网易首页 > 财经频道 > 正文

麻省理工教授警示:人工智能的最大误区是这个!

2020-06-19 08:31:18 来源: 大师 网易号 举报
0
分享到:
T + -

专访诺奖得主基德兰德:没必要抵制美联储加息

本文为网易研究局与罗汉堂联合出品的《新经济思想系列》文章,未经授权请勿转载

·聚焦国际思想市场·解析财经新闻热点·对话国际经济学大师

作者|达隆·阿齐默鲁(美国经济学家、麻省理工学院经济学教授)

凯梅尔·达隆·阿齐默鲁(Kamer Daron Acemoglu),出生于土耳其伊斯坦布尔的美国经济学家,现任麻省理工学院应用经济学教授,2005年度美国经济学会约翰·贝茨·克拉克奖获得者。

自动化对劳动者的冲击

几乎每个人在谈到人工智能这种非常激动人心的技术时,都把它归为一种自动化技术。但更确切来说,人工智能领域更像是一个平台。人工智能可以被应用于各行各业,提高人的生产力并根据数据变化做出实时调整,从而使教育更加个性化、使医疗更加个性化、使生产过程中人和工具的交互更为紧密。但是,当下各行业中龙头企业所采用的商业模式以及技术生态系统引导技术变革的方式,在很大程度上都是面向自动化的。

人们对自动化和机器人最大的顾虑就是对职业以及劳动者带来的影响,这种担忧是有理由的。因为在过去的二三十年里,新技术的发展在很大程度上是沿着自动化的道路进行的,工业机器人就是一个例子。而在此之前是数控和计算机控制机械。现在,人工智能和自动化通过减少昂贵劳动力投入来提高生产力,但也因为人工智能取代了人力,劳动收入在国民收入中的占比会产生波动,进而也可能对薪资、就业产生负面影响。   

自动化研究中最棘手的一点就是它具有多面性,并且经常和其他问题捆绑出现,我们又很难将其分拆、摸清其中的运作方式。以美国农业机械化时期为例,这是自动化过程中非常重要的一个环节。以前农民占总劳动人口的一半,在大约80年的时间里,这个数字快速下降,农业向机械密集转型。如果仅此而已的话,就业形势和薪资水平会受到巨大负面冲击。

但幸运的是,美国经济与此同时创造了许多新的就业机会,无论是在服务业还是制造业,都提供了新的工作岗位、新的高生产率技术,从而成功提高了就业率和薪水。

如何提升劳动者待遇

现在再回顾那段时期,人们可能会说“我们既然有能力扛住变革带来的压力,现在又有什么可担心的呢?”但是今天美国的情况是,因为技术重心过于聚焦在自动化,我们只看到了自动化的部分,却忽略了与之相对的技术。所以,如果说自动化在当今有何意义的话,我认为工业机器人有很好的代表性,你能看到明显的生产率增长,比如汽车工业和电子工业现在有更高的生产率。但工人——尤其是生产人员——并没有从中受益,生产工人的比例急剧下降,在很多行业中,那些占据主要劳动人口的生产工人最终却不得其所。所以回到“自动化具有巨大潜力”的论断,我们需要给它加上一个条件:即只有和技术的其他用途相结合时,才能提升劳动者待遇。

虽然我们创造了新的工作岗位和新的产业,但是在过去30年间,出于各种各样的原因,我们的重心十分单一,除了自动化还是自动化。因此,自动化对劳动力市场的影响,尤其是对中低技术工种,是非常糟糕的。当分析自动化效用的时候,对劳动力来说,有两种相互抵消的作用力十分重要,但我的了解也还十分浅显,也不知道我们在多大程度上可以利用它们。首先,企业正在不断引入自动化技术,因为它利于创收、提高生产力的事实是毋庸置疑的。这种生产力效应对工人有一定益处,但是,工人从生产率效应中受益的程度是有限的。因为我们可以设想一下自动化带来的生产率效益是从何而来,它并没有提高个体劳动者的生产力,而是使雇主可以花更少的钱来替代现有劳动力。因此,我们不能只依靠生产力效应来抵消自动化造成的替代作用,自动化和全行业劳动者福利也无法形成水涨船高的关系。我们必须同时使用新技术来创造新的岗位、新的职业、新的产业、新的生产方式,这样我们才能将劳动力转移到新的岗位需求上,激发更高的生产力。这是我们从过去的经验中所目睹的,也是我相信我们现在有潜力做到的。

以特斯拉为例,它是一个有代表性的,也是许多非常有创造力的企业家都在思考的例子。现在的问题是,想要摆脱人的主观因素影响,把失误率高的人为因素从生产过程中剔除,但特斯拉也证明了这并不管用,哪怕是最理想化的状态下。因为人类独有的一些技能是机器无法复制的,即使这些技能要实现机械化,也需要可能100年、200年,甚至更长时间。所以人是不可或缺的,但我们需要人采用更先进的技术,这是我们长期以来欠缺的,这其中有很多发生在发达经济体。

推动技术和劳动者双向配合

大多数中等收入国家不是自动化,因此,他们有很多时间借鉴美国、德国和日本经验,并使之贴合国情。但我认为这种观点是错误的,因为影响发达经济体的因素会立刻传导到发展中国家。原因在于过去,韩国等国家采用的一个非常成功的发展战略是从劳动密集型产业切入。在全球价值链中,可以像这些国家一样提升劳动力价值、增加他们的附加值,而他们也确实沿着增值链上升。但是,现在这将变得非常困难,因为许多劳动密集型工作都可以由机器人或其他自动化技术来完成,这样一来发展中国家和发达国家是同舟共济的。制造业将继续发挥重要作用,许多我们喜欢和消费的商品需要制造。但很明显,高技能型服务和其他高价值服务也同样重要,不管哪种情况,我们需要找到更好地发挥人类潜能的方式。同时,我们还必须找到与之配套的训练和教育方法,这样他们就能在更现代化的劳动力市场上更有效率地工作。所以,重点就是目前为止我强调的——我们需要新产业,而这些新产业也需要新技术,比如说,在培训男性劳动力的时候重视加强对事物控制力的身体机能。

但是可能就忽视了工作的灵活性、适应能力和在新环境中的计算能力和社交能力。这些技能要求在60年代和70年代是很应景的,但在我们需要创造的新产业中却略显过时了。因此,我们需要同时推动这两个杠杆,鼓励技术朝着有利于劳动力的方向发展,也让劳动者准备好应用这项技术。

我认为经济学有很多有趣的领域,部分原因是经济学家越来越多地参与到更广泛的社会问题中来,反思社会规范是其中一个方面。这真的很令人振奋,我们需要在传统经济学上做更多的革新,要考虑到租金上涨的影响、市场影响力、市场结构。虽然我们所看到的一切并不都与市场结构有关,但是当有一些企业在他们的行业中占据主导地位,我们不可避免地要考虑到垄断资本,但这种垄断资本不会像过去的卡内基(Carnegie Corporation)、美孚(Standard Oil)甚至是美国电话电报公司(AT&T)那样肆无忌惮。所以,这是一种新的思考方式。最后,我认为,技术影响工作的方式、创造赢家和输家,我们需要更全面的思考。

English Version:

Almost everybody who talks about about AI, a very exciting technology thinks of it as an automation technology, but in fact area is a platform. You can use AI to do thousands of things, to increase human's productivity, and to enable humans in real-time adapt to data, making the education more individualized, making healthcare more individualized and making the production process have a much more seamless interaction between precision tools and and workers. But the problem is that the business models of the most dominant companies today and the way that the ecosystem of technology supports the technological change is very much geared towards automation.

There's a lot of concern about what automation and robots are going to do to work and to workers. I think there are reasons to be concerned. Because the way that new technologies have been developed over the last two-and-a-half three decades really is going very much down the path of automation, industrial robots being one example. But before then numerically controlled and computerized control machinery and now AI and automation increases productivity by saving on expensive labor. But it displaces labor and therefore it has negative effects on the labor share of national income and it might have negative effects on wages and employment.

I think one of the confusing things about studying automation is that we see various different facets of automation but they often come bundled with other things and it's very difficult to unbundle them and understand what they're doing. Take for example the period of mechanization of Agriculture in the U.S. that's a very major process of automation , 50% of workers used to work in agriculture and over the course of about 80 years that went down tremendously and agriculture became much more machine intensive . If that was the only thing that happened, it would have been a tremendous negative shock to employment and wages. But fortunately during the same process the U.S. economy generated lots of new jobs also new tasks in services in manufacturing with clerical occupations, with new industries and new types of technologies that were very high productivity and in fact managed to raise employment andwages.

So looking at that period people then say “oh you know we weathered that particular storm very well why should we be concerned “ But what we are seeing in the U.S. today is that because technological focus has been so much on automation, we are seeing the automation part of it but we're not seeing the countervailing technology. So when you ask about the implications of automation today, you know I think they are summarized very well by industrial robots so they increase productivity you see in the autoindustry and the electronics industry there is higher productivity today.  But workers don't benefit, especially the share of production workers falls tremendously and many of the areas where the significant fraction of the workforce was working in these industries as production workers end up suffering as a result. So again that goes back to the same set of ideas that automation has tremendous potential but only when combined with other uses of technology.

So that we lift workers and we create new jobs and new tasks and for a variety of reasons over the last three decades, the singular focus has been on automation, therefore labor market implications formost workers especially for middle skilled low skilled workers has been very bad.

When you think about the effects of automation. There are two important countervailing effects that help workers and I never understand them and understand what the extent to which they we can utilize them. One is that automationis being adopted by firms because it's actually profitable it's increases their productivity. That productivity effect helps workers but the extent to which workers benefit from the productivity effect is somewhat limited and the reason for that is if you think about where does the productivity benefits of automation come from. It's not that it's making workers more productive it's that it's enabling employers to substitute cheaper capital for workers.

So if we only have to rely on productivity effect in order to counterbalance the displacement created by automation, that would be imperfect and we wouldn't be creating the sort of the tide that lifts all boats. So wereally have to at the same time use new technologies in order to create new tasks, new occupations, new industries and new ways of organizing productions, so that we reinstate workers into new jobs in which they can have high productivity and that's again what we have witnessed in the past and I very much believe that we have that potential to do the same thing today.

Take for example Tesla, I think it's a representative of how many very creative entrepreneurs are thinking about. The problem is that you know you want to get out, you wanna get the fallible human factor out of the production equation, but also Tesla proved that it doesn'twork said that well. Because you know humans have unique skills that machine's are not going to be able to replicate for perhaps 100 perhaps 200 perhaps even longer. So we really need the human, but we need the human to work with the better technology and that's what we've been failing it. A lot of this is taking place in developed economies.

Most middle-income countries are not therefore they have a lot of time to learn from what's going on in the U.S. in Germany and Japan and adapt to it. But I think that's the wrong perspective, because what's going on in the developed economies today is immediately affecting the developing countries and the reason for that is because in the past a very successful development strategy which was exploited by South Korea, China is to start from the labor-intensive tasks in the global value chain learn from them build and upskill your workforce and increase this the value-added. They did go up the value-added chain. That's gonna become very hard when many of these labor-intensive tasks now can be performed by robots or by other automation technology. Therefore, developing countries and developed countries are in this boat together manufacturing is going to continue to be important. Many of the good that we like and consume need to be manufactured, but obviously high skill services and high-value other services are going to be very important too. But in both of them I think we have to find ways of better using humans and at the same time we also have to find ways of training and educating humans, so that they can more productively work in a more modern labor market. So if you think of the emphasis that I've placed so far on new tasks, those new tasks also require new skills. So if you educate men for example to be putting emphasis on physical skills manipulating things and not having the flexibility to transform what they're doing not having the adapt ability not having the numeracy skills and the social skills to work in a new environment. Then those are the skills that would have been fine in the 60s and the 70s , but they're not going to work very well with the new tasks that we're required to create. So we really need to push on both levers at the same time encourage technology to go in the direction of helping labor but also make labor more ready to work with that technology.

I think there are so many interesting areas of economics and part of this excitement in my opinion comes from the fact that economists are being more and more engaged with broader social questions. I think thinking about social norms is one of the areas. That is really very exciting and we need to do more on that more traditional economics, think of the implications of rising rents market power and markets structure. I don't think that everything that we're seeing is related to market structure, but when there are a few companies that are playing such a dominant role in their sector and then in the economy you really have to think about monopoly power. But that monopoly power is not gonna play out in the same way that the power of the Carnegie Corporation or the Standard Oil or even the ATT played out in the past. So it's a new way of thinking about it and then finally I think the way that technology impacts work and creates winners and losers. I holistic thinking.

网易研究局(微信公号:wyyjj163) 出品

网易研究局是网易新闻打造的财经专业智库,整合网易财经原创多媒体矩阵,依托于上百位国内外顶尖经济学家的智慧成果,针对经济学热点话题,进行理性、客观的分析解读,打造有态度的前沿财经智库。欢迎来稿(投稿邮箱:cehuazu2016@163.com)。

北京无雾霾?这个冬天 帝都的雾霾都到哪里去了

移驾微信公号 看这里看不到的内容

【精彩推荐】点击进入网易研究局·中国版>>

【精彩推荐】点击进入网易研究局·国际版>>


延伸阅读
李兆元 本文来源:大师 责任编辑:李兆元_B7890
分享到:
跟贴0
参与0
发贴
为您推荐
  • 推荐
  • 娱乐
  • 体育
  • 财经
  • 时尚
  • 科技
  • 军事
  • 汽车
+ 加载更多新闻
×

有些人很少护肤,却也显得很年轻?

热点新闻

态度原创

网易号

查看全部
阅读下一篇

返回网易首页 返回财经首页