*Streets of Gold*

In one of our research  projects, we followed pairs of brothers born in Norway, one of whom left for the United States by 1900 while the other remained behind…Brothers who immigrated to the United States earned nearly twice as much as their siblings back home.

That is from the new and excellent Streets of Gold: America’s Untold Story of Immigrant Success, by Ran Abramitzky and Leah Boustan.

The incidence of rent control

We use the price effects caused by the passage of rent control in St. Paul, Minnesota in 2021, to study the transfer of wealth across income groups. First, we find that rent control caused property values to fall by 6-7%, for an aggregate loss of $1.6 billion. Both owner-occupied and rental properties lost value, but the losses were larger for rental properties, and in neighborhoods with a higher concentration of rentals. Second, leveraging administrative parcel-level data, we find that the tenants who gained the most from rent control had higher incomes and were more likely to be white, while the owners who lost the most had lower incomes and were more likely to be minorities. For properties with high-income owners and low-income tenants, the transfer of wealth was close to zero. Thus, to the extent that rent control is intended to transfer wealth from high-income to low-income households, the realized impact of the law was the opposite of its intention.

That is from a new paper by Kenneth R. Ahern and Marco Giacoletti, via the economics-understanding Kevin Lewis.

Monday assorted links

1. Fertility in various mega-cities.

2. Those old service sector jobs: “Garden hermits or ornamental hermits were hermits encouraged to live in purpose-built hermitagesfolliesgrottoes, or rockeries on the estates of wealthy landowners, primarily during the 18th century. Such hermits would be encouraged to dress like druids and remain permanently on site, where they could be fed, cared for, and consulted for advice, or viewed for entertainment.”  VTESE.

3. Second order effects of the rise of large language models.  Interesting thread.  Will there be so much stratification?

4. The ideas market likes “verbal authenticity.”

5. “Estimates of the number of additional substance abusers during the pandemic presented here suggest that increased substance abuse accounts for between 9 and 26 percent of the decline in prime-age labor-force participation between February 2020 and June 2021.”  Link here.

6. Four-year degree becoming less important as a job requirement (NYT).

7. Algospeak.

When Can/Should We Pull the Plug?

At Less Wrong a plea that “It’s time for EA leadership to pull the short-timelines fire alarm.”

Based on the past week’s worth of papers, it seems very possible (>30%) that we are now in the crunch-time section of a short-timelines world, and that we have 3-7 years until Moore’s law and organizational prioritization put these systems at extremely dangerous levels of capability.[1]

The papers I’m thinking about:

…For those who haven’t grappled with what actual advanced AI would mean, especially if many different organizations can achieve it:

  • No one knows how to build an AI system that accomplishes goals, that also is fine with you turning it off. It’s an unsolved research problem. Researchers have been trying for decades, but none of them think they’ve succeeded yet.
  • Unfortunately, for most conceivable goals you could give an AI system, the best way to achieve that goal (taken literally, which is the only thing computers know how to do) is to make sure it can’t be turned off. Otherwise, it might be turned off, and then (its version of) the goal is much less likely to happen.
  • If the AI has any way of accessing the internet, it will copy itself to as many places as it can, and then continue doing whatever it thinks it’s supposed to be doing. At this point, it becomes quite likely that we cannot limit its impact, which is likely to involve much more mayhem, possibly including making itself smarter and making sure that humans aren’t capable of creating other AIs that could turn it off. There’s no off button for the internet.
  • Most AI researchers do not believe in ~AGI, and thus have not considered the technical details of reward-specification for human-level AI models. Thus, it is as of today very likely that someone, somewhere, will do this anyway. Getting every AI expert in the world, and those they work with, to think through this is the single most important thing we can do.It is functionally impossible to build a complex system without ever getting to iterate (which we can’t do without an off-switch), and then get lucky and it just works. Every human invention ever has required trial and error to perfect (e.g. planes, computer software). If we have no off-switch, and the system just keeps getting smarter, and we made anything other than the perfect reward function (which, again, no one knows how to do), the global consequences are irreversible.
  • Do not make it easier for more people to build such systems. Do not build them yourself. If you think you know why this argument is wrong, please please please post it here or elsewhere. Many people have spent their lives trying to find the gap in this logic; if you raise a point that hasn’t previously been refuted, I will personally pay you $1,000.

There are several interesting things about this argument. First, in response to pushback, the author retracted the argument.

This post was rash and ill-conceived, and did not have clearly defined goals nor met the vaguely-defined ones. I apologize to everyone on here; you should probably update accordingly about my opinions in the future. In retrospect, I was trying to express an emotion of exasperation related to the recent news I later mention, which I do think has decreased timelines broadly across the ML world.

LessWrong is thus one of the few places in the world you can be shamed for not being Bayesian enough!

I am more interested, however, in the general question when will we know to pull the plug? And will that be too late? A pandemic is much easier to deal with early before it “goes viral”. But it’s very difficult to convince people that strong actions are required early. Why lockdown a city for fear of a virus when more people are dying daily in car accidents? Our record on acting early isn’t great. Moreover, AI risk also has a strong chance of going viral. Everything seems under control and then there’s a “lab leak” to the internet and foom! Maybe foom doesn’t happen but maybe it does. So when should we pull the plug? What are the signals to watch?

That was then, this is now

The Red Army collapsed in the first weeks of the war.  This is no criticism of its individual troops.  It is a statement about bureaucratic rule, coercion, lies, fear, and mismanagement.  The problems were not new, nor were they unfamiliar.  Lack of transport, for instance, which was identified by nearly every front-line officer as the reason the retreat turned into a route that June, was a long-standing concern of units based along the Soviet border.  “It is absolutely unknown to us where and when we will receive the motorized transport we need for newly mobilized units”…Spare parts, fuel, and tires were impossible to guarantee.

Circa 1941, that is from the very good Catherine Merridale, Ivan’s War: Life and Death in the Red Army, 1939-1945.  Do not arrive too readily at conclusions about the current situation in Ukraine!  And Merridale books are in general a good place to read about Russian history.

What should we expect from the most accurate sources?

The best estimates of a maximally accurate source would be very frequently updated and follow a random walk. And authoritative sources like WHO are often said to be our most accurate sources. Even so, such sources do not tend to act this way. They instead update their estimates rarely, and are especially reluctant to issue estimates that seem to “backtrack” on previous ones. Why?

First, authoritative sources serve as a coordination point for the behavior of others, and it is easier to coordinate when estimates change less often. Second, authoritative sources need to signal that they have power; they influence others far more than others influence them. Both of these pressures push them toward making infrequent changes. Ideally only one change, from “we don’t know”, to “here is the answer”. But if so, why do they feel pressures to issue estimates more often than this?

…authoritative sources prefer a strong consensus on what are the big sources of info that force them to update. This pushes for making very simple, stable, and clear distinctions between “scientific” info sources, on which one must update, and “unscientific” sources, where it is in considered inappropriate to update. Those latter sources must be declared not just less informative, but uninformative, and slandered in enough ways to make few tempted to rely on them.

Due to the third of these pressures, authoritative sources will work hard to prevent challengers competing on track record accuracy. Authorities will issue vague estimates that are hard to compare, prevent the collection of data that would support comparisons, and accuse challengers of crimes (e.g., moral positions) to make them seem ineligible for authority. And other kinds of powers, who prefer a single authority source they can defer to in order to avoid responsibility for their decisions, will help to suppress such competitors.

Here is more from Robin Hanson.

*Memoria*

This is the new film directed by Apichatpong Weerasethakul and starrting Tilda Swinton in a broader artistic collaboration.  Weerasethakul, in case you do not recognize the name, is director of Uncle Boonme Who Can Recall His Past Lives, arguably the best movie of the last twenty years.

I won’t try to summarize the plot of Memoria, only to tell you it is set in Colombia (the director’s first film outside of Thailand), mostly in Spanish, has A+ cinematography, and it concerns how the revelation of Buddhist Enlightenment might begin to spread through the world.  I’ve read a half dozen reviews of the movie, but none are very insightful and they barely mention the Buddhist angle (and Hinduist animism), central to understanding Weerasethakul’s films.

This movie is almost certainly the best of the year, or more.  And it will cement Weerasethakul’s reputation as the most important director of our time.

Fortunately, it is not available on streaming services for the stay at home small screen barbarians.  It is being rolled out in an unusual fashion, one city at a time (more or less), and currently it is playing for this week only at AFI in Silver Spring, Maryland.  The roll out is poorly organized and publicized.

Recommended, don’t miss it.

The Competition for High-Skill Immigrants Intensifies

The UK has created a new visa for High Potential Individuals. Under the HPI visa any graduate from a top university as defined by “in the top 50 of at least two of the following three ranking systems: (1) Times Higher Education World University Rankings, (2) Quacquarelli Symonds, (3) The Academic Ranking of World Universities” will be allowed to stay in the UK for two (BA, MA) or three years (PhD). Moreover, a job or sponsor is not required and spouses and dependents are also included.

The US is slowly–very slowly–working towards something similar. I wrote this 12 years ago (!):

Behind Door #1 are people of extraordinary ability: scientists, artists, educators, business people and athletes. Behind Door #2 stand a random assortment of people. Which door should the United States open?

In 2010, the United States more often chose Door #2, setting aside about 40,000 visas for people of extraordinary ability and 55,000 for people randomly chosen by lottery.

It’s just one small example of our bizarre U.S. policy toward high-skill immigrants. Every year, we allow approximately 140,000 employment visas, which cover people of extraordinary ability, professionals with advanced degrees, and other skilled workers. The number is absurdly low for a country with a workforce of 150 million. As a result, it can be years, even decades, before a high-skilled individual is granted a U.S. visa. Moreover, these 140,000 visas must also cover the spouse and unmarried children of the high-skilled worker, so the actual number of high-skilled workers admitted under these programs is less than half of the total. Perhaps most bizarrely there is a cap on the number of visas allowed per country regardless of population size. How many visas are allocated to people of extraordinary ability from China, a country of over 1 billion people? Exactly 2,803. The same number as are allocated to Greenland.

The above mostly still holds today. The US Competes Act, which has passed the House, however, would create more visas for high-skill immigrants.

The bill also exempts foreign nationals with a Ph.D. in science, technology, engineering, or mathematics (STEM) from the U.S. or foreign equivalent university or college and their family members’ annual green card limit caps. The language would also add health professions and those in a critical industry to the national or economic security of the U.S. to the definition of a Ph.D. in STEM for purposes of the exemption.

Contemporary music listening

A few of you have asked me what I am listening to lately in terms of contemporary releases.  I don’t feel this list contains any unusual revelations, but here goes:

Poppy, I Disagree.

Iceage, Seek Shelter.

Finneas, Optimist.

Brittany Howard, Jaime.

Courtney Barnett, Things Take Time, Take Time.

Michael Hurley, The Time of the Foxgloves.

Some kind of Moroccan (?) “noise” CD, but I can no longer read the group name or title.

Also, it is not contemporary but I weakened and bought the 30-CD Fela Kuti box.

When your bot is better than you

I expect most written communication will eventually be done by bots. I could train my bot by letting it read all my previous email and other writings. Eventually my bot would answer most of my email directly, though it could hold some aside to ask me whether they merited a personal response.

This sounds convenient, and in many ways it will be. I’ll have more time for taking walks and reading books. But think through the broader equilibrium. If more emails are read by bots, then more emails will be written by bots. Of course that is already the case, but in this new world the bot-composed emails will be at least as good as human emails, and at least as good at getting through whatever filters I set up to protect my time and attention.

A kind of arms race will ensue. Overall, I expect the number of quality messages and emails to rise. Woe unto those who do not have a very good filtering bot.

Imagine negotiating or discussing terms in such a world. I might receive a proposal from your bot. Is it a real, legally binding offer? Or is it simply a ruse to get me to reveal information about my negotiating strategy? In some cases bots might handle these problems smoothly and present both sides with a final settlement. In other cases, negotiators might insist on a face-to-face meeting, both to know they are getting “the real deal” and to limit the potential for back and forth. For some real-world interactions, online written communications will no longer be good enough.

Think about the college admissions essay, for example. Nowadays it is important. But if the bots become good at writing, applicants might have to show up for a personal interview instead. Countermeasures might then evolve. Maybe there aren’t enough admissions officers to conduct all of those interviews. So why not let the applicants spend two days together, tape all the proceedings, and let the bots issue ratings? They might even measure who told the most original jokes.

In this new world, skill at writing will count for much less, and personal charisma for much more. This is not necessarily a positive development. It will be harder to use writing as a measure of broader skill or intelligence.

Here is the rest of my Bloomberg column on the topic.

A ray of good news

Every now and then this country, and for that matter the ACLU, does the right thing:

The American Civil Liberties Union helped scuttle a bill this week that would have enabled the Biden administration to liquidate Russian oligarchs’ assets and turn the proceeds over to Ukraine.

ACLU officials told lawmakers Tuesday that the legislation could run afoul of due-process protections in the U.S. Constitution because it does not allow its targets to challenge the government’s actions in court, according to two people familiar with the matter who spoke on the condition of anonymity to describe private talks. ACLU officials warned that the measure would probably be struck down by the judicial branch if enacted as proposed, giving Russia a potential propaganda victory over the United States, the people said.

Here is the full article.  Shame on all of you who supported this!

Television gets you to spend money

Especially on cars (and other durable goods):

I compare growth in retail sales between areas with and without local TV service over the unanticipated Federal Communications Commission (FCC) Freeze, which halted the licensing of new TV stations from 1948–52. I find three results that corroborate TV’s long-attributed role in American consumerism. First, during the Freeze, total retail sales in counties with TV access increased by 3–4% more on average than in counties without access. Second, the effect of TV was concentrated in the automobile sector, which alone accounted for a third of the total difference.

Here is the full paper by Woojin Kim, via the excellent Kevin Lewis.

What true conservatives should care about

That is the theme of my latest Bloomberg column, here is the opening bit:

If you are a true conservative — and I use the term not as Ted Cruz might, but in its literal sense, as in conserving what is of value in the modern world — then you should be obsessed with three threats to the most vital parts of our civilizational heritage, all of which are coming to the fore: war, pandemic and environmental catastrophe.

These three events have frequently caused or contributed to the collapse or decline of great civilizations of the past. After being seriously weakened by pandemics and environmental problems, the Roman Empire was taken over by barbarian tribes. The Aztecs were conquered by the Spanish, who had superior weapons and also brought disease. The decline of the Mayans likely was rooted in water and deforestation problems.

I think of true conservatism as most of all the desire to learn from history. So let us take those lessons to heart.

Two further points:

1. I don’t think of this as existential risk, rather humanity could be set back very considerably, with uncertain prospects for recovery.  In the median year of human history, economic growth is not positive.  A few thousand years of “Mad Max” would be very bad.

2. I think you should aspire to be more than just a “true conservative.”  You should be a liberal too!  So there is more to the picture than what the column outlines.  Nonetheless I see it as a starting point for reformulating a morally serious conservative movement…

Recommended.