Economists Say Newest AI Technology Destroys More Jobs Than It Creates (2014) | Hacker News Reader

Economists Say Newest AI Technology Destroys More Jobs Than It Creates (2014)

hardware.slashdot.org

58 pointsjoeyespo64 comments

> When the University of Chicago asked a panel of leading economists about automation, 76 percent agreed that it had not historically decreased employment.

> But when asked about the more recent past, they were less sanguine. About 33 percent said technology was a central reason that median wages had been stagnant over the past decade, 20 percent said it was not and 29 percent were unsure.

These two statements are distinct - the first is about employment, the second is about wages.

It seems like the author is making a leap of logic and not the words of economists to connect them.

Econ 101 says lower price means either not enough demand or too much supply. Lower wages imply lower demand for employees, i.e. fewer jobs. Or maybe too much supply of employees.

Suppression of median wages is possible if the tech is particularly complimentary to workers at the top of the earning distribution.

Econ 101 says you're assuming zero friction in the labor marketplace and exactly equal negotiating strength between employers and labor with zero collusion between employers, zero collusion between workers.

Now look at the S&P500 level over that time. Could that frankly stunning increase, GFC and all, have been less if there was a bidding war to get the best talent in labor hiring? Is all labor completely fungible?

Then we compare the salaries of senior management of S&P500 companies, who are also labor. No shortage of labor supply there at those salary levels - you, me everyone would have a go for a year at a salary of $X million. CEO salaries have not stagnated. Compare to demand for really awesome programmers which has outstripped supply by massive amounts and all the big companies desperately want to import people to meet some of that supply from overseas. But we don't see the bidding war with salaries going up to even $500k there.

Econ 101 is notice what your assumptions are in your model and think through all of them when applying that model. The assumptions of perfect competition in the labor market are obviously bogus and we need to think harder about what is going on, why and what a good outcome is (about which I'm saying nothing at all here) and how to get to that good outcome than: "The model says X so by definition there's no problem."

AI will allow a single worker to do work that previously would have required hundreds or thousands or workers, hence increasing the value of each worker and their salary.

Jobs will disappear in many places and appear in others because the world will be completely different.

> and their salary

Computers have made office workers an order of magnitude more productive than they were 60 years ago. Their salary hasn't gone up proportionately.

If we're speaking literally, yes it has. Average salary in US 1957 was just under $5,000.

If you're adjusting for inflation, it's hard to quantify our quality of life difference. We have comforts and luxuries that were hardly conceivable 60 years ago, and they're well distributed.

In real terms worldwide, GDP per capita in current USD went from $445 in 1960 to $10,000 today, a 20x increase.

So I dispute your point, except in inflation-adjusted USD in the USA for the average worker (and where I still don't think you can quantify).

>We have comforts and luxuries that were hardly conceivable 60 years ago, and they're well distributed

The avg amount of hours worked has also gone up considerably in that time, as has the level of education and human capital. If you control for both of those we indisputably get paid less per hour, while at the same time education and healthcare have gotten more expensive.

Not to mention that US poverty rate has not declined at all in the last 25 years, a period of exponential technological progression.

Housing costs in a large percentage of cities has also risen sharply as a percentage of take home pay.

If a tool allows a worker to do N times as much, that doesn't mean the labor is worth N times more, only the labor+tool system.

More like it destroys more white collar jobs & that is scary for economists that actually are part of white collar America.

(technology has been destroying blue collar jobs for a while now)

Industrialisation has been destroying blue collar jobs. Computerization has been destroying white collar jobs for a while. You don't need an army of accountants to manage your books. All you need is an SAP-like system and a small team of experts. You don't need hundreds of engineers doing industrial design all day, you just need a few good guys and a CATIA-like system.

In fact a few years ago I read the memoir of a former officer in the french secret services (DGSE). He explained at length how a large part of the job of the DGSE was to accumulate mundane information about everyone, what article who published where, who is close to who, who thinks what, what rumor X says he heard from Y on Z. Over many years of painful accumulation of data through various sources, and with a critical understanding of the reliability of each source, the DGSE built this massive file on all potential persons of interest, which is crucial for the day they need it.

Now 90% of that is available in free access on linkedin, facebook and gmail...

My favorite example is you don't need a graphics department for presentations. Just that one guy with a graphics design background.

It's scary for everyone.

The original "lump of labor" economic argument was that every level of automation lead to more abstract and higher productivity/value work, blue collar labor "just" had to get white collar jobs.

Now that automation has deeply infiltrated the entire manufacturing/retail process, it's starting to eat auxiliary "white collar" jobs.

We're running out of room for people.

Of course it does. That's kind of the point of technology.

Take farm equipment, for example. One combine replaces hundreds of people with sickles. All those people can't find jobs in the combine factory, either.

So what happened? Mass unemployment? No, those people got jobs making automobiles and refrigerators and washing machines and ten thousand other things. The only reason society could do that is because they weren't wasting all those people working on farms.

For a society, jobs cost people. The society can only have the level of material culture it can create with the people it has to spend. If in some area, we can get by with fewer people, that's progress. We can now spend those people on something else, and our society will be richer for it.

That's the high-level view. In the short term, however, if we can't figure out what to spend those people on, it can get bumpy...

At what point should society decide that robots have provided sufficient work that people should have the choice to work or not? This is why Universal Basic Income needs to become a mainstream concept ASAP.

At the point where the people can all stop working and nobody starves.

Well, if that is your criteria, and if you are working on robots you are working on the wrong problem. (Disclaimer: Robots are by day job).

Starvation is a distribution problem. We grow enough food to feed the world. Poor distribution systems cause some people to starve and some food to be wasted. Robot farmers won't fix that.

Part of the problem is politics. You can't fix that with technology or with basic income. (I mean, maybe you can fix the political problem in, say, the US, with basic income. You can't fix the political problem in Sudan with basic income, because it's a tribal and religious war.)

But, sure, there's enough food today. Then all the farmers quit. You need robotic tractors and combines. Plus you need seed, fertilizer, and pesticides, all produced (and distributed) by robots. Once the food is grown and harvested, you need every step of the distribution done by robots. You also need robots producing the gasoline, vehicles, tires, and roads that are needed for that distribution. Otherwise, if the people stop working, the food stops, and people starve.

Are we there yet? No. We're getting close enough that we can see that it might be possible to get there, though.

There's a basic, and I think unfounded, assumption we'll find new jobs for people.

Look at horses: thousands of years of utility, now essentially decorations and luxury items. Their population virtually gone in 100 years.

No one has explained to me why that won't happen to humans except "it hasn't happened yet" and "humans aren't horses (duh!)", neither of which really addresses the issue.

Capitalism is supposed to serve humanity, not the other way around. When this stops happening, if we don't have an alternative system...well, I shudder to think.

who said that? seriously, the goal of capitalism is to maximize capital however it may be defined!

once serious automation takes holds not all meat popsicles will be of worth equally.. even today humans in US are worth a lot more than say iraq. this will only get worse as haves can have entire production capacity without giving 'handouts' to the poor burger flipper just made obsolete by robot. IMO the time for UBI or such schemes is before serious automation hits so we can scale it up NOT after as it'll be too late then. It rarely happens that people give up power & influence voluntarily.

I'm not sure the population of horses has declined all that much in the last 100 years. They've gone from tools to pets, but there's still a lot of them...

Horse population in the US is half of what it was in 1920, while US population is three times what it was.

So per capita, there's 1/6th the horses of 100 years ago.

This is liked to Slashdot. At least link to the real source.[1]

A poll of economists probably isn't that helpful here. Crunching on data from unemployment offices might be more tied to reality.

[1] http://www.igmchicago.org/surveys/robots

Normally we would, but that link doesn't say much. Is there a better article to link to?

Note that all of this is from 2014, also.

Nah this is cool, man. We just keep convincing big companies and government to hire Accenture, SAP, Oracle consulting, IBM Global Services, Computer Associates, McKinsey, BCG, etc. etc.

They will happily destroy any gains in labor efficiency by hiring hundreds of people to sit in a room hacking enterprise strength (ie non-functioning) stuff to justify billion dollar fees.

It's scary how good your tech has to be to succeed in startup land and yet just how awful the tech is that people deal with at work on a day to day basis.

Anonymous has hacked big co, how to bring them to their knnees?!? "Install sharepoint and let the consultants bleed them dry."

Take this idea to its ultimate end point and their are no jobs.

Its the Super Rich that have to change their viewpoint as there ultimately has to be a new way to distribute wealth.

Everyone is entitled to a wage without work, and the Rich and Powerful must redistribute to all others is there will be anarchy.

A slow but sure increase in a basic income is the only peaceful pathway to that new future.

Of course it's the new technology that causes this and not economists running monetary experiments, that keeps wages stagnant. (who would expect the purchasing power of money to drop when increasing the money supply!?)

For the past thousands of years we've seen incredible productivity gains thanks to technology (the wheel, the loom, the steam engine!), and purchasing power always kept increasing across the population.

On the other hand, we've been running this kind of monetary intervention for a very short while, at an accelerating pace, but no way this is causing an increasing pay disparity.

This time it must be the AI.

I think globalisation (as in outsourcing to cheaper countries) had a major role. It put out of a job low skilled workers in the US, and made (part) of the cost of living much cheaper to those with a job (all the stuff imported from China).

Money printing and the over accumulation of debt I think had less of an effect. It had inflationary effects on upper middle class activities (housing in big cities, college tuitions, etc), but I believe the bulk of the effects is yet to come: a period of high inflation that can have severe effects on our way of life.

AI's doom day scenario needs to take a ticket and wait in the queue...

People confuse "the economy" with certain popular commodities, such as physical paper cash or the concept of the dollar.

In practice extreme inequality of commodities remove those commodities from the general purpose commodity market. Do you buy food with baseball cards? Do you pay the rent with beanie babies? Is your 401K in collectible Elvis plates?

Or lets look at it from top down, instead. Currently the marketplace for fine art is irrelevant to the general population. Lets say a Picasso sells for $100M in a trade between two billionaires. Frankly no one on the planet cares. Picasso's are no longer an economically relevant commodity currency. Who in America owns more than $100M of confederate money? Some coin collectors might know, but its not worth as much as you'd think.

This is where the dollar, or perhaps currency as a concept, is inevitably heading. Eventually someone in America will collect all of what we call money in 2017, one guy will have all the money, and frankly it won't matter because we'll be living day to day trading HN Karma or whuffie or bitcoins or ... The point I'm making is hyperconcentration of wealth doesn't mean the death of the economy it means the death of that form of wealth.

Is collectible Dale Earnhardt memorabilia still wealth? And if it died as wealth does it matter to you today (assuming you weren't speculating in that market?)

The key is debt. If your debt is in US dollars and you are forced to pay in US dollars, then you need to have US dollars. If you don't have them, you have to trade what you have (labor, land, beanie babies) with someone who has US dollars and want what you have. Hyperconcentrate the dollars, get most people in debt, and maintain the power to collect debts. This will make the economy not work well for most people.

Student loans, house loans, car loans, credit cards. Here we are.

I'm wondering if/when will signs of something like the Butlerian Jihad https://en.wikipedia.org/wiki/Butlerian_Jihad show.

In Frank Herbert's Dune books, the objection of humanity against "thinking machines" was not that the machines took jobs, but that intelligent machines shaped humanity too much. And signs of that can certainly be seen today. Business models and customer experiences are built around solutions that are easy to implement with computers. As AI gets used more, and AI will do the jobs of humans, human experiences will be shaped by AI. There will be a point where lawmakers will declare on certain things "no you can't do that with computers, you have to do it yourself".

My guess is that new laws will be made against specific uses of computers and AI. And that the issue of jobs will be regulated in those laws.

For example this recent law looks like something fitting in that category: http://ew.com/article/2016/12/16/obama-law-ticket-bots/

Let's illustrate this with a graph. Let's call it the Laugher Curve, and borrow the inverted U from the much loved Laffer curve. But here, the Y axis is The Value of Human Life, and and the X axis is The Value of Technology.

To the left of the curve's peak, as the value of technology rises along X, the value of humans also rises along Y. Clearly technology is augmenting human labor and increasing the value of human workers.

Further to the right the Laugher Peak is approached, and the rise in human value slows because for some jobs, augmentation has turned to automation. Machines begin to automate more tasks more than they augment. Ergo, when the number of automated tasks rises sufficiently, machine replaces man.

Then as displaced humans seek new jobs, they find themselves taking positions of lesser value because they lack the skills or ability needed to qualify for those remaining jobs of equal value due to those jobs' increasing complexity, which was caused by their rising tech augmentation. What many humans do best, now a machine can do cheaper, and the retraining of those humans for equally valuable jobs has become unprofitable.

Further to the right, human augmentation by tech fades and automation is in full swing -- replacing those humans with marginally greater and lesser skills. Finally when we reach the right edge of the curve, all humans have been made redundant by tech; they're unemployed. At last, the net present value of mankind falls to zero.

[It's called the Laugher Curve because it's believed the original Laffer Curve was never about taxes. It was just a pretext to encourage those folks who valued tech more than people. It also cleverly refers to the oxymoronic outcome of a temporal mobius loop, in which the final laugh was on those who could laugh, by those who couldn't.]

I don't believe in an efficient market response when it comes to improved accuracy of models and ultimately AI. More jobs will be destroyed before they are replaced. The volatility that results will decimate wages one industry at a time (assuming applications are gradual). I would be curious to see % of population employed over the periods where productivity spiked.

Generally, the way it works is like this, a new technology comes around and every one in that generation that had that career is out of a job in that industry.

So a whole generation loses a job. Most of them don't get new jobs.

Their kids learn new vocations.

https://en.wikipedia.org/wiki/Power-loom_riots

Happened in the textile industry. In New England, Portuguese immigrants lost their jobs building cloths to more efficient machines and outsourcing, their kids became lawyers and doctors.

This will most likely happen in today's computerized automation wave.

One generation gets screwed but their children adapt to new opportunities.

Just because it worked out in the past doesn't mean it will continue to work out in the future. There is no fundamental reason that pattern is guaranteed to continue and some evidence exists already that this time really is different.

If industry displacement starts occurring more than once a generation?

Our rate of advancement is increasing. We will get to the point where new job opportunities will be automated before there's even a real workforce.

New jobs will only be created if people can consume (and afford to consume) more. We're getting pretty close to consumption limits.

I don't agree. Whats better than a fictitious human-like AI of tomorrow? A literal human slave 200 years ago. The US South had millions. What did the free people do? Work of course.

You can point to the "Poor White" as an example of what AI could create tomorrow but the "Poor White" existed well before and after slavery was instituted.

The solution to poverty is always the same. More education and more infrastructure. AI should free up enough resources to help make those two goals much more attainable.

When will people realize that increased productivity is what makes society better to live in and not jobs. If all you needed were jobs then every country in the world would have no excuse for low living standards - they'd just need to get people a job doing anything, picking crops, whatever. [1]

1. https://en.wikipedia.org/wiki/Productivity#National_producti...

Increased productivity isn't making society better to live in for people who don't have jobs.

Yes it does. Increased productivity means more tax revenue for government programs, cheaper consumer products and improvement in living standards globally across the population. Its like giving your entire economy an +offset which shifts your distribution in a positive direction.

This is the point. Make computers do work, then sit back and drink a mojito ( or your drink of choice ). Get a new hobby if you want to work.

How do you pay for the mojito? Society has been conditioned to the idea that "if you work, you get stuff". In fact the current system is rigged to ensure that, so if jobs go away instead of mojito, you get nothing, and instead of hobbies you get to beg.

I'd love nothing more than a post-scarcity world where everyone get enough to live on, and people who do more get more. However there are X things that will stop that:

1. We have the resource to do this for everyone, but it means some people will have to reduce their quality of life

2. People like to work, they need to have meaning - some people will find hobbies, but some people just aren't creative thinkers and need some direction from society or otherwise. Here is education and basic wage, now go away won't suit everyone

3. It means the people at the top end need to sacrifice some. That isn't going to happen.

Less than half the population works and pays for everything everyone else uses. The majority are not in the "people like to work" camp. Its not going to be a major problem. Over half the population either strives (no change necessary) or simply exists (no change necessary)

There's a lot of political sloganeering about redistribution and basic incomes that make it politically impossible. What seems likely to me is utility-ization of more services and abstraction of who pays (if anyone). That's a free public utility sounds a lot more likely politically than "forced redistribution of wealth at gunpoint" or "basic income" or "universal welfare" or whatever. Even if it ends up being the same thing with a different name.

You don't need to wave cash in the air to obtain a policeman's service and this seems quite normal. Obviously "somehow" some kind of shuffle game is producing salary for that policeman, but its kinda abstract and out of sight. Likewise our supermarkets are likely to go extremely downscale in quality, but will have no cash registers.

People will continue to trade labor for money and trade money for services and better class of stuff. I'm just thinking in 2075 it'll seem conceptually weird to pay an electric bill, just like it would seem conceptually weird today to obtain a policemans services by waving a $20 in the air. Paying an exit fee at a register at a public food store will seem as weird as paying an exit fee to a public library. Why would I pay for medical care, I don't pay for water or oxygen or wifi? No one expects a bill for storm drain service today, its very abstract. Likewise no one will get a bill for natural gas service, although in a very abstract sense someone will get paid for it.

Now the public food store will have little better than human kibble able to keep alive but if you want the fancy stuff, get a service job and trade for really good food. I suspect the black market will flourish.

How the new economies will deal with legacy economies like real estate or FIRE sector in general or the legacy IP economy will be interesting to watch.

Currently that police officer is motivated by money. That money comes from taxes collected by siphoning off a bit of value every time value changes hands. How do you continue motivating that police officer? Or the engineers/technicians operating the free-electricity boxes?

Exactly my point. How do you pay for social workers in a poor city, how do you pay for the welfare state now? Surely you can't expect a social worker to carry around a cash register, their clients don't have money after all.

I'm not sure how the back end will roll out but I know for sure the front end will roll out as a cultural shift in how we view most of the economic transactions we perform today.

We can't slip "universal welfare" or "socialized medicine" past politics today, the concept of basic income is unacceptable to too many. Because we need it anyway it'll be rolled out as the death of the shopkeeper transaction economy.

I'm sure someone pays the snowplow driver, somehow. I don't wave $20 bills in his face before he plows the street in front of my house, nor does he threaten to kneecap me if I don't pay up after he plows. He doesn't send me a bill either. And god forbid in red blooded America we have "socialized plowing" or "basic plowing" or "universal road care". None the less somewhere in the back there is a massive shell game going on with prop tax revenues and budget and somehow that plow gets filled with diesel and 99.99% of the population pays it no attention while simultaneously obtaining completely communist street snow plowing services based solely on need. Just make sure to never point out in polite company that we have "socialized plowing" or a "universal basic snowplow".

A lot of people talk about end user UI which is not going to happen in the USA. 100% change we will have socialized medicine and the only thing 100% certain about the scheme is we will not call it socialized medicine. It'll be like "medicine as a service" or "public utility medical services" or "One of many benefits of living in this fine city much like the free public parks"

Who will make sure you get your fair share of the automation profits?

As a controversial idea intended to provoke thought and not knee jerk political reactions, consider that in human history, much as we are in a bubble of fossil fuel burning, we are in a bubble of equality. If in the infinite oscillation of all human activity, is there any reason to exclude the past being the future? Is a future stone age impossible because we have run out of stone?

A common theme in ancient history was omnipotent leader getting pissed off at the sea (for sinking a fleet, or just because he's a jerk, it don't matter why) so he walks out in the sea and whips it or orders his army to slash at the sea. No matter how much you might dislike the idea of the end of the equality bubble, disliking it really really hard is about as effective at preventing the future as whipping the sea is at teaching hurricanes not to vaporize naval fleets.

Hey those blue collar jobs are going away, best get used to it, learn something new. Hey that human equality thing is going away, best get used to it, learn something new.

Even analogies don't work. You can say we learned the technologies of freedom and equality and they only increase over time because... well because of luck so far. Just like we learned the technology of burning coal or slavery, we learned it and it can only increase over time because ... well it did out of luck until it stopped. One's gone, the other's starting to go away, maybe everything in this paragraph is over, eventually.

I don't like it, isn't a valid rebuttal. Personally I don't like it, but I'm willing to devils advocate it a bit and its one of those things where if your mind is open and not dogmatically shut you can't discount the highly likely (but not 100%) probability.

You can tell mother nature "we must have a solution meeting my criteria A B and C" and if those are unsolvable, she is quite unconstrained in actual practice.

You have the King Canute story all wrong.[https://en.wikipedia.org/wiki/King_Canute_and_the_waves]

Its a common cross cultural snark about obnoxious authoritarian leaders, personally I was thinking of good old Xerxes but it just keeps appearing everywhere ...

Look at how the proceeds of the big tech exits get distributed and you see that even the people building the automation itself barely benefit, save for VCs and management

How do you bridge the gap between now and what you've described without recreating the Russian Revolution?

This trend has only accelerated, will only continue to do so, and will inexorably force a rearrangement of the productive forces in society.

(2014)

Yes. Added.

Economists always fail in their predictions, but are very good to explain things in retrospective.

AI would not directly create new jobs, but rather allow people to take up other jobs.

> allow people to take up other jobs

What jobs are these? Why aren't they "taking up" these other jobs without the influence of AI and automation?

Maybe these new jobs that they are "allowed to take up" are worse than their jobs that are being done by machine intelligence.

"Take other jobs" where the Return on Investment (ROI) of AI or Automation can not exceed employing minimum wage earners. The higher the income for that worker the more likely that position can be justified in being replaced by automation.

Isn't that the point?

Summing up the perspectives on AI and Automation in 13 years in just 2:39:18 of qualitatively worthwile (my opinion) mealpieces of TED-Talks in the playlist below. If you find an equaly rich source of views and shapes on the same topic please comment below, I'd love to read and learn more about it.

Getting back to business: I am to blame, because I'll automate away at least 40 Jobs until June 2017 in just one site (at the begining) of a large multinational company using Deep-Leearning. Kind of feel bad, but I'm not the evil and I certainly don't destroy jobs. Whoever is using the word destroy in topics related to AI is strongly trying to polarize. I really enjoy critical thinking, arguments that make you think, but most often these kinds of discussions sadly end in fear-mongering or finger-pointing akin low quality results. That's with the people who have the strongest opinion about these topics, those are also usually the only ones who bring up these topics, so that's a dilema.

How do you argue against and with people who are mad about you, beceause they think you destroy jobs?

I know it's not universally true, but try to replace the word destroy with augment: Economists Say Newest AI Technology Augments More Jobs Than It Creates (2014)

And the whole thing becomes a new meaning. Now take a look back at how the common consensus about AI and Automation and the accompanied fears have changed and formed over the years using following TED playlist, I've compiled for your pleasure.

• David Autor: Will automation take away all our jobs? [1]

TEDxCambridge · 18:37 · Filmed Sep 2016

• Olivier Scalabre: The next manufacturing revolution is here [2]

TED@BCG Paris · 12:26 · Filmed May 2016

• Anthony Goldbloom: The jobs we'll lose to machines — and the ones we won't [3]

TED2016 · 4:36 · Filmed Feb 2016

• Jeremy Howard: The wonderful and terrifying implications of computers that can learn [4]

TEDxBrussels · 19:45 · Filmed Dec 2014

• Andrew McAfee: What will future jobs look like? [5]

TED2013 · 14:15 · Filmed Feb 2013

• Erik Brynjolfsson: The key to growth? Race with the machines [6]

TED2013 · 11:56 · Filmed Feb 2013