Some doubts about Chinese industrial policy

by  Tyler Cowen in

Each year governments worldwide spend an enormous amount of money subsidising businesses. This column investigates the relationship between the allocation of government subsidies and total productivity for Chinese listed firms. The authors find little evidence that the Chinese government consistently ‘picks winners’. Firms’ ex-ante productivity is negatively correlated with subsidies received by firms, and subsidies appear to have a negative impact on firms’ ex-post productivity growth.

And this:

Our sample includes all firms listed on the Shanghai and Shenzhen stock exchanges from 2007 to 2018 except financial services firms, for which computation of productivity presents a number of challenges.

That is from a new study by Lee Branstetter, Guangwei Li, and Mengjia Ren.  Such estimations are typically fraught with questionable assumptions, but at the very least the results do not come up positive.

EA, AI, and the rationality community

by  Tyler Cowen

More broadly, I think AI Alignment ideas/the EA community/the rationality community played a pretty substantial role in the founding of the three leading AGI labs (Deepmind, OpenAI, Anthropic), and man, I sure would feel better about a world where none of these would exist, though I also feel quite uncertain here. But it does sure feel like we had a quite large counterfactual effect on AI timelines.

That is from “habryka, Ben Pace” on LessWrong blog.  As you might expect, I would give those comments a different valence, nonetheless they are insightful.  Here are my points:

1. It is truly remarkable how much influence the cited movements have had.  Whether or not you agree in full (or at all), this should be recognized and respected.  Kudos to them!  And remember, so often ideas lie behind technology.

2. Anthropic has announced a raise of $5 billion and is promoting its intention to compete with Open AI and indeed outdo them.  The concept “Solve for the equilibrium” should rise in status.

3. You cannot separate “interest in funding AI safety” (which I am all for) from “AI progress.”  That by now should be obvious.  No progress, no real interest in safety issues.

4. To this day, the Doomsters are de facto the greatest accelerationists.  Have you noticed how the Democrats (or Republicans) “own” certain political issues?  For instance, voters trust the Democrats more with Social Security, and the mere mention of the topic helps them, even if a Republican has a good point to make.  Well, the national security establishment “owns” the ideas of existential risk and risk from foreign powers.  The more you talk about doomsday issues, the more AI risk gets slotted into their purview, for better or worse.  And they ain’t Brussels (thank goodness).  To the extent the Doomsters have impact, their net effect will be to place the national security types in charge, or at least to raise their influence.  And how do they think that is going to work out (on their own terms)?  Perhaps they would do better to focus on mundane copyright and libel issues with LLMs, but that is not their nature.

Saturday assorted links

by  Tyler Cowen

1. Mary Lou Williams appreciation (NYT).

2. Is D.C. becoming a bagel city?

3. Arthur Allshire on whether LLMs will be an anti-censorship tool.

4. Filipinos nailed to crosses on Good Friday despite objection by Catholic Church.

5. The economy in Zambia.

6. “Pretty incredible that the Indian government has released a version of @OpenAI Whisper fine tuned on a Hindi data set.”  And AI and religion is a match made in heaven (NYT).  Really.  “The flood of papal imagery has been so voluminous that some people in online generative A.I. forums have begged for creators to use another inspiration.”  I really like this one.

Lessons from the Baruch Plan for Nuclear Weapons

by  Tyler Cowen in

The invention of atomic energy posed a novel global challenge: could the technology be controlled to avoid destructive uses and an existentially dangerous arms race while permitting the broad sharing of its benefits? From 1944 onwards, scientists, policymakers, and other technical specialists began to confront this challenge and explored policy options for dealing with the impact of nuclear technology. We focus on the years 1944 to 1951 and review this period for lessons for the governance of powerful technologies, and find the following: Radical schemes for international control can get broad support when confronted by existentially dangerous technologies, but this support can be tenuous and cynical. Secrecy is likely to play an important, and perhaps harmful, role. The public sphere may be an important source of influence, both in general and in particular in favor of cooperation, but also one that is manipulable and poorly informed. Technical experts may play a critical role, but need to be politically savvy. Overall, policymaking may look more like “muddling through” than clear-eyed grand strategy. Cooperation may be risky, and there may be many obstacles to success.

That is by Waqar Zaidi and Allan Dafoe, at the Centre for Governance of AI, exactly the kind of work people should be doing.

Matt Yglesias on movies vs. TV

by  Tyler Cowen in

But I’ve gotten really disgruntled with the “prestige TV” landscape and am trying to redirect my content consumption accordingly. One thing that makes movies really great in my view is that before they shoot a movie, they write a screenplay and the screenplay has an end. Both the screenwriter and other people have read that screenplay all the way from beginning to end and they’ve tweaked and changed it and gotten it into a position where they are ready to start production. Then after a movie is filmed, the editor and director work with the footage and come up with a complete movie that has a beginning, a middle, and an end. They then ship the movie out, and it’s screened by critics who watch the entire movie before writing their review.

This does not guarantee that every movie that comes out is good. But it does guarantee that if someone tells you “‘The Menu’ is good,” they are evaluating a completed product…

By contrast, TV shows have this quasi-improvisational quality where the showrunners are constantly needing to come up with new balls to toss into the air. In old-fashioned non-prestigious “adventure of the week”-type shows, this actually works fine because the writers are not building up tension or setting unexplored plots in motion. But as serialized TV storytelling has gotten more and more common, we’re more and more often asked to show patience through early episodes or to try to find things intriguing with no ability to know whether any of it will pay off. Creators often have no idea where they’re going with the story.

Back in HBO’s heyday, the tradeoff was that The Sopranos and The Wire got to paint on a giant canvas and tell stories that are just too capacious for the movie format. But eventually networks got tired of spending that kind of money and cut back the sizes of the casts to something more normal for television.

That is from his Friday mailbag ($).  The bottom line is that, like Matt, you should watch more movies and less TV.

Did Ottoman Sultans ban print?

by  Tyler Cowen in

Did printing transform the Ottoman Empire? And what took the Ottomans so long to print? Much of the scholarship surrounding the topic of Ottoman printing, or the occurrence of printing within the Ottoman Empire (1453–1922), is structured around these two related frameworks. In this essay, I argue that these frameworks are ahistorical because they predicate Ottoman printing on the European experience of print. To support this point, I examine the disproportionate role played by certain early modern European accounts of Ottoman printing within Western and Arabic historiography. In particular, I examine the life cycle of scholars’ belief that Ottoman sultans banned printing, which I contrast with extant documentation for the imperial Porte’s stance on printing. I argue that the sources available to scholars today do not support the notion that the sultans banned printing. Rather, they demonstrate that this claim arose from early modern European scholars’ search to articulate their sense of Ottoman inadequacy through explanations for why Ottomans did not print. The history of this particular line of inquiry is significant, I argue, because many scholars continue to probe the issue of why Ottomans did not print. In so doing, they maintain the expectation that print would revolutionize society, even though they have begun questioning the existence of the ban.

That is from Kathryn A. Schwartz, in Print History (jstor).  Via Benedikt A.

Friday assorted links

by  Tyler Cowen

1. Gloria the young sloth.

2. ” Our theory sheds light on the mechanism of disastrous individual decisions in highly institutionalized authoritarian regimes.”  With reference to Putin.

3. White House issues update to cost-benefit guidelines.  And an EA analysis.  And is the best kimchi made in earthenware pots?

4. The gambler who beat roulette (Bloomberg).

5. “Inflation is conflict.”  New paper, not exactly my view, but it is not as if previous theories have been entirely successful.

6. Peter Sellars version of St. Matthew’s Passion, with Berlin, is now ungated for Easter.

The Nuclear Non-proliferation Treaty and existential AGI risk

by  Tyler Cowen in

The Nuclear Non-Proliferation Treaty, activated in 1970, has been relatively successful in limiting nuclear proliferation.  When it comes to nuclear weapons, it is hard to find good news, but the treaty has acted as one deterrent of many to nation-states acquiring nuclear arms.  Of course the treaty works, in large part, because the United States (working with allies) has lots of nuclear weapons, a powerful non-nuclear military, de facto control of SWIFT, and so on.  We strongly encourage nations not to go acquiring nuclear weapons — just look at the current sanctions on Iran, noting the policy does not always succeed.

One approach to AI risk is to treat it like nuclear weapons and also their delivery systems.  Let the United States get a lead, and then hope the U.S. can (in conjunction with others) enforce “OK enough” norms on the rest of the world.

Another approach to AI risk is to try to enforce a collusive agreement amongst all nations not to proceed with AI development, at least along certain dimensions, or perhaps altogether.

The first of these two options seems obviously better to me.  But I am not here to argue that point, at least not today.  Conditional on accepting the superiority of the first approach, all the arguments for AI safety are arguments for AI continuationism.  (And no, this doesn’t mean building a nuclear submarine without securing the hatch doors.)  At least for the United States.  In fact I do support a six-month AI pause — for China.  Yemen too.

It is a common mode of presentation in AGI circles to present wordy, swirling tomes of multiple concerns about AI risk.  If some outside party cannot sufficiently assuage all of those concerns, the writer is left with the intuition that so much is at stake, indeed the very survival of the world, and so we need to “play it safe,” and thus they are lead to measures such as AI pauses and moratoriums.

But that is a non sequitur.  The stronger the safety concerns, the stronger the arguments for the “America First” approach.  Because that is the better way of managing the risk.  Or if somehow you think it is not, that is the main argument you must make and persuade us of.

(Scott Alexander has a new post “Most technologies aren’t races,” but he doesn’t either choose one of the two approaches listed above, nor does he outline a third alternative.  Fine if you don’t want to call them “races,” you still have to choose.  As a side point, once you consider delivery systems, nuclear weapons are less of a yes/no thing than he suggests.  And this postulated take is a view that nobody holds, nor did we practice it with nuclear weapons: “But also, we can’t worry about alignment, because that would be an unacceptable delay when we need to “win” the AI “race”.”  On the terminology, Rohit is on target.  Furthermore, good points from Erusian.  And this claim of Scott’s shows how far apart we are in how we consider institutional and also physical and experimental constraints: “In a fast takeoff, it could be that you go to sleep with China six months ahead of the US, and wake up the next morning with China having fusion, nanotech, and starships.”)

Addendum:

As a side note, if the real issue in the safety debate is “America First” vs. “collusive international agreement to halt development,” who are the actual experts?  It is not in general “the AI experts,” rather it is people with experience in and study of:

1. Game theory and collective action

2. International agreements and international relations

3. National security issues and understanding of how government works

4. History, and so on.

There is a striking tendency, amongst AI experts, EA types, AGI writers, and “rationalists” to think they are the experts in this debate.  But they are only on some issues, and many of those issues (“new technologies can be quite risky”) are not so contested. And because these individuals do not frame the problem properly, they are doing relatively little to consult what the actual “all things considered” experts think.

*A New History of Greek Mathematics*

by  Tyler Cowen in

I have read only about 30 pp. so far, but this is clearly one of the best science books I have read, ever.  It is clear, always to the point, conceptual, connects advances in math to the broader history, explains the math, and full of interesting detail.  By Reviel Netz.  Here is a brief excerpt:

And this is how mathematics first emerges in the historical record: the simple, clever games accompanying the education of bureaucrats.

One of the best books of the year, highly recommended.

Thursday assorted links

by  Tyler Cowen

1. “We find that using algorithmic responses changes language and social relationships. More specifically, it increases communication speed, use of positive emotional language, and conversation partners evaluate each other as closer and more cooperative. However, consistent with common assumptions about the adverse effects of AI, people are evaluated more negatively if they are suspected to be using algorithmic responses.”  Link here.

2. Living with non-alignment.  And a very sane take on AI risk.  Very good thread.  And should American VCs be funding Chinese AI?  And Leopold Aschenbrenner responds to me on AI risk, very good piece.

3. Why isn’t Europe doing worse?

4. “I hereby challenge professors from universities around the world to submit assignments that they believe are AI-immune.”

5. Which individuals are most likely to believe that AI is likely to destroy society?