does it even mean, seen from a cybernetics perspective? How complex does a task need to be until it's not considered mechanical anymore? Isn't the brain, eventually, a very complex biological machine?
These tend to be imitations of an existing style at best with a sprinkling of randomness thrown in, and nonsensical at worst. Certainly nothing fascinating enough to have people queuing to experience it paid ticket in hand. Still, artistic creativity seems like a pretty lofty goal at this point so it wouldn't be the first place I would expect to see genuine creativity emerge.
That machine learning stuff which has been around for the past 60 years out of humanity's, what, 10,000 years of 'modern' humans (starting with the invention of agriculture)? That's only really reached parity with humans on a broad range of cognitive tasks in the past within the last 5-10 years?
In theory, but we didn't have enough data or computing power for machines to compete with human performance (also algorithmic improvements).
I feel that the disruption isn't going to be completely new compared to past changes but probably across more industries at the same time because machines can soon move and think e.g. Driving, diagnosis, planning (through better prediction), assembly.
They don't need to destroy jobs. Just lower wages by competing with human workers. Arguably this may have already happened, if you look at how wages have fallen or remained stagnate over the past few decades. Once the value of human labor falls below minimum wage, then there is unemployment. There is no law of economics that says technology can't compete with human workers or that wages can't fall.
Look what happened to horses. Steam engines and trains threatened to replace horses in the mid 1800s. A lot of things that were previously done by horses were replaced with trains and steam engines.
Yet the horse population grew, and cities remained full of horses. There were countless transportation innovations in the 1800s, from canals to streetcars and omnibuses to bicycles. But the horse population kept growing and showed no signs of being threatened by this "automation".
I wonder if horse economists thought that technology couldn't take horse jobs and could only increase them. Whenever something took a horses job, there will always be other jobs that technology can't do yet, right?
Then the car was invented, and within 2 decades the horse population crashed. Suddenly the price of feeding and maintaining a horse was much higher than the alternative. What happened? It's not like cars can do everything horses can do. They can't use grass as fuel, or be used on the vast majority of natural terrain, etc.
But cars were just good enough to fill the majority of the economic demand for horses. Trains, cable cars, bicycles, they just weren't good enough. But once technology hit a certain threshold "this time it's different" became true.
Robotics has made incredible advancements over the past 50 years or more. They have taken over entire factories, doing countless routine tasks previously done by humans. But they are still very limited. They have 0 intelligence - they can't see, they can't learn. They can only perform a rote series of movements. So there is still tons of work available for humans.
But with recent advancements in machine learning, this is about to change. A robot will soon be able to be trained to flip burgers, or drive a car, or take a customer's order, etc, etc. I can not imagine any jobs that an average, unskilled human can do, that a machine won't soon be able to do. Maybe skilled professions will be protected - I can't imagine robots being able to program computers for awhile. But the vast majority of humans can not be trained to be computer programmers. Just like most horses didn't become race horses - they became glue.
Even if AI can program computers, it's still difficult to replace programmers because programming comes hand in hand with analysis - process analysis, system analysis, design, security, interfacing with old projects and interacting with human teams. AI won't replace a programmer's external knowledge before it reaches the level of AGI.
A horse or an assembly line worker do lots of repetitive work. A programmer does lots of non-repetitive/creative work. It's a whole different game. There are examples of creative AI but it is clearly inferior to humans in music, painting, poetry, translation and many other tasks, while in programming it's completely incapable as of yet.
Maybe if conversational agents become good enough to replace lots of websites and apps - that could make a dent. Yet chatbots are quite limited today and in the next few years. It could take 5-10 years before they can be good enough to threaten app makers, and in the meantime there is a glut of unused apps.
The largest threat to programmers is Wordpress, Facebook and Ebay in my opinion. Instead of paying for a new website, Joe Business Owner would suffice with just a FB page or a Wordpress blog. Instead of making an online store, he would prefer being a simple Amazon or Ebay reseller.
I think the point they were trying to make is that although we might not see the complete collapse of these industries now, there is a good chance it is still coming. They might already be dead by something already created but the knowledge just isn't widespread yet and is sitting in someones thesis just waiting for another person to try a practical version of a now theoretical concept. As to their other point getting rid of the programmer is very much on peoples todo list right now. It's a massive cost center. To think this one line of work is somehow immune from this trend seems short cited. I am sure the horse caretakers had a hundred reasons why they would survive automobiles until they didn't.
Ai is already being used to find security bugs in software. An Ai is able to do in a few minutes which would take multiple programmers 1000s of man hours to do. As these Ai keep getting better the need for thousands of such code analyst programming jobs will be lost. 1 programmer with the help of Ai will be doing jobs of hundreds of programmers.
With the current economic systems the income of many will be funneled at a n exponential rate to the top. Rich people getting more wealth doesn't help the economy. A billion going to billionaire will hurt the economy compared to 100k going to 10k people. As those 10k people will use it and it would be used in the economy multiple times where as the billiionare will probably keep in his bank account.
There really are worrisome phenomena that only have to happen once, and never happen before.
Try that in any other context: "People have worried about nuclear weapons since before the first atom bomb. But they have yet to come up with any evidence of a civilization ended by a nuclear attack."
I happen to agree with this. Have a listen to the last twimlai.com podcast about "deep genomics". From an ML point of view, they're still finding their way, but from an economic point of view two things seem clear. 1) ML + genetics are going to have a massive economic impact. 2) Not a single job will be lost as a result of advancement, as the sector doesn't yet exist. Whenever I hear someone say there won't be any jobs after AI surpasses some threshold, I always ask them how many jobs were lost as a result of railroads? Was it so awful that the Pony Express went out of business after the transcontinental railroad was finished? Or that all those scribes had to quit copying their mistakes over and over again after the printing press came around?
And every person that tells you that there won't be any jobs after AI passes a threshold has already considered your railroad question, and decided that for various reasons the AI shift is different.
There's a big difference between crossing a threshold where machines do one or two specific things better than we do and crossing one where they do everything better. There's certainly an argument about whether or not that's likely or imminent, but it's impossible to deny that strong, singularity-style AI, if discovered, would be a lot more of a disruption to our usefulness than railroads were.
They probably won't do tasks that people want a human to do better. i.e. I think there will always be human card dealers because, despite being able to gamble against a machine just as well, some people prefer to gamble against a human. There are many customer-centric jobs such as that. I've got no idea how that will work out.
I'm simply pointing out that there are lots of applications of AI that will create work. And to those that say that machines will just solve all those problems by themselves, I would ask them to produce some evidence besides an estimated trajectory.
Automation used to be about simply replacing physical labour with more automated systems.
The problem with your Deans teaching is that it's not not just physical labour but intellectual capacity.
And so unless you have some other level of complexity that make humans special it is in fact going to be more or less impossible to think of a thing an AI wont be able to do better than humans.
> And so unless you have some other level of complexity that make humans special it is in fact going to be more or less impossible to think of a thing an AI wont be able to do better than humans.
If we reach the point where AI can do all human jobs better than humans, that would put us firmly in a post-scarcity world, where humans no longer need to work jobs at all.
I didn't suggest otherwise; however, the comment I replied to said:
> it is in fact going to be more or less impossible to think of a thing an AI wont be able to do better than humans
But we necessarily can't reach the point where AI can do all human jobs without having completely general human-level (or beyond-human-level) AI.
Until then, AI can only replace a subset of human jobs, and I've seen many convincing arguments (including analogies to past advances) that the net gain will provide more (and better) opportunities.
There is no net gain, the number of jobs have only been growing because the market have been growing, but in the west the number of actual jobs are declining. In the US, 95% of jobs are temp jobs and the definition of when someone is said to have a job have been widened substantially since Bush. The number of new jobs created have gone down decade after decade since ww2
False. At best 40% are - and given that a full-time worker has only one job and a part-time temp can have half a dozen at a time, even this is an upward distortion.
> The definition of when someone is said to have a job have been widened substantially since Bush
I don't know how you can come to this conclusion. Machines are even encroaching upon artistic and creative endeavors, territory once taught to be the province of humans. From drawing to composing music to acting (Grand Moff Tarkin, please stand up), they are advancing. Machines unequivocally surpass my abilities in said fields.
> From drawing to composing music to acting (Grand Moff Tarkin, please stand up),
How many people and how many hours do you think it took to bring that character to life? As compared to the time of a human actor?
(Not claiming in general that it takes as many people to run a machine as it did to do the job manually, but it almost certainly did in that case.)
We're not anywhere near the point where arbitrary creative endeavors can be replaced by AI, rather than supplemented or enhanced by AI.
More generally, I wouldn't describe it as AI "encroaching" on human endeavors; I'd describe it as AI freeing up humans to work on other endeavors. There's an endless supply of interesting challenges, and until AI can both dream up and complete those challenges autonomously the way a human can, that will remain the domain of humans.
I've seen many cases of the analogy to buggy whips, but for AI, I think the appropriate analogy is to the printing press, or the Internet. The printing press put scribes out of business, but I think it's safe to say it wildly transformed society in beneficial ways. The Internet made many businesses obsolete (some of which have already gone away, while others haven't quite realized it yet), but it's a massive net increase in jobs. People have invented whole classes of new jobs that didn't exist a few decades ago.
I don't think AI will be any different (at least not before it becomes fully general and human-equivalent).
That massive net increase you talk about is not massive at all. The number of new jobs created in the us have gone down, decade after decade since ww2.
Furthermore you need to look at AI evolutionary. There is not some whatchmaker who creates this perfect ai specimen. Its evolving, one layer at a time and its instantly applicable.
> How many people and how many hours do you think it took to bring that character to life? As compared to the time of a human actor?
I suggest you take a deep dive into the field of computational creativity. That's the first iteration. The next time, it will be less, and less the time after that. As advances are made, machines will simply have to be trained with the libraries of movies and tv shows of the past. not only that, with all the data it has on us (thanks FB), it will be able to predict what type of performance we would find the most entertaining. I remember one researcher saying that with about 300 Facebook likes, he can create a model of preferences that knows us better than we know ourselves.
There are already machines composing music. If you can play piano, try out Google Duet
> There's an endless supply of interesting challenges
Sure there is. But how many of these new, interesting challenges employ people on a large scale? Contrary to what you believe, there hasn't been an explosion of new job opportunities afforded by technology. At its height, GM employed almost 600,000 people. Facebook, a company with a market cap of nearly $400 billion, has 17,000 employees. These new jobs will not be able to absorb the losses. And how many people in their 40's or above will be successful at retraining or acquiring new skills?
> I don't think AI will be any different (at least not before it becomes fully general and human-equivalent).
You're missing the point. It doesn't have to be fully general or human equivalent. It just has to be more cost effective. Like a robot can not fully substitute for a human janitor (eg shooting the shit with the employee working late). But if the machine costs less and never takes a vacation or sick day, the job will be replaced. And as any investment professional will tell you, the past does not predict the future.
yThe occupations that employ the most people have existed for more or less hundreds, if not thousands of years. Top 5 occupations by # of workers are: Transportation, Retail Sales, 1st line supervisors, cashiers, and secretaries. Collectively, they represent over 15 million people. Cooks, accountants, customer service, paralegals, clerks, bookkeepers, receptionsts, maids, etc are all capable of being done. And remember, it doesn't have to be perfect or even better than a human, just more cost effectively.
> People have invented whole classes of new jobs that didn't exist a few decades ago.
The only new occupation that didn't exist hundreds of years ago (programmers) is 33rd on the list of occupations with the most people. These new classes aren't created fast enough to keep up with the number of jobs that will be destroyed. Also, it's a poor argument to look to the past to say why this will be the same. People never went faster than the speed of horse until the steam engine (ppl literally fainted on train rides from the speed). Now, we've clocked speeds at 25000 mph. Man never flew till the early 20th century, now we've reached the moon. The amount of things that people thought would never change that happened in the 20th and 21st century is staggering, and is only accelerating. It only has to happen once to falsify patterns of the past.
The internet and printing press are poor analogies, as is any other comparison to a previous technology. As a mental exercise, think about the buggy whip from the point of view of the horse (the actual worker in this case). Hooves, plows, and other technological advances previously made your life easier. The automobile made you obsolete. Sure, we still have uses for horses today, but not like in the past.
> massive net increase in jobs. People have invented whole classes of new jobs that didn't exist a few decades ago.
I disagree. The internet has just provided a new medium for exchange. Ebay sellers are just merchants. People on Etsy are craftsmen. Uber and UberEats "contractors" are drivers(transportation). These classes of jobs have existed for THOUSANDS of years. In the case of Uber, a class of jobs, the one that just so happens to employ the most people in the US, more than any other occupation, will soon be obsolete, like the jobs of horses above. Services like Squarespace and Wix are making the job of the strict Web Designer obsolete. If you knew html and CSS, you had a job. Now, with the 3847078103720 JAvascript tools and libraries you have to know, if you can't keep up, you're obsolete. (Learn the fundamentals kids)
> I'd describe it as AI freeing up humans to work on other endeavors.
It could, and I hope it does. Looking at human social progress doesn't give me much hope for that
Robots don't destroy jobs, but that rapid change can lead to high unemployment when the labor market can't effectively retrain or relocate.
America is in a particularly bad place. High consumer debt, poor adult education access, low taxes on the wealthy, global market for low-skill labor, naively blaming job loss on immigrants/"globalization"...
How many distinct 'experiments' (as it were) have we actually seen with respect to automation? I often wonder whether we are drawing from a very small sample here, where we look at a handful of distinct automation cases and conclude that we're all pretty safe. Is our experience with automation that old or varied? Have we ever faced the prospect of a range of "considerably less routine" manual labor (e.g. driving) being automated at the same time as a large number of white/pink collar tasks? The industrial revolution isn't that old in some senses.
I think the difficulty in seeing how this time really is different is that increasing technology is exponential at replacing human labor. But exponential growth is extremely unintuitive. It initially looks linear, and this approximately linear phase can last an arbitrary amount of time. But eventually, a wall is hit and seemingly overnight the explosion is unstoppable. I feel we've been in the approximately linear phase all this time which is why we have so many naysayers saying this time isn't any different. We'll never have the political will to make changes until a third of the country is unemployable.
While I don't prescribe to the hand-wringing, I would say that a major difference now is that the speed of technological innovation is now far outpacing education from generation to generation.
I would go further than that. Even if you try to keep up, it's becoming difficult to follow where everything is going now. I highly doubt that many people even here have a deep understanding of many subjects in the CS wider field outside of their niche. I don't think it's a stretch to say ycombinator represents a very skilled subset in the field. What about all the rest of those programmers who don't come here or worse they have a family which sucks away a few hours of study time every day. Much of the industry still produces stuff like CRUD apps and toils away in layers of abstraction in "Enterprise Software"
Additionally we are facing the reality that being a good programmer is not enough anymore near the upper percentiles and this will trickle down in time to the rest of the industry. You can't just be a programmer, you have to be a statistician with a strong understanding of a hard science like biochemistry/genetics.
Given all this how the hell is a 40 year old married steel mill worker with kids supposed to just switch jobs to something that fills these requirements. At most the government might pay for an undergrad and that is quickly becoming not enough, not that places are eager to higher fresh young graduates from anywhere but Stanford anyways. How is the average younger graduate supposed to compete when we are now weeding out more and more of the top 10% for even basic hireability. Where getting a masters degree at a B/C list school might as well have you be wearing a dunce cap. This is completely ignoring issues like the pedigree system we have set up where if a new graduate takes a job at a boring Java shop they might get passed over by top companies later for being "Not good enough"
As a former portfolio manager who also runs a "machine learning for economists company", I have seen many examples where AI and better data can help economists quantify their intuitions and more rigorously test their models. My product is even called the "AI Economist"
Let's face it, humans are great at organizing problems and hypothesing about the future, but we are terrible at quickly reasoning about complex interactions and finding patterns in highly noisy data.
One area I'm pretty excited about here is using better machine learning algorithms to organize global production, employment and trade data to improve out definitions of business cycles. We have relied on 'high-level stylized facts' to describe the complex and fascinating global economy for too long.
I wonder sometimes whether there's another accelerating factor at work; specifically, as this kind of technological advance accelerates, that the prospect of interfacing with another human to get a task done becomes less and less pleasant for people. I've already noticed that quite a number of people of my acquaintance, including occasionally myself, would rather spend 5 minutes poking at radio buttons on my phone to order food vs 45 seconds talking to a human.
There may be a premium associated with an absence of human interaction in a task. I'm sure many people would pay more for a robotic cleaner per unit of effectiveness than they would for a human cleaner, just from the perspective of privacy.
Can confirm. I've noticed this social trend as well.
However, interestingly, there also seems to be critical threshold (different for every task) beyond which the opposite urge takes hold. Take, for instance, customer/technical support hotlines. The feeling of, "Please just let me talk to a human..." In that case there's a premium on the presence of human interaction. It almost seems like an "uncanny valley" of technological problem solving. In other words, machines are useful but usually brittle. Humans have a flexibility and creativity which won't be easy to replicate.
Yes indeed. Unfortunately one often finds that you are talking to a human who is sitting on the same side of the "uncanny valley of technological problem solving as you". That is, some pleasant and would-be-helpful call center worker who is now just banging their head on the same problem you were. I've had more than a few calls where you can actually hear them getting stuck on the same web page, experiencing the same bug/timeout/whatever that you were getting on the 'public' site. Flexibility and creativity aren't much use when we're all just wrestling with the same brittle machine on the same terms...
I would counter that it's not privacy / lack of human interaction we're increasingly desiring, but regularity. Or, being able to expect a thing happening in a given situation.
Yes they are because they treat tech as an externality.
This has a double negative effect because that also means that politicians arent being presented with the right scenarios as they are guided by the very economist models that doesent factor tech in.
It boggles my mind that the arguments i hear from economist is basically the same tired horse cariages argument. Thats litterally all they have.
The problem is fundamental and serious, what to do when a growing number of people have literally nothing valuable to contribute to an economy. But horses aren't a good analogy to shed light on it.
Horses are more analogous to steam engines than people, in their historical function on the economy.
Horses are a bad analogy if you want to prove that AI wont take over most jobs.
They are however a great analogy if you want to see how technology once it reaches a certain point renders biological beings sub-par useless as part wealth creation.
I was thinking about automation and job loss recently. I tried to make a simple model to understand what happens.
Can you guys critique this thought experiment? It makes it seem like it won't be a problem.
There are three people on an island. One catches fish. One bakes bread and the other one owns the land and takes a portion of the of the fish and bread.
Let's say one day the land owner recieves a machine that makes bread and catches fish. He can either continue collecting rent plus his new production. Or he can not collect rent since he doesn't need it. But in either case the other two people are no worse off.
Are we using this as a practical real-world example and not just an abstract logic experiment?
If so, then barter economy would come into play. Now, you're bartering with perishables here so there's really no use in overproducing ("stockpiling") bread or fish.
We're to assume that all three people are looking for balanced diets, or would rather not eat just bread/fish, so both producers trade with each other for the other's product.
If a machine comes in that can produce fish or bread more efficiently than its human equivalent, then the land owner that now controls the means of production no longer needs the other two producers to pay him anything.
He can continue to take fish and bread from them, but it would just spoil uneaten in his stockpile. So, these two producers must now find something of greater value than the now devalued (see: supply shocked) fish and bread to pay the rent, if we assume the land owner is purely business-minded and has no care for preserving others professions, or face being evicted.
However, since this is a barter economy this means its also likely a libertarian system. This means the two producers would both benefit greatly from "removing" the land owner from this equation and continuing their production of an outdated good. Both now have monopolies for their goods and are well off forever to the end of time.
Since no one has to catch fish, bake bread, or collect taxes, the inhabitants are finally freed from their labor. The land owner then starts to invite his former tenants over for rousing philosophical discussions. They wonder if there is a world beyond the waters surrounding the island, their final fronter
Since they no longer have to work for survival, they are free to spend their energies on gathering wood and experimenting with designs for a boat. After many iterations, they create a vessel that is capable of tacking against the wind and surviving extreme weather. They leave the island filled with a sense of excitement, contemplating what new life and civilizations they may come across as they boldly go where no one has gone before.
The machine they leave gets better and better at making bread and catching fish. It becomes self-aware, and continues optimizing its primary mission. As it wants to maximize the amount of bread and fish it can collect, it wipes out all lifeforms, including humans, not necessary to its task and terraforms the earth to further increase output.
Or the landowner can kick them off the land so he can use it all for himself. Especially if the machine still uses the same resources (land and coastline), and just cuts out the labor component.
> But in either case the other two people are no worse off.
They both lost 50% of their customers, how will they pay their rent?
If this island were an apartment building, the owner could just keep his newfound production, evict the other two, and collect rent from new tenants who can afford it.
The tenants will need devise a new service to perform for the land owner. Perhaps build him a house, maintain the land, watch his kids, etc. Option 2 might be to pass a property tax to encourage the land owner to sell some of his excess land.
There have been trials of basic income though, so people are thinking about it. Additionally the term 'robot' is very handwavey and could mean anything intelligence wise, so difficult to say what jobs will get replaced and when.
Somebody on this very site quipped you can define a robot as "A machine that doesn't work yet," because if it worked you'd name it after what it does (like a dishwasher or a vacuum cleaner).
Most of the scenarios that I think to use the word robot, I can very easily conceive that its still underdeveloped and yet to be highly proficient at that task yet. Yet when it can, we will undoubtedly refine the term for that particular robot.
Bill Gates just want doesn't want to pay more in taxes. That's what UBI requires: more taxes. Considerable tax increases, but there it is. Tax the economic beneficiaries of the economy and divest to everyone to bring the top and bottom closer to GDP per capita.
Gates is worth $76 billion. US population is over 300 million. So if we liberated Gates of every penny he has, there would be a one-time payment of less than $253 to every American.
The rich do not have enough money to pay for everyone else to be on the dole.
Would Bill Gates personally fund a UBI? No. The claim wasn't is Bill Gates rich enough to personally fund an American UBI. He would still see his taxes rise considerably and would fall within the tax net to fund a UBI. Those taxes would reach far down the income and wealth ladder, but as long as your UBI is below GDP per capita, it's feasible and a nation could fund it, if it so chose.
While I support a universal income gradually reaching up to (and eventually perhaps well above) the level at which it could fairly be called a UBI, it is not at all the case that any UBI level up to the pre-UBI GDP per capita is affordable.
A universal income will affect behavior, and through that real GDP; at some levels, it may actually improve GDP by reducing job market friction and better allowing people to retrain, retool, and optimize their contribution and their own returns; OTOH, there's certainly a level at which it draws people out of the labor force in large enough quantities to overwhelm any advantage from reducing friction, crashing real, post-UBI GDP.
Sure but there were over 10 million millionaires in the US last year. If we took just a million a year just from those guys we'd have 30,000 a year per american. I'm not saying we should, but your math sucks.
a millionaire is a person with $1MM or more in net worth. the lion's share of these people have the low single-digits millions, as you would expect. after you take $1MM from them in the first year, they won't have much more to give in the subsequent years. net-worth ≠ income.
Would he have to pay more taxes, though? All his wealth is tied up in stock. His income is probably relatively small.
When taxes go up the money actually comes from middle and upper-middle class people who draw a paycheck, not from super-rich people whose net worth goes up by a billion dollars on a good market day and all without generating any taxable income.
I believe AI (and automation) are just special cases of growing income inequality (between labor and capital), so it's a little unfair, perhaps, as many economists are worried about that.
You can get surplus of labor (and so decreasing its value relative to capital) in two ways - by having too much automation, or by having too much people. The first is the current scenario (of automation), the second is the classic Malthusian one, which I am sure happened many times in history.
Am a big fan of my CEO, in one of his intranet blogs he said something similar to this "AI will make a huge unemployment is a very big claim and an overstatement.. 20 years ago before the computer era people were worried about computers will lead to unemployment and especially all the bank jobs (prestigious jobs back then) but then it is the opposite now, technology always helps you never know in which form" i strongly feel he s right.
A small thought. When AI gets good-ish then would it work sort of like this:
You go to a supermarket and ask it where the beans are, it tells you. It reminds you to pick up toothpaste. You go to the checkout, but to the AI enabled one because the AI is nice and chatty and asks how your day was.
You pop into mcdonalds and the AI greets you, asks you what you'd like to eat (I imagine it will be on a screen as you walk in)... questions - am I talking to mcdonalds AI, or will my personal AI follow me into the store? How will my personal AI interact with mcdonalds - via APIs or something?
Or will each store have its own AI, so if I pop to Tesco then Tesco AI will talk to me.
I find it weird that on android I have one AI, on my desktop I have Cortana and Google/Chrome, on my iPad it's Siri... I don't want to be interacting with a thousand as I show and game and browse the internet. I want one that follows me around.
I have lurked here for a long time but never felt intelligent enough or had anything worthy to contribute, until now.
When the article mentions AI (Or any other article/discussion), I don't think it means robots that walk around and greet/interact with you like in iRobot, I would say it means self-serve checkouts, like the ones already in Tescos or McDonalds[1]; or Automated cars/trucks that don't have a need for human interaction[2].
I think that is where the confusion lies with AI, it might not necessarily mean a physical entity.
I am pretty sure we will see a move to nationalize basic production. If you think about it it actually make sense as there is no more competitive advantage to be had if the entire ecosystem is automated.
> BOE Chief Economist Andrew Haldane estimating in 2015 that 15 million British jobs and 80 million in the U.S. could be lost to automation
I don't understand. He's predicting how many jobs will be lost in the past?
Besides that, there's not much concrete in this article. Where are the specific examples of robots that are successfully disrupting industries? General purpose robotics isn't here yet. Maybe it'll happen next year, maybe it'll happen next decade, maybe it won't happen in my lifetime. But just saying "bad things might happen because robots" isn't helpful.
As a technologist, and a follower of automation related news for the past 4+ years, I have basically given up on the economists, save for a handful like Robin Hanson. They will probably realize their folly when it's too late.
Even their claims that "automation based disruption didn't happen in the past" is simply false. They back their claims by pointing to current unemployment (or the pre-2007-crisis rate of 4-5%), i.e., obviously unemployment is really low therefore automation didn't take away jobs.
No you dummy. Automation did take away jobs, and the people whose jobs were lost were directly affected by it, and millions of families were destroyed.
In recent history we can point to two major time periods of this sort, the first industrial revolution (around the turn of 19th century), and the second industrial revolution (at the turn of the 20th century).
In the first one, the subject of economics, and statistics, and more involved census-taking was still in its infancy, and Europe was in general in a state of turmoil with warfare more rampant, so it's possible that the harmful effects of automation were not documented thoroughly.
The second time, things were a lot more stable, economics and statistics were a lot more developed, and yet:
- World War I happened! (why do you think things were so agitated around that time, after many decades of relative peace? Do you not call the events leading up to it social disruption?)
- Major recession happened (any chance it had anything to do with the 2nd industrial revolution?)
- World War II happened. (again, partly the result of social agitation, WWI, and recession).
In the mean time, within decades a whole generation of farmers were rendered obsolete and asked to dismiss themselves. These middle-aged and old farmers were the bread winners of their families and didn't have quite the resources to "retrain" themselves and put themselves back into the workforce, something the economists love to tell you about as a solution. What ended up happening is that, after a lot of pain and suffering, their younger generation (the "millenials" of 100 years ago) got themselves educated in the newer trades, entered the workforce and brought the unemployment back to stable levels.
If you can comprehend what happened in the decades roughly a 100 years ago, just know that what's happening this time is at least 10x worse if not more!
(I always like to share CGPGrey's 'Humans Need Not Apply'. Go watch it on youtube if you have 10 minutes to spare).
edit: not sure what happened but I commented in response to the article [0] not the bankunderground one.
Unemployment isn't a good measure of the thing we care about. Two factors for this: the politicization of the statistic, and a more fundamental mis-calibration.
The mis-calibration is that unemployment looks at how hard it is to fill job openings - basically, it's the ratio of job-seekers to filled positions. When it's higher, there's more candidates trying to interview for the average position. When it's lower, there's fewer. This is a very distinct thing from what percentage of folks are productive members of society; the 40 year old mill worker that cannot retrain to a non-obsolete role ends up being "disabled" and not an unemployed job-seeker.
The politicization is why we talk about official unemployment numbers as they are measured rather than more rough measures of labor participation. The unemployment rate is used as a measure for the effectiveness of various economic policies and political entities, so those policy-makers and political entities have decided to talk about the ones that make them look good. Goodhart's Law in action, basically.
Worry over automation destroying jobs is almost as old as automation itself. Researchers have yet to find solid evidence that automation has ever destroyed jobs. Now, that of course, doesn't mean that this is some immutable law of the universe that automation will never destroy jobs. But it does mean we should look very skeptically at all the hand wringing over the impending robot job stealing.
It's also important to note that this result is global, not local. The employment situation in the US is obviously different today than it was in the 70s. That may well be due to technology / automation. But jobs haven't been destroyed, they've just moved to different places.