I have studied health economics, development economics, and educational economics to the extent that there are definitely around 100 people in the top 300 around the world. I am also one of the 10 editors of the second-best journal in my field of development economics. With the Major League having 40 people per team * 30 teams total around 1200, at least in my field, I’m a pretty good starting player.
However, AI has completely surpassed my level of knowledge and analysis in my major field. I am no match for this. Just a year ago, when AI wrote a review report, there were some great contents, but mistakes were noticeable. But now, AI reviews that are released in just a few minutes are better than those that I spent hours to write thoroughly. Mistakes have almost disappeared. AI is more than Shohei Ohtani + Aaron Judge. It has exceeded the collection of only the strengths of all the players that exist.
The same is true of finding research topics. Now, AI finds more diverse policies than I do. Even the methodology to analyze the policy is much more systematic than the way I was contemplating alone.
Similar stories are told at a gathering of top professors in the medical community. Even professors at Yonsei University and Seoul National University acknowledge that AI has already passed their expertise. Economists are no different. However, AI was at the level of resident knowledge a few months ago.
Not long ago, I gave AI researchers public data that is well used in health studies, and suggested that they find research topics on their own and make 10 papers a day, 300 a month. 300 corresponds to the lifelong research achievements of scholars who work hard in this field (though not creative). I wanted to let them know that trivial papers examining the degree of connection are meaningless now. But their response was that it was “already such a natural story” that it was not worth trying. Wow…
Looking at this trend, it seems clear that AI will not be able to easily write high-quality papers that I want to write in the near future. So I’m currently focusing on research that AI can’t do yet, or research that requires data that AI can’t learn.
Let me tell you what it is,
- Large-scale social experiments (RCTs) — Data such as social experiments on tens of thousands of people cannot yet be created by AI itself.
- Combined analysis of administrative data with limited access — It is a method of analyzing administrative data such as health, education, and welfare that was previously unavailable.
For the time being, AI cannot completely replace such research due to ‘lack of data’. However, this is also replaced as soon as AI has access to data.
Now that AI has already surpassed the knowledge and capabilities of a world-class scholar, as a researcher, I feel like a “frail human being standing alone in front of a huge wave.” It is hard to guess what the role of human researchers will be in the future. In that sense, I also changed the cover photo. I am afraid.