经济与社会科学2019economics.mit.edu
人工智能与劳动力需求的未来
Artificial Intelligence and the Future of Labor Demand
入门 · 约 60 分钟读完前置知识：劳动经济学基础、宏观经济学基础
导读
这篇论文探讨 AI 对劳动力需求的影响，提出 AI 不一定只能用于自动化替代人类任务，还可以通过“增强型”AI 创造新的高生产率劳动任务，从而缓解劳动需求停滞与不平等。作者认为市场机制可能无法充分引导这类“重建型”AI 的发展，需要政策干预。
HN 关注点：AI 对就业的长期影响，讨论焦点在于 AI 究竟是替代还是增强劳动力，以及市场能否自发产生“正确类型”的 AI。
HN 评论精选
这几条评论里：质疑 3 补充 1
质疑论点抽象上成立，但作者对 AI 的理解脱离当前技术现实
我认为作者想论证 AI 不必局限于自动化直接取代人类工作的任务，也可以增强人类劳动，让人们做以前无法想象的新事情。这类似于经济学里推土机的老故事：推土机不是让人拿着铲子失业，而是让人能盖摩天大楼而不是只能搭棚屋。这个论点抽象上成立，但和许多外行作者一样，他对 AI 的认知主要来自《奇点临近》《超级智能》这类书，与当前技术现状及可预见的未来发展完全脱节。
DebtDeflation查看英文原文
I think he's trying to make the argument that AI need not be limited to automating tasks that directly replace human jobs but can also augment human labor in a way that enables people to do new things that couldn't previously be imagined. It's a variation on the old economics story of how the bulldozer does not replace humans with shovels, it enables those humans to build skyscrapers instead of being limited to building shacks. He's making a valid argument in the abstract. However, like so many authors from outside the field, his concept of AI is primarily informed by works like The Singularity Is Near and Superintelligence, and completely disconnected from the current state of the art and its development for the foreseeable future.
质疑教育 AI 市场会自发发展，医疗则因监管根本算不上市场
有趣的论文。他们说市场未能催生某些类型的“重建型”AI，并举了教育等三个例子。但教育 AI 我不这么看：很多人正在开发能评估并因材施教的 AI，因为大家意识到一旦做出成果市场有多大，我认为市场会产出这一类。至于医疗，那里充斥着监管和立法保护的特殊利益集团，怎么能叫市场？如果真是市场，现在早就大不一样了，甚至可能激励出“正确类型的 AI”。
geomark查看英文原文
Interesting paper. They say that markets fail to bring forward some types of "reinstating" AI and they give three examples, education being one. But I think otherwise when it comes to education AI. Lots of people are trying to develop that kind of thing, an AI that assesses and tailors to individual learners' needs. And I think so many are trying to do that because they recognize what a huge market there will be if they can deliver results. I think the market will yield that one. Then they mention healtcare, which is such a quagmire of regulations and legislatively enabled special interest groups that how can one call that a market. If it really was a market it would already look very different than it does now and might even incentivize the "right kind of AI".
质疑摘要论断缺乏正文证据，AI 创造的新岗位问题本质是老问题
我没看到摘要中“当前劳动需求停滞和贫富不均是由 AI 导致”这一论断在正文中有证据支撑，如果属实我会非常惊讶。就个人见闻而言，AI 也在创造新岗位，比如数据科学很大程度就是 AI 的直接产物。这不代表没有问题：被 AI 自动化淘汰的岗位（如呼叫中心和司机）从业者未必能进入新领域，但这更像“底特律问题”的老版本，而非新现象。
lucozade查看英文原文
I don't see where the assertion made in the abstract, that current labor demand stagnation and wealth inequality is due to AI, is evidenced in the body of the paper. And I'd find it very surprising if it were true. Anecdotally, I'd also argue against the thrust of the paper. New roles are being created as a consequence of AI. One could argue that a significant proportion of data science is a direct consequence, for example. That's not to say that it won't be problematic. It's not at all obvious that the occupants of jobs that will likely disappear through AI automation (call centres and drivers in the immediate future) are the people that will be employed in the new fields. However, this strikes me as a version of the "Detroit problem" rather than something new.
补充自动化并非错误方向，关键是让被替代者也能用上 AI 工具
我不同意“自动化就是错误类型的 AI”这一前提。文章所说的赋能工人、创造“许多新的高生产率劳动任务”，大多需要先把较简单的例行任务自动化出来才能委托。所以关键问题在于：如何让 AI 工具被那些任务被自动化的人自己获取和使用。
pseudosimus查看英文原文
> Most AI researchers and economists studying its consequences view it as a way of automating yet more tasks. No doubt, AI has this capability, and most of its applications to date have been of this mold — e.g., image recognition, speech recognition, translation, accounting, recommendation systems, and customer support. But we do not need to accept that this as the primary way that AI can be and indeed ought to be used. I don't agree with the premise that automation is necessarily "the wrong kind of AI". Empowerment of workers as it is described in the article as a means of creating "many new, high-productivity tasks for labor" mostly requires automation of simpler routine tasks to be delegated. In that sense, I think the main problem is making AI tools accessible to the same people whose task is automated in the first place.
历次讨论
- 2019Hacker News 讨论帖146 分 · 33 评
同领域还值得读
一切的回报率，1870–2015
The Rate of Return on Everything, 1870–2015
利用1870–2015年多国的长期数据，估算股票、住房等各类资本的回报率，并将其与经济增长率进行比较。
经济与社会科学2019 · 487 分 · 260 评进阶 · 约 100 分钟编辑精选一切资产的回报率，1870–2015
The Rate of Return on Everything, 1870–2015
构建了 1870–2015 年 16 个国家的长周期资产回报数据集，覆盖股票、住房、政府债券和短期票据，系统比较各类资产的真实回报与波动性。
经济与社会科学2018 · 215 分 · 151 评进阶 · 约 120 分钟编辑精选一切资产的回报率，1870–2015（2019）
The Rate of Return on Everything, 1870–2015 (2019)
这篇发表于《经济学季刊》的论文构建了 1870–2015 年间 16 个主要发达国家的股票、债券、房地产等资产收益率的长周期数据库。
经济与社会科学2024 · 190 分 · 148 评进阶 · 约 80 分钟编辑精选想法是否越来越难找？
Are ideas getting harder to find?
探讨“想法是否越来越难找到”：通过衡量多个领域的“研究生产率”，发现维持相同的增长需要投入指数级增长的研究资源，即研究效率在持续下降。
经济与社会科学2016 · 182 分 · 106 评入门 · 约 60 分钟编辑精选比特币与区块链的经济极限
The Economic Limits of Bitcoin and the Blockchain
芝加哥大学 Eric Budish 的论文从经济学角度分析比特币与区块链的根本限制：核心论点是只要攻击网络（双花或砸盘做空）的收益超过挖矿成本，51% 算力攻击就有利可图，而挖矿…
经济与社会科学2018 · 176 分 · 37 评进阶 · 约 90 分钟编辑精选