- It’s hard to escape the feeling that we’re living in an uncomfortable moment in which the capabilities of AI models to perform research and economic analysis have far outpaced our understanding of how best to deploy and oversee them.
- “Every conversation I have with a colleague at Harvard or a think tank in Washington is a combination of excitement, anxiety and confusion,” Harved economist Jason Furman told Axios.
State of play: Furman, who was a top White House economist under President Obama and now co-teaches Harvard’s introductory economics course, created a site called Econvitals, which went live Friday, with a wide range of data and analysis.
- Nothing on the site is revolutionary new or cutting edge. But neither are most economic analyses.
- What’s amazing is that this is a site that a very busy professor made in his spare time, with a little help from research assistants and a lot of help from Claude (with some cross-checking by the OpenAI Codex).
- It offers, for free, the kind of in-depth calculation and analysis that previously required thousands of man-hours and a seven-figure budget.
Zoom: For example, the site includes an AI analysis of Federal Reserve officials’ speeches that shows how hawkish or dovish each official is and how their views have evolved over time.
- It’s certainly not rocket science, but it’s the kind of work that Fed watchers at banks and consulting firms have been doing on a more artisanal basis for ages, serving paying customers.
- It produces email notes summarizing interesting economic research and analyzing the latest data that may not have the human touch of the best email tip sheets (ahem), but are accurate, effective, and free.
- Furman shows samples of economic documents written entirely by AI, using short prompts. These are not the kind of results that would land a young economist a job, but they are remarkably competent considering the little human effort involved.
What they say: “In terms of factual reliability, at this point, if they’re done correctly, they’re better than any human being,” Furman says. “If you have fact-checkers, what comes out is completely reliable. The idea of hallucination is a combination of outdated or incompetent usages.”
Yes, but: “It’s still not as creative and exciting as the best humans,” he adds, and it’s more effective at replacing routine analysis than at generating breakthrough ideas. That said, the definition of what constitutes a routine continues to change with each new version of the boundary models.
- “It’s really exciting to see what AI can do for us and empower all of us,” he said. “Part of it is scary. I find he can do a lot of things that I do better than I did, and I thought I was pretty good at what I did, and that can be discouraging sometimes.”
- “My hope for economic research is that we see not an increase in quantity but in quality. Any idiot could write 300 articles a year now, and they would be perfect and factually accurate. But I hope that people will not write more articles but will write better articles,” he said.
The plot: Furman says that if he were leading the White House Council of Economic Advisers today, he would rely heavily on AI for his analyses.
- “If I were running CEA today and had a choice between 20 staff with both current and future doctorates and the AI systems we have today, I would choose AI systems with some sadness,” he said.
- “It would take people at CEA interacting in meetings, talking to the press, talking to Congress. There are a lot of things that only humans can do, but that’s only a fraction of what we’ve done at CEA. For the most part, at this point, I think machines are faster and more reliable than humans.”
The bottom line: So what does Furman say to the undergraduates he teaches, especially those who aspire to do research or shape economic policy?
- “I think people need to understand how they can add value and do something distinctive and unique. Repackaging and reinterpreting data and numbers is not that unique and exciting compared to other things people can do.”
Gn bussni
