On Thursday, May 11, Kevin Bryan joined Markus’ Academy for a lecture on A User’s Guide to GPT and LLMs for Economics Research. Bryan is an Associate Professor, Strategic Management Area, Rotman School of Management at the University of Toronto.
LLMs have been shown to increase productivity in sales and programming in the past; academia should capitalize on this technology
6 takeaways: (1) Controlling the output of LLMs is difficult, (2) the “Raw” ChatGPT online is far from state of the art, (3) hallucinations are mostly fixable, (4) the technology’ rate of improvements is fast, (5) most use cases for economists require using API+code (this will give you much more control on the output), (6) it is cheap to do so
The main uses for economists are: (1) cleaning data, (2) programming/making graphs, (3) spelling checks, (4) summarizing literature
Some best practices: (1) provide the model with as many examples as you can of what you need to be done, (2) do not use LLMs to do math, (3) use an “ensemble approach” to find the optimal prompt to ask the model