Jesse Perla
@jlperla.bsky.social
Associate Professor, UBC Vancouver School of Economics. Interested in the (macro)economics of growth/AI/ML, and use of ML within economics itself. Swiftie.
jesseperla.com
jesseperla.com
Oh we might have our answer to the breakdown, at list. Not sure about the ordering quite yet: bsky.app/profile/tyre...
I honestly think LLMs are very useful tools, but this kinda shit shows the real danger with these models:
Stupid people with important roles will use them for purposes they are not well suited for.
Stupid people with important roles will use them for purposes they are not well suited for.
April 3, 2025 at 3:32 PM
Oh we might have our answer to the breakdown, at list. Not sure about the ordering quite yet: bsky.app/profile/tyre...
I am willing to believe the reverse-engineering of the ridiculous numbers, but is there any sensible theory of the ordering of the list?
April 3, 2025 at 1:43 AM
I am willing to believe the reverse-engineering of the ridiculous numbers, but is there any sensible theory of the ordering of the list?
"ChatGPT, please generate a list of countries and islands along with a random number between 0 and 99%. Remove
Russia and ensure it is sorted randomly rather than going by country name or size"
Russia and ensure it is sorted randomly rather than going by country name or size"
April 2, 2025 at 10:06 PM
"ChatGPT, please generate a list of countries and islands along with a random number between 0 and 99%. Remove
Russia and ensure it is sorted randomly rather than going by country name or size"
Russia and ensure it is sorted randomly rather than going by country name or size"
Reverse engineering worries me. I love good data work, and I love applied theory - but trying to reverse-engineer a model from stylized facts is ill-posed. Which explanation to use? My sense is that insisting on data+theory+quantiative application has been a detrimental to publishing in econ
February 6, 2025 at 8:22 PM
Reverse engineering worries me. I love good data work, and I love applied theory - but trying to reverse-engineer a model from stylized facts is ill-posed. Which explanation to use? My sense is that insisting on data+theory+quantiative application has been a detrimental to publishing in econ
One concern is that much of the economics is learned by tinkering with simplifications/FOCs/edge-cases. These tools might make it easy to write down models, solve them numerically/symbolically/etc., but ultimately provide papers with fewer insights. At least for applied-theory papers.
February 6, 2025 at 5:11 PM
One concern is that much of the economics is learned by tinkering with simplifications/FOCs/edge-cases. These tools might make it easy to write down models, solve them numerically/symbolically/etc., but ultimately provide papers with fewer insights. At least for applied-theory papers.
A snippet of the 5 minute stream of conscious output before it answered my question.
January 29, 2025 at 8:56 PM
A snippet of the 5 minute stream of conscious output before it answered my question.