The Age of AI
Image made with the help of ChatGPT
It has been 14 years since my first blog post: What is a Supply-Side Liberal? I have written an anniversary post every year since:
My blogging continues at a slower pace; in addition to my more immediate duties as a professor, husband, father and grandfather, a key part of my research has heated up: my work with Dan Benjamin, Ori Heffetz, Kristen Cooper, Jiannan Zhou, Tushar Kundu and Hannah Solheim toward developing the principles for national well-being indexes that can stand as coequals with GDP and thereby help create a world where what matters most to people matters most to rulers.
More and more, I wonder if the rulers “creating a world where matters most to people matters most to rulers” refers to will be artificial intelligences. In that case, developing comprehensive measures of well-being will be a key part of AI alignment: it is not enough to make AIs care about human welfare. They also need to have a very sophisticated, comprehensive and open-ended concept of human welfare, lest they foster certain human values while crushing others. Jack Williamson’s classic science fiction story “With Folded Hands”—which he later expanded into the novel The Humanoids—is a warning about this. He tells the story of a future where the robots advance human safety at the expense of many other human values.
Like many others, I have been thinking about AI a lot in the past year. I have avidly read news and commentary about AI. I have worried about unaligned AI. I have made sure I was invested heavily in the stock market on the view that AI will lift the economy as a whole, but for large-cap have invested in an equally-weighted fund, which has no theoretical attraction for me, but is simply a way to bet on AI while also betting that the Magnificent Seven (Nvidia, Alphabet, Apple, Microsoft, Amazon, Tesla, Meta) don’t have good enough moats to avoid facing fierce competition in the future, so that their valuations will leak out to other firms, deflating them in relative terms. I have used AI for the first time to help write exams (especially the false answers in multiple choice exams). I have used songs I created with the help of Suno AI to help students remember key ideas in my Intermediate Macroeconomics class. I have used ChatGPT as a tutor to help me learn about three-dimensional and higher-dimensional geometry. And I have experimented with using AI for blog posts. I know I haven’t gotten things down yet. I have tried to represent interesting dialogues I have had with ChatGPT, but ChatGPT is naturally quite long-winded. Nevertheless, here are the posts I have done in the last year using ChatGPT—most of them based on dialogues I had with ChatGPT simply out of curiosity that I then copy-pasted:
The one full post I wrote this past year without involving AI was
I don’t know how many people share this impression, but subjectively speaking, I feel I was in the same technological era from when I was born in 1960 all the way up to 2022. Then, since November 30, 2022, when ChatGPT was released into the wild, it seems everything has been changing with incredible rapidity. We are going into a period in which human-AI teams will beat humans. What is beyond that is hard to foresee with any clarity.
So, to preserve one’s own competitive position, my advice is to experiment with AI and act like a fan, even if you aren’t. And to the extent their are dangers from AI, it will be those who have been doing that experimentation who will be best equipped to cope with those dangers (though those who won’t touch AI will help in warning of the dangers). We are living inside the fabled Chinese curse: “May you live in interesting times.”