Thursday assorted links

by  Tyler Cowen

1. ChatGPT taking U. Minnesota law exams.

2. The story and background of Scholz.

3. What is going on with PredictIt, and the legal case against them.

4. Macca reviews from 45s from 1967.  Lennon does the same from 1965.  Both are a bit harsh, but right on, and ruthless with those who will not innovate and progress with their styles.

5. Why is East Asia less happy?

6. Construction Physics on Goolsbee and Syverson on construction productivity.  And commentary on some recent gender gap results.

7. Overview of the OpenAi story.

The Great (British) Stagnation

by  Alex Tabarrok in

David Wallace-Wells in the NYTimes:

In December, as many as 500 patients per week were dying in Britain because of E.R. waits, according to the Royal College of Emergency Medicine, a figure rivaling (and perhaps surpassing) the death toll from Covid-19. On average, English ambulances were taking an hour and a half to respond to stroke and heart-attack calls, compared with a target time of 18 minutes; nationwide, 10 times as many patients spent more than four hours waiting in emergency rooms as did in 2011. The waiting list for scheduled treatments recently passed seven million — more than 10 percent of the country — prompting nurses to strike. The National Health Service has been in crisis for years, but over the holidays, as wait times spiked, the crisis moved to the very center of a narrative of national decline.

It’s not just the NHS

By the end of next year, the average British family will be less well off than the average Slovenian one, according to a recent analysis by John Burn-Murdoch at The Financial Times; by the end of this decade, the average British family will have a lower standard of living than the average Polish one.

Wallace-Wells puts the blame on “austerity”. I see austerity as more obviously a consequence than a cause of stagnation. Government spending in Poland and Slovenia is modestly less than in the UK and the central government in Poland and Slovenia spend far less than the UK does on health. The question is not why the UK spends less–it doesn’t–the question is why it spends so much and gets so little.

See also Tyler and Tyrone on this issue.

My excellent Conversation with Paul Salopek

by  Tyler Cowen in

Here is the transcript and audio, here is the summary:

Paul Salopek is a Pulitzer Prize-winning journalist and National Geographic fellow who, at the age of 50, set out on foot to retrace the steps of the first human migrations out of Africa. The project, dubbed the “Out of Eden Walk,” began in Ethiopia in 2012 and will eventually take him to Tierra Del Fuego, a distance of some 24,000 miles.

Calling in just as he was about to arrive in Xi’an, he and Tyler discussed his very localized supply chain, why women make for better walking partners, the key to crossing deserts, the most difficult terrain to traverse, what he does for exercise, his information prep for each new region, how he’s kept the project funded, why India is such a good for walkers, which cuisines he’s found most and least palatable, what he learned working the crime beat in Roswell, New Mexico, how this project challenges conventional journalism, his thoughts on the changing understanding of early human migration, and more.

Here is one excerpt:

COWEN: What’s true is true. How is it that you crossed the desert? You’ve been through some of the Gulf States, I think.

SALOPEK: Yes, I’ve been through several deserts. The first was the Afar Desert in north Ethiopia, one of the hottest deserts in the world, and then the Hejaz in western Saudi Arabia, and then some big deserts in Central Asia, the Kyzyl Kum in Uzbekistan.

You cross deserts with a great attentiveness. You seem to want to speed up to get through them as quickly as possible, but often, they require slowing down, and that seems counterintuitive. You have to walk when the temperatures are congenial to your survival. Sometimes that means walking at night as opposed to the day. It means maybe not covering the distances that you would in more moderate climates.

Deserts are like a prickly friend. You approach them with care, but if you invest the time, they’re pretty inspiring and remarkable. There are reasons why old hermits go out into the deserts to seek visions. I was born in a desert. I was born in the Mojave Desert of Southern California, so I’m partial to them, maybe even by birth.

COWEN: Do you find deserts to be the most difficult terrain to cross?

SALOPEK: No, I find alpine mountains to be far trickier. Deserts can be fickle. Deserts can kill you if you’re not careful. Of course, water is the most limiting factor for survival.

But alpine mountain weather is so unpredictable, and a very sunny afternoon can turn into a very stormy late afternoon in a very quick time period. Threats like rock falls, like avalanches, blizzards — those, for me, are far more difficult to navigate than deserts. Also, I guess having been born in the subtropics, I don’t weather the cold as well, so there’s that bias thrown in.

COWEN: What do you do for exercise?

Recommended, interesting throughout.

National Average is Over

by  Tyler Cowen in

This paper considers the implications for developing countries of a new wave of technological change that substitutes pervasively for labor. It makes simple and plausible assumptions: the AI revolution can be modeled as an increase in productivity of a distinct type of capital that substitutes closely with labor; and the only fundamental difference between the advanced and developing country is the level of TFP. This set-up is minimalist, but the resulting conclusions are powerful: improvements in the productivity of “robots” drive divergence, as advanced countries differentially benefit from their initially higher robot intensity, driven by their endogenously higher wages and stock of complementary traditional capital. In addition, capital—if internationally mobile—is pulled “uphill”, resulting in a transitional GDP decline in the developing country. In an extended model where robots substitute only for unskilled labor, the terms of trade, and hence GDP, may decline permanently for the country relatively well-endowed in unskilled labor.

That is from a 2020 IMF working paper by Cristian Alonso, Andrew Berg, Siddharth Kothari, Chris Papageorgiou, and Sidra Rehman.  Via Eric Yu.

Who is locally influential these days?

by  Tyler Cowen in

Um…um…uh-oh:

Who do people think are influential in their own community? This question is important for understanding topics such as social networks, political party networks, civic engagement, and local politics. At the same time as research on these topics has grown, measurement of public perceptions of local influence has dried up. Years ago, researchers took active interest in the question of community influence. They found that most ordinary Americans could identify a person who they thought had influence in their community. Respondents usually named business leaders. Where does the public stand today? In three different ways, we ask respondents who has local influence. The vast majority of respondents today cannot think of anyone. Those who do identify someone as influential rarely choose a businessperson. This article aims to reintroduce the public opinion of community influence and situate findings in related scholarship.

Here is the new article by Joshua Hochberg and Eitan Hersh.  David Brooks, telephone!  Don’t even ask how the “religious leaders” fare in the polling…

AI Is Improving Faster Than Most Humans Realize

by  Tyler Cowen in

That is the topic of my latest Bloomberg column, here is one excerpt:

I have a story for you, about chess and a neural net project called AlphaZero at DeepMind. AlphaZero was set up in late 2017. Almost immediately, it began training by playing hundreds of millions of games of chess against itself. After about four hours, it was the best chess-playing entity that ever had been created. The lesson of this story: Under the right conditions, AI can improve very, very quickly.

LLMs cannot match that pace, as they are dealing with more open and more complex systems, and they also require ongoing corporate investment. Still, the recent advances have been impressive.

I was not wowed by GPT-2, an LLM from 2019. I was intrigued by GPT-3 (2020) and am very impressed by ChatGPT, which is sometimes labeled GPT-3.5 and was released late last year. GPT-4 is on its way, possibly in the first half of this year. In only a few years, these models have gone from being curiosities to being integral to the work routines of many people I know. This semester I’ll be teaching my students how to write a paper using LLMs.

We are now at or close to the point where LLMs can read and accurately evaluate the work of…LLms.  That will accelerate progress considerably.

And to close I wrote this:

I’ve started dividing the people I know into three camps: those who are not yet aware of LLMs; those who complain about their current LLMs; and those who have some inkling of the startling future before us. The intriguing thing about LLMs is that they do not follow smooth, continuous rules of development. Rather they are like a larva due to sprout into a butterfly.

It is only human, if I may use that word, to be anxious about this future. But we should also be ready for it.

Recommended.  Remember my old Wilson Quarterly piece about “invisible competition”?

On censorship of LLM models, from the comments

by  Tyler Cowen in

IMO, censorship is a harder task than you think.

It’s quite hard to restrict the output of general purpose, generative, black box algorithms. With a search engine, the full output is known (the set of all pages that have been crawled), so it’s fairly easy to be confident that you have fully censored a topic.

LLMs have an effectively unbounded output space. They can produce output that is surprising even to their creators.

Censoring via limiting the training data is hard because algorithms could synthesize an “offensive” output by combining multiple outputs that are ok on their own.

Adding an extra filter layer to censor is hard as well Look at all the trouble chatGPT has had with this. Users have repeatedly found ways around the dumb limitations on certain topics.

Also, China censors in an agile fashion. A topic that was fine yesterday will suddenly disappear if there was a controversy about it. That’s going to be hard to achieve given the nature of these algorithms.

That is from dan1111.  To the extent that is true, the West is sitting on a huge propaganda and communications victory over China.  This is not being discussed enough.

Gender and tone in recorded economics presentations

by  Tyler Cowen in

You’re going to see a lot more research papers like this one:

This paper develops a replicable and scalable method for analyzing tone in economics seminars to study the relationship between speaker gender, age, and tone in both static and dynamic settings. We train a deep convolutional neural network on public audio data from the computer science literature to impute labels for gender, age, and multiple tones, like happy, neutral, angry, and fearful. We apply our trained algorithm to a topically representative sample of presentations from the 2022 NBER Summer Institute. Overall, our results highlight systematic differences in presentation dynamics by gender, field, and format. We find that female economists are more likely to speak in a positive tone and are less likely to be spoken to in a positive tone, even by other women. We find that male economists are significantly more likely to sound angry or stern compared to female economists. Despite finding that female and male presenters receive a similar number of interruptions and questions, we find slightly longer interruptions for female presenters. Our trained algorithm can be applied to other economics presentation recordings for continued analysis of seminar dynamics.

Some people might just stop going to recorded conferences, of course.  That paper is by Amy Handlan and Haoyu Sheng, via the excellent Kevin Lewis.

Ethnic Remoteness Reduces the Peace Dividend from Trade Access

by  Tyler Cowen in

This paper shows that ethnically remote locations do not reap the full peace dividend from increased market access. Exploiting the staggered implementation of the US-initiated Africa Growth and Opportunity Act (AGOA) and using high-resolution data on ethnic composition and violent conflict for sub-Saharan Africa, our analysis finds that in the wake of improved trade access conflict declines less in locations that are ethnically remote from the rest of the country. We hypothesize that ethnic remoteness acts as a barrier that hampers participation in the global economy. Consistent with this hypothesis, satellite-based luminosity data show that the income gains from improved trade access are smaller in ethnically remote locations, and survey data indicate that ethnically more distant individuals do not benefit from the same positive income shocks when exposed to increased market access. These results underscore the importance of ethnic barriers when analyzing which locations and groups might be left behind by globalization.

That is from a new NBER working paper by Klaus Desmet and Joseph F. Gomes.