@DarioAmodei has a nice essay arguing AI could allow us to build "a country of geniuses in a data center", and that could lead to a century of progress in 5-10 years.

Some thoughts on why I think opinions on this differ and rarely converge. darioamodei.com/machines-of-lo… A peculiar phenomenon in debates about the likely impact of AI is that people can differ dramatically in their views, exchange all their information, and remain just as far apart as when they started.

https://x.com/mattsclancy/status/1768030659585409391

It’s the third anniversary of the “living literature review” on social science research about innovation I created!

I think more academics should consider writing living literature reviews and grant support is available. Thread!newthingsunderthesun.com “Living literature review” is the best name I’ve come up with for what I write. Key characteristics:
- collection of articles
- synthesize recent academic research
- written for non-specialists
- rigorous
- updatable
- one primary author
More details here. openphilanthropy.org/research/what-…

One thing from this interview I want to say more on, because I’ve never seen anyone make this point (apologies if I’ve just missed it). It’s about whether the Industrial Revolution is evidence that AI could lead to 10x faster economic growth.

https://x.com/robertwiblin/status/1794739244810055827

The argument is we should be open to 10x faster growth in the future because it’s happened before. UK GDP per capita growth accelerated from an average of 0.07%/yr over 1252-1652, to 1.13%/yr over 1800-2000 (all data from Angus Maddison in this thread). That’s a 17x speedup!

No single social science paper is decisive, but I really like this one. Seen in conjunction with other papers using a variety of methods, I think the case is strong that something troubling is afoot in science. Thread (references at the end). nature.com/articles/s4158… The paper’s main contribution is to show papers and patents have become less disruptive on average. Disruption is measured with citations: you’re disruptive if people who cite you don’t cite your own references (you rendered your forebears obsolete).

Was just talking to someone about whether they needed to go back to school to learn economics. What are your favorite online resources to teach yourself economics? A few of mine below:

https://twitter.com/BrianCAlbrecht/status/1602715987572408321

@coreeconteam has a fantastic free introductory textbook to economics: core-econ.org/the-economy/en/

One proposal for reforming scientific grant making is to look for proposals where reviewers disagree (some champions, some doubters), rather than consensus. I don't know of much evidence on this, but here's one datapoint from the appendix of this paper: direct.mit.edu/rest/article-a… This is ave citations of Econ papers submitted to top journals by sets of peer reviewer recommendation. Left image is raw data; right includes adjustments for journal and year of review. If disagreement is good, we should see off-diagonal papers do unusually well.

Long plane ride let me finally read @kaushikcbasu’s excellent Republic of Beliefs, after years of delay. Basic question: why do people follow laws? Easy answer: because if they don’t, they go to jail!

Not good enough!

Basu argues this is analogous to partial equilibrium thinking. We need to ask why do the police and judges etc enforce the law?

For @michael_nielsen a "short" explanation of why (TFP growth) = (real GDP growth) - 0.7*(labor growth) - 0.3(capital growth)

Composing as a thread rather than a set of replies, so I can go back and edit if needed.

https://twitter.com/michael_nielsen/status/1494674891399843842?s=20&t=NMeqqkcJpmXa42lvRviXvg

Assume we have a general aggregate production function linking GDP (measured in real output) at time t (denoted Y(t)) to capital and labor at time t via a function f(.) and TFP at time t (denoted A(t)):

Y(t) = A(t) f(K(t),L(t))

I want to riff a bit on this interesting post by @ArtirKel on tacit knowledge and his suggestion that there may be scope to disseminate tacit knowledge at scale via curated video libraries of people working on concrete examples. nintil.com/scaling-tacit-… I think you can think of tacit knowledge two ways:
- knowledge that can't jump from one head to another by being written down (or codified in another way)
- knowledge that can be transferred this way but isn't

A lot of tacit knowledge is the latter type.

Rather than fearing the rise of remote work, we should be promoting it. There are three main reasons.
1. To reduce geographic inequality
2. To reduce carbon emissions
3. To increase productivity growth 1. Geographic inequality: Agglomeration effects have boosted US growth for the past several decades, but at the cost of ever widening geographic inequality. This spills over into our current political predicament.

For all the profs thinking thinking about how to do asynchronous teaching online in the Fall, let me share a great experience I had with a new (free) platform called LearnItFast.io. Students loved it and did better than usual in my class (tho lots else changed). The basic idea is you create course pages with short flashcard-type questions mixed throughout your content. What's really cool is students create personal accounts and the questions they answer get added to a review stack that uses a spaced repetition algorithm.

Maybe 1/3 of the US labor force is working remotely now, due to covid-19. Will this persist? What do we know about the productivity of remote work, in normal times? A thread... Let's start with Bloom, Liang, Roberts, and Ying (2015). This study is remarkable because one of the coauthors is the co-founder of a large Chinese travel booking company. nbloom.people.stanford.edu/sites/g/files/…

With academic conferences all being cancelled, a few weeks ago a lot of people wondered why we need in-person conferences at all. Can't we just present over zoom?

One objection is that networking at these conferences matters. Does it? Here's a 🧵. #ThursdayThreads One place to start is to ask people where they met their academic coauthors. That's what Freeman, @inaganguli, and @ravivmg (2015) do (among other things). Among coauthors not living in the same city, about 15% met at conferences. nber.org/chapters/c13040

Are bigger populations more innovative? Let's look at a few papers emphasizing the distant past to see what they say. #ThursdayThreads There are a few reasons to think they would be:
- more brains ➡️ more ideas
- more people ➡️ more specialization (?)
- more consumers ➡️more profit for innovating

Remote collaboration in innovation is on the rise. The average distance between all the inventors listed on a patent has tripled between 1975 and 2015. #ThursdayThreads Of course, the number of inventors on a patent has also risen. But we see the same thing if we restrict attention to patents with just two inventors. More of these 2-inventor patents belong to distant collaborators.

In light of this recent article, here's a long thread about copyright (which I think is way too long). #ThursdayThreads nytimes.com/2020/01/06/art… The economic argument for copyright is that there is insufficient incentive for artists to create because art can be cheaply copied. Granting a creator monopoly rights over their creation makes it profitable (or maybe less unprofitable) to make art.

Where do new ideas and technologies come from? One school of thought says they are born from novel combinations of pre-existing ideas. It’s an appealing notion, but does this perspective get us anywhere useful? #ThursdayThreads Let’s look at a few papers.

One nice thing about the combinatorial perspective on innovation, is that if you know all the possible technological building blocks you can exhaustively enumerate all possible inventions, including ones that are never actually invented.

Happy Thanksgiving! Let's talk corn. In 2016, the NYTimes presenting the following figure to argue that the introduction of genetically engineered (GE) corn did not increase crop yields! Can this be true? Let's look at some research. #ThursdayThreads
nytimes.com/interactive/20… (Disclaimer: this thread isn't going to touch the controversy about the impact of GE crops on health or the environment. It's an important issue but I don't have anything interesting to say about it.)

We should nurture the rise of remote work. It’s one of the only economic forces countering rising regional inequality due to agglomeration effects.

An ongoing 🧵... My original blog post on this: matt-clancy.com/rehabilitating…

Here’s an obvious thought: more scientists, more discoveries; fewer scientists, fewer discoveries. But is that true? Let’s take a look at a few papers studying two (traumatic) upheavals in the scientific labor force. #ThursdayThreads #ProgressStudies. First; Nazi Germany's 1933/1940 dismissal of Jewish (and other) academics. Because the dismissals varied by university, this can be used as a (grim) experiment on the impact of rapidly reducing your scientific labor force.

True or false:
Science ==> Technological progress.
Here's a thread on recent research trying to answer that question. #ThursdayThreads #ProgressStudies Many would say answer is obviously "true." But some think technological breakthroughs mostly come from tinkering, and aren't guided by science. Others say ivory tower academics are disconnected from the real world and focus on irrelevant problems.