Norman Topping Chair in Medicine & Public Policy, USC, Sol Price School of Public Policy

Los Angeles, CA
Joined August 2014
A colleague said today "We really are medieval institution." Maybe we should lean into this.
57
This is why we need TWFE difference-in-differences analysis.
I recently learned that heavy use of the emdash is considered evidence that the writing is AI, or AI assisted -- yes, I'm late once again on trends. I've always been a heavy user. I bet the causal inference chapter in 8e of my intro econometrics book uses it nearly 100 times.
129
"It's like we shoot at the side of a barn and then draw a bullseye around the bullet hole." That right there kids is called "researcher degrees of freedom."
Replying to @paulnovosad
Academics are really good at convincing you that what they did is important, even when what they did was kind of arbitrary and inspired by what was feasible or readily at hand. It's like we shoot at the side of a barn and then draw a bullseye around the bullet hole.
1
157
Perception is filtered through discrimination. The weak orders commonly used in econ are, in practice, semi-orders--indifference is intransitive. "Shrinkflation" capitalizes on semiorders as do add ons at the point of sale, and price creep, and platform decay online.
TIL another econ Nobel laureate has taken an interest in psychology. Woodford, M. (2020). Modeling Imprecision in Perception, Valuation, and Choice. Annual Review of Economics.
2
127
Given Fama/Shiller, Myrdal/Hayek, and Schultz/Lewis the AI will just disagree with itself on major points so humans don't have to.
You’re deluding yourself if you think AI advances like in math aren’t coming to economics and other social sciences
96
Gary Becker paper "A rational number theory of the reals"
Economists love rationality so much they want to assume it for numbers
3
117
Jason Doctor retweeted
The most important briefings of the day. Karl Deisseroth, professor of bioengineering and of psychiatry and behavioral sciences, shares the news of his Nobel Prize with his kids.
530
4,658
935
36,684
4,629,562
[@ben_golub ]: Are eigenvalues involved? [narrator]: Eigenvalues are the primary tool used to decide how many components to keep for the final regression stage. [@ben_golub ]: I like it!
What's the best way to solve linear regression? (No, it’s not least squares.) In our new paper, we show that principal component regression (PCR) beats, up to constants, every monotone spectral filter (incl. gradient descent & ridge regression) on every problem instance! (1/8)
3
282
I sometimes think about Vonnegut's "Ice-9," now a cautionary tale about madness and instability in a world where intelligence could run wild. Maybe like water in that book, intelligence is "metastable"; a small advance keeps the world the same(2), but a large one destroys it (3).
I think we really need to reject this argument that it’s fine if we at Anthropic do what would otherwise be morally wrong because it’s important that the morally superior people do it and we are morally superior.
2
122
Writing can involve descriptions of introspection, feelings and states of being that are not relevant to conversation or debate. Therefore, writing is essential to thinking.
1
90
Jason Doctor retweeted
We have thousands of years of religious, philosophical, and legal traditions trying to “align” humans and make our behavior “safe.” Now we’re trying to speed-run toward those goals with AI. As we have often failed with humans, we should expect and plan for failures with AI.
3
5
1
16
2,191
Christopher Guest but instead of naming nuts he is naming leadership roles at journals: "Editor, Deputy Editor, Associate Editor, Assistant Editor, Guest Editor..."
1
99
If you’re young, one of the main ways you can add value at work is by taking advantage of a lack of family responsibilities to just take annoying shit on and be helpful. slowboring.com/p/feed-the-li…
22
34
24
976
78,459
This is a good example of the psychology of risk perception. Two factors predict high-risk judgments: "dreadfulness" and "uncontrollability." Plane crashes are dreadful and uncontrollable, but air travel is safe. You don't control the Waymo; driving seems safer, but it is not.
Bayesians of the world unite! How does this evidence impact your beliefs about the robots? As the value of a Statistical life rises faster than GNP grows, why aren't more people celebrating Waymo's achievements? I thought we all want "safe assets"?
108
No more algebra in school just sigma-algebras.
We should replace geometry with real analysis.
212
I'd say it is relevant :P(patient alive X yrs) uses measure theory. Time of death has infinite points. So we must break it up into a countably infinite disjoint point sets. Lebesgue measure has no countably additive extension to all subsets of the reals. Hence, measure theory.
first quarter first year stats PhD measure theory was fun I guess, but one of the most useless technical classes I've taken in my life. I don't remember one thing from it that helps me use probability in my job today
1
278
I kind of agree . Singular points have empirical meaning fixed points are a trick to prove a theorem. Every singular point is a fixed point of every automorphism. But, not every fixed point is a singular point. In prospect theory for example, the status quo is a singular point.
we should nuke fixed point theorems from the 1st year econ curriculum. in the last 1000 econ papers I've read, and maybe the last 200 theory papers, I have seen 0 fixed point theorems
1
1
450
A fixed point theorem proves that one map sends at least one point to itself. This is not anything to do with the theory, just a math fact that can help get a proof to the finish line.
1
1
42
I am not against fixed point theorems but there are not big E or P (economics or psychology) they are big M (mathematics) little "e" or "p" (economics or psychology). And it is the social science that we are theorizing about.
1
22