新晋诺奖得主讲座：《人工智能如何影响我们的工作和未来的繁荣》

在今年的桑贾亚拉尔纪念演讲中，牛津大学赛德商学院的客座教授达龙·阿斯莫格鲁敦促我们不要相信技术进步自然会带来更繁荣和更美好生活的神话。

在美国尤为盛行的技术乐观主义，让我们相信时间和市场会奇迹般地抹平人工智能等技术的粗糙边缘和短期问题，为我们所有人创造利益。阿斯莫格鲁将此称为 “生产率浪潮”：即随着技术的进步，生产率也会提高，工人的工资也会随之增加。

借用他最近出版的《权力与进步：我们为技术和繁荣进行的千年斗争》一书，阿斯莫格鲁提出了他的论点，说明了在英国工业革命期间，织布等工序的自动化不仅导致工资低迷（纺织工人的实际收入在90年间下降了约 50%），还导致监控加强、生活条件恶劣、过度拥挤、肮脏和疾病。直到 1840年左右，随着制度的变革和工会的强大，情况才开始好转。

他说：“我认为，要了解谁将从新技术中获益，考虑这两点非常重要，因为这两点是历史上谁能获益的驱动因素，即新技术是否只是自动化，还是引入新任务从而提高劳动的边际产出，以及我们如何定义权力关系。”

如果今天的企业继续专注于人工智能自动化现有工作的能力，而不考虑如何利用技术创造新的任务和更好的岗位，那么它们就没有动力提高工资或以其他方式与工人分享生产率的提高。经济不平等和权力差异只会加剧。

但一切并非毫无希望。他说，事实上，如果政府进行适当监管，劳工运动和公民社会采取行动纠正权力失衡并重塑技术发展方向，还有很大的影响空间。人工智能应该用来帮助人类，而不是取代人类。例如，他建议人工智能可以通过提供更好的信息和培训，来帮助解决目前电工和其他行业的用工短缺问题。

因此，我们将面临选择——谁将在使用人工智能的过程中获得更多的权力，我们是要利用人工智能来提高工人的边际生产力和能力，还是要试图让工人靠边站，历史表明我们有很多选择……我认为，过去的经验表明，希望新技术会自然而然地造福民主、自然而然地造福工人、自然而然地造福我们所有人的想法有点过于一厢情愿，但潜力是存在的。如果我们为技术创造正确的权力平衡和正确的发展方向，我们就会有更多的希望。

——达龙·阿斯莫格鲁教授

2024年诺贝尔经济学奖得主

讲座后的讨论环节更深入探讨了全民基本收入、担忧与恐惧之间的界限以及全球不平等等多个话题。他总结道：“在未来十年左右的时间里，我们将看到人工智能的发展程度、人工智能的类型以及谁将从中受益，因此我认为，现实地看待我们所处的位置和能做的事情将至关重要。”

欢迎播放下方视频，观看完整讲座和对话讨论。

In this year’s Sanjaya Lall Memorial Lecture, Visiting Professor Daron Acemoglu urges us not to fall for the myth that technological advancements lead naturally to greater prosperity and better lives. 

Techno-optimism, particularly strong in the US, would have us believe that time and the market will miraculously smooth out the rough edges and short-term problems of technologies such as AI (artificial intelligence), and create benefits for us all. Acemoglu calls this the ‘productivity bandwagon’: the narrative that as technology improves, productivity increases and results in higher wages for workers. 

Drawing on arguments in his recent book, Power and Progress: Our Thousand Year Struggle Over Technology and Prosperity, Acemoglu shows how the automation of processes such as weaving during the Industrial Revolution in Britain led not only to depressed wages (the real earnings of textiles workers declined by about 50 per cent over 90 years) but also to greater surveillance, poor living conditions, overcrowding, squalor, and disease. Things only started improving after about 1840 as institutions changed and trade unions became stronger. 

He said 'I think to understand who will benefit from new technologies it is really important to consider these two things that have been drivers of who gets the benefits in history, whether new technology just automates or introduces new tasks and hence increases the marginal product of labour and how we defined power relations.'

If firms today continue to fixate on the ability of AI to automate existing jobs and do not consider how to use technology to create new tasks and better jobs, there is no incentive for them to raise wages or otherwise share productivity gains with workers. Economic inequality and power differentials will only rise. 

But all is not lost. Indeed, there is all to play for, he says, if governments regulate appropriately and labour movements and civil society act to redress power imbalances and reshape the direction of technology. AI should be used to help humans, not replace them. For example, he suggests that AI could help combat the current shortage of electricians and others trades by providing better information and training. 

So we're going to have these choices - who is going to get more power with the use of AI and whether we're going to use AI to increase the marginal productivity and the capabilities of workers or we're going to try to side-line the workers, and history shows we've have a lot of choices...I think the past shows that automatic hope that new technologies will naturally benefit democracy, will naturally benefit workers, will naturally benefit all of us is a little bit too much wishful thinking but the potential is there and if we create the right power balance and the right direction for technology we have more hope.

Professor Daron Acemoglu

Nobel Prize Economic Sciences Laureate 2024

The post-lecture discussion covers topics as varied as Universal Basic Income, the thin line between concern and fear-mongering, and global inequality. He concludes: ‘It's going to be the next decade or so where we're going to see how much AI, what type of AI and who's going to benefit from it, so I think being realistic about where we are and what we can do is very important.’

Play the video below for the full lecture and discussion.