人工智能如何影响你的工资

如果无法播放，请点击此处在新窗口打开

人工智能如何影响你的工资(上)
Finance and economics
财经版块
What does AI mean for your pay?
人工智能对你的工资意味着什么？
A dispatch from the front line of an economic revolution.
从经济革命前线发来的消息。
Around a decade ago Carl Benedikt Frey and Michael Osborne, two economists, published a paper that went viral.
大约十年前，两位经济学家卡尔·贝内迪克特·弗雷和迈克尔·奥斯本发表了一篇论文，这篇论文在网上被疯传。
It argued that 47% of American jobs were at risk of automation.
论文称，47%的美国工作岗位面临被自动化的风险。
A deluge of research followed, which suggested the poorest and least-educated workers were most vulnerable to the coming revolution.
随后出现了大量研究，研究表明，最贫穷和受教育程度最低的工人最容易受到即将到来的工作革命的影响。
Such fears have intensified as artificial-intelligence (AI) capabilities have leapt ahead.
随着人工智能发展的突飞猛进，这种担忧也在加剧。
On November 2nd, speaking after Britain's AI summit, Elon Musk predicted: "There will come a point where no job is needed."
11月2日，埃隆·马斯克在英国人工智能峰会结束后发表讲话时预测道：“终有一天，人们不再需要工作。”
Yet at the same time, economists have become more optimistic.
与此同时，经济学家们也变得更加乐观。
Recent studies have found that fewer workers are exposed to automation than Messrs Frey and Osborne supposed.
最近的研究发现，面临自动化风险的劳动者要比弗雷和奥斯本以为的更少。
In 2019 Michael Webb, then of Stanford University, showed that AI patents are more targeted at skilled jobs than those for software and robots.
2019年，当时在斯坦福大学的迈克尔·韦伯表明，人工智能专利更多地针对技术型工作，而不是软件和机器人能完成的工作。
New AI seems better at coding and creativity than anything in the physical world, suggesting low-skilled jobs may be insulated.
而新的人工智能似乎比现实世界中的任何东西都更擅长编码和创意类工作，这表明低技能工作可能会免受影响。
In March Shakked Noy and Whitney Zhang, both of the Massachusetts Institute of Technology (MIT), published an experiment showing that ChatGPT boosted the productivity when writing of lower-ability workers more than that of higher-ability workers.
今年3月，麻省理工学院的沙基德·诺伊和惠特尼·张发表了一项实验结果，表明ChatGPT对低技能办公者的写作效率的提升比对高技能办公者更明显。
Although AI is still in its infancy, some industries have been eager adopters.
尽管人工智能仍处于初级阶段，但一些行业已经热切地开始使用这一技术。
A close look at three of these-translation, customer service and sales-is broadly supportive of the optimistic shift among economists, though not without complications.
仔细看看其中三个行业--翻译、客服和销售，可以发现观察结果大体上支持经济学家的乐观预测，尽管也还是有一些更复杂的问题。
In translation, perhaps the first industry to be heavily affected by language modelling, workers have become copy editors, tidying a first draft undertaken by AI, which eases the path of newbies into the industry.
翻译行业可能是第一个受到语言大模型严重影响的行业，这个行业的工作者已经成为文案编辑，只需整理由人工智能完成的初稿，这为新手更容易入行。
In customer service, AI has helped raise the performance of stragglers.
在客服行业，人工智能帮助提高了落后者的表现。
But in sales, top performers use the tech to find leads and take notes, pulling away from their peers.
但在销售行业，能力强者会使用人工智能来寻找线索并做记录，从而领先同行。
Will AI boost the incomes of superstars more than those of stragglers, much as the internet revolution did?
人工智能会像互联网革命那样，给超级明星带来比落后者更高的收入吗？
Or will it be a "great equaliser", raising the incomes of the worst off but not those of high flyers?
还是会成为一个“伟大的均衡器”，提高境况最差的人群的收入，而非高薪群体的收入？
The answer may depend on the type of employment in question.
答案可能取决于所讨论的是哪种就业类型。
Roland Hall has been translating board games and marketing material from French to English for 27 years.
罗兰·霍尔的工作是将桌游和营销材料从法语翻译成英语，他从事这种工作已经有27年了。
He recalls that even in the 1990s software was used to render specific words from one language to another.
他回忆说，早在20世纪90年代，软件就被用来将特定的某些单词从一种语言翻译成另一种语言。
Today the tools are more advanced, meaning the types of job available have split in two.
如今，这些工具更加先进，这意味着现有的工作已经分成两类。
One type includes texts where fluency is less important.
一类包括对行文流畅性要求不那么高的文本。
An example might be a several-thousand-page manual for an aircraft, says Mr Hall, where readers simply need to know "what part to look for" and "do you turn it left or right".
霍尔表示，这种文本可能是长达数千页的飞机说明手册，读者只需知道“应该找哪个部件”，以及“你是把它向左转还是向右转”。
The other type includes literary translations, where the finest details matter.
另一类包括文学翻译，文本的细枝末节都非常重要。
The first type has been most affected by AI.
第一类受人工智能影响最大。
Many workers now edit translations that have gone through a machine similar to that underlying Google's translation service.
许多工作人员都是先用与谷歌翻译的底层技术类似的机器生成译文，然后再对译文进行编辑。
They are paid at a steep discount per word, but more work is available.
他们按字数计算的报酬大打折扣，但工作也变多了。
Lucia Ratikova, a Slovakian who specialises in construction and legal translations, reckons that such work now makes up more than half of listings on job sites, up from a tenth a few years ago.
专门从事建筑和法律翻译的斯洛伐克人露西娅·拉蒂科娃估计，这类工作现在占求职网站上发布信息的一半以上，而几年前这一比例为十分之一。
A larger pool of businesses, many eager to expand into global markets, are taking advantage of the drop in price.
更多的企业正在利用翻译价格下跌的机会，其中许多企业渴望进军全球市场。
If machines are able to do what humans do more cheaply, employers will turn to computers.
如果机器能够以更低的成本做人类做的事情，雇主就会转而使用计算机。
But as prices fall, overall demand for a service may rise, and possibly by enough to offset the increased use of machines.
但随着价格下降，整体需求会上升，从而可能抵消使用机器（带来的工作减少）。
There is no law to determine which effect will dominate.
没有定律可以决定哪种影响将占主导地位。
So far in America the number of translators has grown, yet their real wages have fallen slightly-probably because the profession now requires rather less skill.
到目前为止，美国的译员人数有所增加，但他们的实际工资却略有下降，可能是因为这一职业现在对技能的要求要低得多。

轻音乐播放地址

如果无法播放，请点击此处在新窗口打开

人工智能如何影响你的工资(下)
Customer service offers more difficult terrain for AI.
客户服务的人工智能应用情况更为棘手。
Firms have been trying to automate it for years.
多年来，公司一直在尝试将客服自动化。
Thus far they have mostly just annoyed customers.
到目前为止，这种做法主要只是惹恼了客户。
Who doesn't try to game the chatbot in order to speak to an actual human?
为了和真人对话，谁不会试着对聊天机器人扯谎呢？
The American Customer Satisfaction Index has been falling since 2018, and workers also appear fed up.
自2018年以来，美国消费者满意度指数一直在下降，员工们似乎也受够了。
Turnover in American "contact centres" hit a record high of 38% last year.
去年，美国“联络中心”的员工流动率达到了创纪录的38%。
But there may be consolation: the workforce is becoming more welcoming to the low-skilled.
但还有一些安慰：客服行业对能力较差的人开始变得更加友好。
Erik Brynjolfsson of Stanford, as well as Danielle Li and Lindsey Raymond of MIT, studied the roll-out of an AI assistant to more than 5,000 customer-support agents earlier this year.
今年早些时候，斯坦福大学的埃里克·布林约尔松以及麻省理工学院的丹妮尔·李和林赛·雷蒙德研究了向5000多名客服专员推出人工智能助理的情况。
The assistant offered real-time suggestions to workers.
人工智能助手向客服工作者提供了实时建议。
This lifted the productivity of the least-skilled agents by 35%, while the most-skilled ones saw little change.
这将能力最差的客服专员的工作效率提高了35%，而能力最强的客服专员的工作效率几乎没有变化。
It would be reasonable to assume that the impact on salespeople would be fairly similar to the one on customer-service workers.
由此可以合理地假设，人工智能对销售人员的影响与对客服人员的影响会非常相似。
But that is not the case.
但事实并非如此。
Marc Bernstein of Balto, a firm that creates AI software for both sales teams and call centres, notes that "style points" (ie, charisma and the ability to develop a relationship) matter much more in sales than in customer service, where the important thing is getting the right answer quickly.
Balto是一家为销售团队和客服中心开发人工智能软件的公司，该公司的马克·伯恩斯坦指出，“风格特点”(即个人魅力和建立关系的能力)在销售中比在客服中更重要，客服看重的是迅速给出正确答案。
AI might even create sales superstars.
人工智能甚至可能创造出销售行业的超级明星。
Skylar Werneth has been in the industry for eight years and is now at Nooks, a startup that automates sales.
斯盖拉·沃尼斯已经在销售行业工作了八年，现在在Nooks工作，这是一家将销售自动化的初创公司。
Software analyses his calls, identifying which tactics work best.
软件对他打的电话进行分析，确定哪种策略效果最好。
It also helps him call many people at once.
还能帮助他同时呼叫许多人。
Most customers do not pick up; dialling in parallel ensures Mr Werneth is talking more and listening to dial-tones less.
大多数客户不会接电话，同时拨打电话可以确保沃尼斯多说话，少听拨号等待音。
He reckons the tools Nooks offers makes him three times more productive, earning him a solid amount more than before.
他认为，Nooks提供的工具使他的工作效率提高了三倍，为他带来了比以前更多的可观收入。
What does this mean for labour markets?
这对劳动力市场意味着什么？
Sales representatives are given bonuses based on the number of clients they bring in over a threshold.
销售代表根据他们拉来的超过一定门槛的客户数量而获得奖金。
When productivity grows across a firm, bosses tend to raise the threshold.
当一家公司的工作效率普遍提高时，老板往往会提高这个门槛。
Because not everyone is able to meet it, low performers are pushed out of the workforce, since demand for products does not grow in parallel with sales performance, as would be necessary to justify retaining them.
因为不是每个人都能拉到更多客户，所以表现不佳的员工会被挤出行业，因为对产品的需求并不会随着销售业绩的增长而同步增长，而需求的增长才能证明留下销售人员是合理的。
The result is a shrinking set of highly productive salespeople.
其结果是，销售人员减少，以效率高的销售为主。
At least, given high turnover in the industry, the shift to this state of affairs might mean hiring fewer people, not mass firings.
至少，考虑到销售行业的高流动率，这种转变可能意味着雇佣更少的人，而不是大规模裁员。
If AI eventually becomes superhuman, as many attendees at Britain's recent summit believed possible, all bets are off.
如果人工智能最终成为超人，正如最近的英国峰会上的许多与会者所认为的那样，那么结局非常难以预测。
Even if AI advances in a less epochal fashion, labour markets will see profound change.
即使人工智能的发展不那么惊天动地，劳动力市场也将发生深刻变化。
A study by Xiang Hui and Oren Reshef of Washington University in St Louis and Luofeng Zhou of New York University, published in August, found that earnings for writing, proofreading and copy-editing on Upwork, a freelancing platform, fell by 5% after ChatGPT was launched last November, compared with roles less affected by AI.
圣路易斯华盛顿大学的项辉和奥伦·雷谢夫以及纽约大学的周罗峰在8月份发表的一项研究发现，与受人工智能影响较小的工作相比，去年11月ChatGPT上线后，自由职业平台Upwork上的写作、校对和文案编辑的收入下降了5%。
A survey of 400 call-centre managers by Balto found that the share using at least some AI grew from 59% in April to 90% by October.
Balto对400名客服中心经理进行的一项调查发现，至少使用部分人工智能的客服中心的比例从4月份的59%上升到10月份的90%。
Mr Bernstein thinks that although "today AI is not capable of replacing a human [in call centres]…in ten years, quite possibly five, it will be there."
伯恩斯坦认为，尽管“现在的人工智能还不能[在客服中心]取代人类......但十年后，很可能是五年后，人工智能就可以做到这一点。”
The flipside of AI disruption is new jobs elsewhere.
人工智能颠覆工作的另一面是其他地方会出现新的就业机会。
Modelling in 2019 by Daron Acemoglu of MIT and Pascual Restrepo of Boston University suggests that the impact of automation is worst for workers when productivity gains are small.
麻省理工学院的达伦·阿斯莫格鲁和波士顿大学的帕斯古尔·雷斯特波在2019年进行的建模表明，当生产率提高较小时，自动化对工作者的负面影响最大。
Such "so-so" automation creates little surplus wealth to increase the demand for workers in other parts of the economy.
这种“一般般”的自动化创造的额外财富很少，无法增加其他经济部门对劳动力的需求。
Our investigation of industries at the front line of AI change suggests that the new tech has a shot at leading to much greater efficiency.
本刊对处于人工智能变革前沿的行业进行的调查表明，人工智能有望带来更高的生产率。
The picture on inequality remains murkier.
收入不平等的情况仍然不明朗。
Better to be a superstar than a straggler, then, even if only to be safe.
那么，即使只是为了稳妥起见，努力成为超级明星也比当落后者要好
轻音乐播放地址