Arnold Kling  

The Great AI Shift?

PRINT
Live and Let Live... My Speech at Western Kentucky ...

Jaron Lanier says,


If you had talked to anyone involved in it twenty years ago, everyone would have said that the ability for people to inexpensively have access to a tremendous global computation and networking facility ought to create wealth. This ought to create wellbeing; this ought to create this incredible expansion in just people living decently, and in personal liberty. And indeed, some of that's happened. Yet if you look at the big picture, it obviously isn't happening enough, if it's happening at all.
...There are just a lot of people who feel that being able to get their video or their tweet seen by somebody once in a while gets them enough ego gratification that it's okay with them to still be living with their parents in their 30s, and that's such a strange tradeoff. And if you project that forward, obviously it does become a problem.
...What's happened now is that we've created this new regimen where the bigger your computer servers are, the more smart mathematicians you have working for you, and the more connected you are, the more powerful and rich you are. (Unless you own an oil field, which is the old way.) I benefit from it because I'm close to the big servers, but basically wealth is measured by how close you are to one of the big servers, and the servers have started to act like private spying agencies, essentially.
...Basically, people can expect free stuff from the Internet but they don't expect wealth from the Internet, which to me makes it a failed technology at this point, although I hope it's revivable.
... whenever you improve efficiency, when you save money, it's only the same thing as making money if you're already rich. If there are people who aren't rich enough to benefit from that, it just makes them poorer because they have less to do, and less ways to earn money.

Pointer from Tyler Cowen. Robin Hanson has warned that as artificial intelligence improves, there will be a dramatic movement toward unequal income distribution, with the owners of capital having much and those who only offer labor having little. One way to interpret what Lanier is saying is that we are moving in this direction.

I recommend reading Lanier's entire spiel, even though I do not think all of it makes sense. Let me try to re-tell his story in my own words, without saying that I agree with it.

One way to think of it is as follows. People will be paid for doing what on the margin cannot be replicated. In an industrial economy, it costs something to replicate what I do with my hands and brain in a factory. In an artificial-intelligence economy, it is costless to replicate stuff.

In an industrial economy, if I do not work on building a car, then at the margin someone loses out on a car. In an Internet economy, if I do not write a blog post, no one loses out on anything. Not because I am such a lousy blogger, but because if I fail to write the blog post, then someone else will, and that blog post can be copied an infinite number of times. The blog post does not really need me to produce it in the way that the car does in an industrial economy. However, in an artificial-intelligence economy, the car becomes as easy as a blog post to produce and re-produce.

So, lots of people can become ZMP workers (zero marginal product). Meanwhile, a relatively small set of people will be able to figure out the artificial-intelligence economy and be able to position themselves to reap its rewards.

Happy Labor Day.

[update. Timothy Taylor finds this quote from Keynes:


We are being afflicted with a new disease of which some readers may not yet have heard the name, but of which they will hear a great deal in the years to come--namely, technological unemployment. This means unemployment due to our discovery of means of economising the use of labour outrunning the pace at which we can find new uses for labour.

A reader points out that this is line with PSST. In fact, I could have used this quote in an article I wrote on PSST that is forthcoming in a professional journal.]



COMMENTS (15 to date)
J Storrs Hall writes:

You might try reading chapter 21 of Beyond AI.

"... The key to beneficent AIs is simple: ensure their interactions with each other and with us are economic, based on universal rules of property and reciprocity, rather than political, in which takings are the norm.
..."

jeff writes:

I recommend reading Lanier's entire spiel, even though I do not think all of it makes sense. Let me try to re-tell his story in my own words, without saying that I agree with it.

We could use such a translation of every spiel Lanier has ever written.


Thomas DeMeo writes:

There is always a political tipping point where acceptance of the property rights of the few is disolved. The majority will not accept limitless inequality.

jb writes:

AI is going to be relentlessly applied to automation of larger and more complex tasks.

The more closely you can align yourself with that application, the longer you'll be gainfully employed.

I expect that social strife will be associated with this process, and I am quite concerned about how that will play out. When only 49% of the people in a country can find a decent job, can our current meretricious social order survive at all?

Vladimir writes:

J Storrs Hall,

That's unfortunately far from a sufficient condition for the AI to be beneficient to most humans. If you live from selling your labor, as most people do, and if the price of your labor falls below subsistence, you're dead -- unless of course someone keeps you alive out of charity, which is however not at all implied by property rights. And AI implies pretty clearly that the cost of human labor will fall to an infinitesimal fraction of what it costs to keep a human alive (let alone in tolerable conditions).

Or to put it differently, pecuniary externalities can be just as deadly as predation and regular externalities.

Vladimir writes:

Thomas DeMeo:

There is always a political tipping point where acceptance of the property rights of the few is disolved. The majority will not accept limitless inequality.

That seems empirically false to me. Historically, the most extreme inequality can be observed in -times of disastrous famines, when a significant percentage of people starve to death, and the majority struggles for the barest survival, while the elite still lives decently, possibly even in luxury. What we observe in practice is that as long as the elites rule with a minimum of competence and without serious intra-elite strife, the people will just starve quietly and obediently, without any threat to the established order. (With perhaps a few isolated small incidents, which are easy to deal with.) For a very recent example, look at North Korea.

Hoosier writes:

And this is a good thing?!

Dornier Pfeil writes:
Vladimir writes:
Thomas DeMeo: There is always a political tipping point where acceptance of the property rights of the few is disolved. The majority will not accept limitless inequality.
That seems empirically false to me. Historically, the most extreme inequality can be observed in -times of disastrous famines, when a significant percentage of people starve to death, and the majority struggles for the barest survival, while the elite still lives decently, possibly even in luxury. What we observe in practice is that as long as the elites rule with a minimum of competence and without serious intra-elite strife, the people will just starve quietly and obediently, without any threat to the established order. (With perhaps a few isolated small incidents, which are easy to deal with.) For a very recent example, look at North Korea.

The French Revolution doesn't constitute empirical evidence?

sean writes:

I wonder if we start to move towards an extensive zmp economy, if the rationale for income redistribution starts to make a lot more sense. any thoughts?

Vladimir writes:

Dornier,

No, it doesn't. The French Revolution would have been easily suppressed before it escalated beyond a few pitiful incidents if the king had decided to do so instead of naive appeasement. And the initial phases of the revolution, and even much of its subsequent escalation, were principally the work of radical reformist elements within the ruling elite. (For example, the main patron of the Jacobins and their key protector before they seized power was Louis Philippe II, Duke of Orléans and king's cousin.) The lack of effective response to the mob excesses was due to these rifts in the ruling elite and the king's naivete and indecisivenes, not the power of the mob itself.

Generally speaking, the notion of real grassroots popular uprisings against inequality and oppression is largely a modern myth.


Gian writes:

Where is AI in the Internet economy right now?

There may be technical unemployment due to automation but why conflate it with nonexistent AI?.

Evan writes:

Lanier's thesis seems internally consistent, but I can't see how it lines up with historical facts. Farming automation made it nearly costless to replicate the labor of farm workers, but they found work elsewhere. Industrial machines did not cause unemployment either, even though they replicated much of the labor of industrial workers.

You might say that those machines were only better than humans in some ways, whereas AIs are better in every way. But that doesn't make sense either. There are lots of humans who are better workers than you or me in every way, but they don't create unemployment. I can't really see the difference between creating AI and creating more people. Our population has grown tremendously in the past century, but there has not been a similar increase in unemployment. Fertility does not cause unemployment. Why would it differ if the new people it created were made of silicon instead of carbon?


The only real qualitative difference I can find between creating new people and creating new AIs is that humans have limitless and insatiable desires. Therefore they constantly demand more and more goods and services no matter how wealthy they become, creating a constant new demand that must be filled by hiring more and more employees. AIs, by contrast, might not be programmed to have insatiable desires, or even be fully sentient, and therefore fail to create new demand. Is that the difference? It's the only one I can think of that even begins to make sense.

Floccina writes:

From my admittedly narrow perspective the internet and China eliminated certain jobs faster than jobs could be added in other areas. In response Greenspan pushed interest rates down low enough that housing absorbed enough workers to mask the situation. Now that bubble has burst we need some deep changes. There is as always plenty of work to be done but it takes time for new services to evolve and be accepted.

Colin K writes:

The problem I see is that Lanier (and many others) seem to be equating "wealth" with having a job.

A hundred or two years ago, being able to sleep in a warm place and eat a full meal including meat and beer with some entertainment after, and not having to work all day in the fields, would have fit most peoples' definitions of being "wealthy."

Today, provided you have no larger aspirations in your life, it is possible to get there with little or no labor. In China or India, if you do not toil in the field or factory floor, then you do not eat. Most of the US underclass has access to a car, cable television, internet porn, a video game console, and is far more likely to suffer ill health from obesity than lack of calories. Is it no surprise that the least motivated and future-oriented among us would choose more leisure?

Anyway, if a lot more goods and services become subject to some form of Moore's Law, then most people will need to work a lot less. Work does produce status along with cash, but employment-related status can be easily substituted with status provided by leisure activities, as the idle rich are wont to do.

Costard writes:

So if we made it essentially costless to produce apples -- everyone would starve? And if a computer does not fill an entire room and cost $500,000 to build, nobody will be able to afford one? And basically every invention since mammoth-killing has only driven us deeper into our poverty?

Does this theory explain how, in an age of free information, a man can say something so globally stupid?

He would make a good Keynesian, though. Modern man has 1/1000th the marginal product of neanderthal man... one wonders how much stimulus would be required to fix THAT.

Comments for this entry have been closed
Return to top