人工智能会不会导致大量失业

从去年开始我遇见了不少讨论这个主题的文章。首先关于会不会有很明确两边对立的观点,各自都有所理据,最近甚至讨论到大量替代之后带来的包括经济衰退社会动荡等等不良后果。我隐隐约约是相信不会的,但是也经常提醒自己是不是小看了人工智能的发展,或者轻信了这次会一样。

在我看来首先一个关键性要确定是人工智能能到达的工作能力的程度。如果能在各种方面十分接近人类的话,我觉得这个话题就谈都不用谈了,不是导致失业了,是人类没有存在意义了。所以我个人还是觉得限制在人工智能能高效完成当前社会的某个单线程工作这么个前提条件下来看问题。

然后最关键的一个问题就是,人工智能会不会在将来的某一时间导致突然的,大量的,长期性的失业。

否的理论原因还在思考,先记录几个我想到的问题,日后总结。

最近日本就业市场求人比率到2,失业率达到了几十年来最低,但是工资仍然不上升,为什么?失业率的下降是不是因为现在科技发展的速度还赶不上少子化速度?

我学校里有时候有什么会议的时候,会见到有人站在学校旁的路口,就拿一块牌子指着场地方向。这大概我能见到的最简单的工作了,为什么没有被更加高效的方式替代?我们系至今为止还发布信息竟然还是靠玻璃告示栏,电子化程度不堪入目,为什么?

而像投行这样的工作,高强度长时间高工资而公司人员精简,新招一个低层级员工的边际收益很高,为什么却不这样做?

人工智能最被期望与代替重复劳动的工作,我从五年前投资A股的时候就有机器人热点,说某某工厂现在导入了全自动生产线,现在怎么样了,为什么太大声响了,为什么反而为了税到美国去建厂了?

我最近想买个扫地机器人,irobot,非常贵,相当贵,犹豫了半天,放弃了。你说如果他要是定价能和小米手机一样,岂不是家家一个,卖上天?所以为什么让他无法降价?

古有马车,今有司机,都是最典型的被替代的命运,也没什么再就业手段,似乎很说的通,但等等,马车夫失业以后司机是怎么冒出来的?

Noah Smith的一篇文章在我看来非常好的解释了为何以上的一些无效工作案例说明机器人彻底消灭工作岗位是不太可能的。

See, the so-called lump of jobs fallacy really is a fallacy. Many people think that if you eliminate an existing job, the number of available jobs in the economy goes down by one. But this is simply not support by the facts. The rate of job churn in any modern economy — even Japan’s — is much, much larger than the rate of net job creation or destruction. When people’s jobs are eliminated, they almost always find new things to do, unless the country is in a deep recession. 

Of course, there are costs involved with eliminating an existing position — for example, if people have to retrain. But the skill of pushing an elevator button or guarding an empty lot isn’t exactly something that requires one to go back to school. And government can help laid-off people look for work, through websites and other services. 

如果看过去二十年的通货膨胀的话,伴随着科技进步,TV软件电话服务大幅度通缩,而教育儿童医疗餐饮则通货膨胀。或许人工智能推崇者最后会发现,被替代最多的不是别人,而是他们自己。

 

2017/08/10 update

The Robot Takeover Is Greatly Exaggerated

It makes sense to be worried:
[1] the extreme version: automation simply makes human workers obsolete, just as cars made horses redundant
[2] a less apocalyptic possibility is what economists call “skill-biased technological change” — people who are technically savvy, mentally flexible and educated will reap greater and greater rewards while everyone else sees their wages decline.
→ some/all either impoverished or reduced to living off of the government dole

Although adoption of machines in the past didn’t make human labor obsolete so far:
[1] if it had, we’d be seeing faster productivity growth and higher unemployment
[2] claims that companies are substituting machines for humans more than in the past are not few, but so far the evidence remains scant
[3] the example of the Luddites: they could always find ever-more-lucrative jobs that made use of new tools → previous technological revolutions always ended up making the mass of humanity more valuable, not less
[4] reviewing economic history, new technology has complemented human labor rather than replacing it (most macroeconomic models assume that the relationship between technology and humans is basically fixed)

[5]  though the paper by Acemoglu&Restrepofinds in MIT evidence that robots are already costing American jobs, Mishel&Bivens of the Economic Policy Institute(a think tank) examine MIT paper in detail, and note that capital investment, and use of computers specifically, tend to increase jobs, as Acemoglu&Restrepo themselves wrote

“(comparing to industrial robots – fully autonomous multipurpose machines with no human operators -) other types of capital equipment and even computers tend to increase the demand for labor…thus a very different impact on employment and wages.”

[5.1] + workers haven’t been changing occupations as much as they did in past decades — if people were losing their jobs to automation at a faster rate, we’d expect them to have to retrain more frequently
[5.2] productivity growth and corporate investment in IT has fallen, which also doesn’t fit with a story of accelerating automation (actually an investment drought in rich countries, as companies sit on cash)
[5.3] negative impact of Chinese competition on U.S. jobs was more than three times larger than the effect of robots

there’s no guarantee that future technology will work the same:
[1] robot revolution may be more of a long-term concern, driven by the rapid advances in machine learning → a world machine do anything better than human? / capital-labor substitutability would increase dramatically — you don’t need a human to operate the machine tool or the computer (← it’s always hard to predict future technology)
[2] before 1800 before Industrial Revolution, an economist would conclude technologies of the past have never allowed the mass of the species to escape from poverty and malthus trap  → new technologies were qualitatively different (though resulted in an abrupt acceleration of wealth generation)
[3] If the elasticity of substitution between capital and labor (how easy it is to replace humans with machines) goes up, labor’s share of income can go down and down + Skill-biased technological change rewards the top workers while punishing everyone else
[4] more extreme, huge inequality: people who won at the beginning of the artificial-intelligence revolution would start off in a privileged position, able to outcompete new entrants in winner-take-all markets / people who started out with the capital would amass essentially all the wealth in society, while those who started out with less would be out of luck.
[5] there will always be more work for humans to do, despite more absurd, esoteric jobs ← the problem is that such jobs will not necessarily pay a living wage ← if capital-labor substitutability increases dramatically, wages may fall a lot ← either humans will survive via redistribution, or they won’t survive at all ← redistribution depends on politics, and in an age of robots, the masses may have very little power to compel the wealthy to give up even a small sliver of their wealth
“Once the military became robotic, revolution became impossible.”

when is next difference? (source)

1x-1

 

PS.
more on real, pressing problems:
[1] Inequality in developed countries remains severe and potentially harmful to civic society.
[2] The productivity slowdown is disturbing and imposes a drag on growth.
[3] Reduced economic dynamism, including falling startup rates and less job-switching, is a concern.
[4] Monopoly power is a looming threat to the health of the economy.
[5] Although wages have been rising recently, those gains come after many yearsof stagnation or decline.

 

Update 2017/9/11

「AI失業」は起こらない

Robots and Artificial Intelligence

 

Update 2017/10/19

Lessons from history for the future of work

 

 

 

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s