Ryan Murphy

20.9K posts

Ryan Murphy banner
Ryan Murphy

Ryan Murphy

@ryanhmurphy

Research Associate Professor @SMUCoxBridwell. Co-author of the Economic Freedom of the World index and the Human Freedom Index.

Dallas Katılım Ocak 2010
129 Takip Edilen979 Takipçiler
𝔐𝔽𝓩
𝔐𝔽𝓩@mean_field_zane·
@ryanhmurphy @PhilWMagness Compared to the models of the time, they don’t seem very formal to me! They seem pretty simple… do you find it surprising or unsurprising that the calculus of consent framework hasn’t really evolved into a full literature in applied theory?
English
1
0
0
12
𝔐𝔽𝓩
𝔐𝔽𝓩@mean_field_zane·
I mean, it’s entirely silly to refer to one of the most foundational economic theorists of all time as “niche”. His work is the foundation for literally every single model in modern macroeconomics and also the study of adverse selection & moral hazard.
English
2
4
62
5.1K
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@mean_field_zane @PhilWMagness this sub-question started with how many models public choices theorists developed. So are you ignorant, crazy, or a bot? They clearly developed as many models as was needed from the period.
English
1
0
0
12
𝔐𝔽𝓩
𝔐𝔽𝓩@mean_field_zane·
@ryanhmurphy @PhilWMagness I mean, they spent too little time on anything but price theory, which is a great subject but is inherently limited for this application. So it’s unsurprising people don’t really use their models.
English
1
0
0
16
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@mean_field_zane @PhilWMagness If you want to say, "we have superior theoretical models of political economics now," that's fine. If you want to say that Buchanan or Tullock spent too little time on theory, that's crazy
English
1
0
0
15
𝔐𝔽𝓩
𝔐𝔽𝓩@mean_field_zane·
Sure, but it hasn’t kept up, and just as with Friedman (who I am a megafan of dispute his mediocre formal modeling ability), they ended up losing control of the field’s direction bc they chose to go into politics (or were bad at making friends like Tullock). This means nobody really picked up working on the theory, unlike with market power stuff.
English
1
0
0
13
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@mean_field_zane @PhilWMagness The work of both Buchanan and Tullock was built on formal models (of their time)... that was their methodology. That was their thing. They were fighting Samuelson on his own terms. This wasn't Friedman Chicago.
English
1
0
0
18
𝔐𝔽𝓩
𝔐𝔽𝓩@mean_field_zane·
@ryanhmurphy @PhilWMagness Exactly. I like public choice but because they didn’t contribute too many formal models economists don’t work nearly as much on it as they should.
English
1
0
0
17
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@PhilWMagness @mean_field_zane ...that's the nexus of disagreement here? If I take the question substantively in terms of "do economists think of these things when doing research" then yes I would agree with this tweet, stylistically. it also isn't the nexus of disagreement and weird for you to say otherwise
English
1
0
1
17
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@PhilWMagness @mean_field_zane You raised the "bullshit citations" thing in response to Stiglitz, right? Why is this on the defensive? I always wanted to just push everything onto a Jamesian positivist scale, if anything, to answer the question.
English
1
0
1
17
Phil Magness
Phil Magness@PhilWMagness·
@ryanhmurphy @mean_field_zane But no answer to the question of where those citations are coming from, which was the claim about "quality," no? If the question was "Is Stiglitz cited in cumulative more than Buchanan?" then that could have been ascertained by a simple google scholar search.
English
1
0
0
23
Phil Magness
Phil Magness@PhilWMagness·
I'm fine if you want to limit it to just econ. I was thinking econ & poli sci, since Buchanan's work had strong influence in poli sci. But however you wanted to parse it, I suspect that Buchanan has very few citations in humanities journals compared to Stiglitz, whose popular work gets picked up in those journals out of political agreement
English
1
0
0
34
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@PhilWMagness @mean_field_zane this seems like weird question about what defines "adjacent fields." Stiglitz, I'm guessing, would win most definitions of "adjacent fields."
English
1
0
1
22
Phil Magness
Phil Magness@PhilWMagness·
@ryanhmurphy @mean_field_zane Suppose we were able to parse the aggregate citations of each into (a) econ and closely adjacent fields and (b) everything else. Who do you think would have the higher "everything else" share?
English
1
0
0
19
Phil Magness
Phil Magness@PhilWMagness·
@ryanhmurphy @mean_field_zane The book has 25K citations. The paper with Dixit, which is often said to be his most important paper, has 15K. And Stiglitz has several other highly cited popular press books with lots of low-quality citations. Buchanan did not write popular press books
English
1
0
1
117
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@PhilWMagness @mean_field_zane Stiglitz doesn’t lose much by losing that, right? I was just raising it because they both represent “low quality citations”.
English
1
0
0
104
Phil Magness
Phil Magness@PhilWMagness·
@ryanhmurphy @mean_field_zane Because I'm not disputing that Stiglitz is cited. The other guy is, by contrast, disputing that Buchanan is cited, and further claiming without any evidence that citations of Calculus of Consent are from low quality sources
English
1
0
2
109
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@PhilWMagness @mean_field_zane The Use of Knowledge in Society is number one for Hayek, but his next few books were accessible to other readers and undoubtedly cited by non-economists. Why not focus on a metric that works here?
English
2
0
1
121
Phil Magness
Phil Magness@PhilWMagness·
@mean_field_zane @ryanhmurphy Just going to note that this has 10K more citations that Dixit-Stiglitz 77, probably making it Stiglitz's most "influential" work. I'll let you speculate about the quality of citations it has accrued.
Phil Magness tweet media
English
1
0
2
112
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@PhilWMagness @mean_field_zane I’m afk, but that’s one paper? I would prefer to do a bill James style aggregate to answer the question, but if we set aside finance and econometrics like, it’s not a question here
English
2
0
0
114
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@PhilWMagness @mean_field_zane So the work of James Buchanan was super niche and can be more casually dismissed than how you dismissed the work of Stiglitz. Way more niche.
English
2
0
2
167
Ryan Murphy
Ryan Murphy@ryanhmurphy·
@mean_field_zane @PhilWMagness I also wouldn't even frame it necessarily in terms of macro. Not crazy to say that Stiglitz is the most important public finance economist since the second world war! His arguments may be shit, but that's a different question.
English
2
0
1
132
Phil Magness
Phil Magness@PhilWMagness·
@mean_field_zane By "foundational" what you really mean is he wrote some important papers in macro theory that are very influential among people who specialize in the same... ...which is to say niche, particularly since not everyone values macro the same way you do.
English
2
0
9
564
Ryan Murphy
Ryan Murphy@ryanhmurphy·
The model also clearly interpolates and extrapolates into the space of text that has not yet previously existed. If you haven't seen it, you are using very outdated models or haven't honestly given one a go.
English
0
0
0
26
Ryan Murphy
Ryan Murphy@ryanhmurphy·
There is also a "temperature" variable that makes hewing to the highest probability or more unlikely tokens to be chosen. The context (including, well, what you tell it you want) will adjust the temperature variable.
English
1
0
0
29
Ryan Murphy
Ryan Murphy@ryanhmurphy·
This is false. The algorithm starts by assigning probabilities to the next token, but it doesn't follow the rule of "always take the highest probability token" for the very reason that it flattens its output. Somewhat basic stuff, no?
Daniel Friedman@DanFriedman81

AI writing is produced by reading a tremendous volume of human writing and then using a statistical algorithm to produce new text that is as close as possible to the exact average of everything in its training set. In other words, when you tell it to write something that sounds literary, it produces figurative language that resembles good writing in superficial ways but lacks intent, meaning and purpose, the things AI is incapable of. That is how you get passages like: “If the village had a mouth, it was hers.” At first read, it sounds like strong characterization containing a witty observation, because it is mimicking texts with those qualities, but it actually has no content. It says nothing about Marsha. Does it mean she is a leader who speaks for the village? Does it mean that she is a gossip? Does it mean she eats everything? “Big in the way of women who never apologise to furniture.” Why would anybody apologize to furniture? Once again, this is an algorithmic system fundamentally incapable of understanding humor trying to replicate it. In the training set, it is statistically likely that writers will humorously describe a fat person by the way their weight affects furniture, and this is the AI’s attempt at approximating such a description. “She knew the ways of men hollowed by want until only one thing remained.” What? Who are these men? What is the thing that remains after they are hollowed by want? How does she know the ways of these men? This is just an approximation of literary language that is entirely empty of meaning or purpose. And then we get to: “People talk about bush like it dumb. But bush keeps memory the way hair keeps scent.” This, like everything else here is meaningless. But it sounds really ethnic and exotic, and that veneer of exoticism is what short-circuited whatever mechanism the white liberal judges of the Commonwealth Prize had for perceiving slop.

English
1
0
0
94
Ryan Murphy
Ryan Murphy@ryanhmurphy·
Still need to cite it!
English
0
0
0
22
Ryan Murphy
Ryan Murphy@ryanhmurphy·
The End of History and the Last Man by Fukuyama isn't what anyone thinks it is, a sober if optimistic view of the prospects for liberalism as communism sunsets. Instead it is 400 pages of Hegelian mysticism that doesn't really mean anything.
English
1
1
0
52