Strategic High Skill Immigration

post by Gentzel · 2017-12-06T01:15:19.387Z · LW · GW · 8 comments

Contents

  How to improve the global economy, increase strategic stability, and safeguard the far future.
    Part 1: The current situation: shifting balance of power
    Part 2: The arguments
    Strategic benefits of high skill immigration for the far future:
    Possible objections: 
    Conclusion:
None
8 comments

How to improve the global economy, increase strategic stability, and safeguard the far future.

Part 1: The current situation: shifting balance of power

When trying to help people at scale, it is often prudent to go beyond direct interventions and to seek equilibria which are in the long run interest of humanity, especially if we assign any significant value to the far future. From this we can derive the instrumental value of reducing risks which are multi-generational or permanent in nature, and ensuring the safe development of transformative technologies which could pose immense risks either on their own or in the wrong hands. Historically, wars and state competition have had a great deal of influence on these sorts of concerns: competitions between nations can lead to arms races that produce new dangerous or beneficial technologies, the destruction of wars themselves can have permanent effects, and the victors of wars determine the institutions which govern humanity and its technology going forward in time.

As China’s economic growth continues to outpace US economic growth, and its military expenditure as a percentage of GDP continues to remain stable it is plausible that the US and China may find themselves in a different equilibrium from the current peace they are used to, and plausibly a similar trap to the one the UK and Germany faced just before WW1. Though the British Empire spanned much of the globe and had the largest Navy in the world, in 1898 Germany began a naval arms race with Britain, just before 1914 the German economy caught up, and shortly after World War 1 followed.

Known as the Thucydides Trap, in 12 out of 16 cases where a rising power challenged an established power in the past 500 years, war was the result. Though the pattern may seem surprising, it is rather consistent: with no outside enforcer with coercive power to make peace agreements between great powers trustworthy, it is fairly difficult for countries to credibly commit to one of the many outcomes that leaves both parties better off than war, especially since as the rising power becomes more powerful, it continuously gains negotiating advantage, while the formerly ruling power loses it. In the four cases above where war was avoided, one involved a prior war, and external intervention by the pope, two cases involved competing allies, and the remaining case was the rise of the Soviet Union, which never actually matched the economy of the United States.

While China’s economy could stagnate due to the structure of its institutions, or it could fail to compete with the US on a military level due to a technological research disadvantage and lack of alliances, China will likely still close in on US dominance as it builds a larger Navy, further exceeds US GDP, disproportionately invests in artificial intelligence, and continues to be less inhibited from genetically modifying humans.

Though the overall economic interests of the US Government and the Chinese Government are fairly aligned, and China has much warmer relations with the US than the Soviet Union did, there are likely to be areas where China could benefit more by further expanding its interest in foreign countries at the expense of US interests once China is in a position of relative power. From the considerations above, we can see that the long term pressures on the US and China (established power vs. growing economic power) will likely lead to conflict (though not necessarily war given deterrence, economic interdependence, and the many norms toward peace in the modern era) and that such conflict could further encourage technology races in areas which are of great concern for the far future, such as AI and human genetic engineering.

To handle this situation, international agreements are valuable for slowing military tech races but they may be hard to generate for many reasons. While economic benefits of not having an arms race are often win-win, some benefits are positional (e.g. being a central node in a trade network), and countries may not accept an economic win-win trade that gives their rival a positional strategic advantage. Even worse, as technologies such as AI and human enhancement will impact economic productivity, there is not even necessarily an immediate economic win-win from generating agreement, just the potential to reduce future risks which current decision makers might discount. While the US and Russia/the Soviet Union were able to generate risk reducing agreements (*cough cough* which Russia cheats on a lot), these were partially caused by agreement on the analysis of the situation: both sides could agree peace and survival was more likely in one equilibrium than another. Unlike debating nuclear weapons of specific speeds, ranges, and numbers of re-entry vehicles, AI has many more applications which could be destabilizing in some manner and generate a strategic advantage for the country deploying more advanced AI systems first. Further, unlike nuclear testing, which is hard to conceal (though Russia still does small yield tests despite the Comprehensive Test Ban Treaty), many AI systems can likely be deployed online without being noticed. Being a much more complex space, it may take substantially longer to analyze the possibilities (of which there are more than humans can imagine, as smarter than human AI will be more intelligent than humans) and agree on regulations that would put both parties in a better equilibrium. Even if countries could agree on the strategic implications of AI (which might be its own risk, because understanding the actual situation involves having a sense of each other’s capabilities, which may incentivize further racing) the rate of innovation could quickly render such agreements not useful due to unpredicted changes in the balance of power, and the slow pace of policy change. Unless the consequences of cheating on agreements are unambiguously worse than cooperating in the minds of policymakers, cheating will occur, and it will occur disproportionately with governments that can conceal activities from investigation. If these sorts of problems can’t be resolved (and we should try very hard to resolve them!), it is valuable to have a backup plan for creating a safe policy environment for the development and deployment of transformative technologies, without having to get every world power to agree on exactly what that environment will be.

While historically, monopoly of force has been a mechanism by which coordination problems are resolved, this does not mean war would be necessary. If in anticipation of the nuclear bomb, one country had monopolized uranium (though unrealistic due to its geographic distribution), there would not have been as much potential for nuclear arms race conditions to develop. The country with control of such resources would be so decisively ahead that other countries pursuing such resources would increase their risk via attracting attention more than they would decrease their risk by shrinking the advantage of the first mover.

A different kind of resource, even more important to the development of such powerful technologies is human capital. The ability of countries to gain and hold highly intelligent people greatly affects their economic and strategic outlook. As concrete examples, nuclear weapons and strategy in the US and China were greatly influenced by the migration of John von Neumann and Quan Xuesen respectively. To compete for the best and brightest, countries may engage in a credible virtue signaling contest of sorts. In the 1940s the US gained human capital at the expense of the USSR and Nazi Germany by being more “virtuous” while China gained Quan Xuesen in 1955s as the result of the US’s unfair behavior toward him in the preceding years and hawkish stance toward China at the time. Such competitions on virtue signaling are preferable to arms races which produce weapons, or other projects that generate new types of intelligence that may generate larger alignment risks than they are equipped to handle (eg. a certain kind of surveillance state doesn’t have to keep the public happy to stay in power). As opposed to creating new abilities, competitions for fixed resources mostly don’t generate new risks. At the worst, a weapons engineer may be more productive in one country than another, but the number of skilled weapon engineers is not increased without additional interventions such as espionage. As having the best minds is a key constraint on developing highly sensitive technologies, a party that chooses not to engage in credible virtue signaling, in favor of racing in another direction may lose advantage in the long run which makes this sort of competition potentially stable and may partially offset other more harmful races (both by costing resources, and by causing brain drain which may prevent certain countries from being capable of arms races in the first place).

So far the framework above implies a model of states as unified rational actors, while in reality they consist of coalitions which are not always aligned toward the central goals of a state, but rather toward those of the smaller groups and individuals. In democracies, broad distribution of power helps keep the government more aligned with its population, but reduces the ability to coordinate on contentious issues and to behave strategically in the long run, hence there may be low hanging fruit where democracies can behave more strategically at low cost. On the other hand, in less democratic systems, top down force can resolve internal coordination problems and get things done but, this power is not necessarily aligned with the population, and can be precarious since when some loyalties switch, a new dominant coalition may develop absolute power. Overall this implies both that high skill immigrants are likely to prefer to live in democratic systems, and that such systems are less able to abuse any advantages over other countries that they gain from such immigrants.

Part 2: The arguments

In bullet point form below, we discuss the potential benefits of further liberalizing immigration policy for individuals of high skill or potential, and ways to address the potential risks of such policy changes.

Strategic benefits of high skill immigration for the far future:

(click each section below for justifying bullet points)

With high skill immigration less restrained, the natural coalescing of top talent increases differentials in progress between leaders and followers: some countries would get more of an advantage in AI and other transformative technologies than others, which reduces competitive pressure, allowing more resources to be allocated to safe design and safe deployment of new technologies rather than to just staying ahead in tech races.

Higher mobility for the highly skilled allows them to move to the systems that are most in their interests: the altruistic can move to systems where they have greater leverage, while the less altruistic can go where they can expect to personally benefit the most. As the best run systems will tend to both have more economic opportunities and larger government budgets, both groups are likely concentrate in the same countries, leaving governments where rational altruists can gain influence more resources at their disposal due to higher tax revenues from immigration driven increases in economic growth.

Increased freedom of movement for skilled workers increases their leverage vis-a-vis states: resulting in net brain gain for those states which most credibly signal pro-social policies.

Increased ease of regulating risky technologies

Higher salaries for white collar work in the US discourage riskier but potentially high paying shady black hat(ish) industries.

Increased alignment between the elites of different countries via overlapping and interlaced social networks

Possible objections:

(click each section below for supporting details and our counter arguments)

Increased high skill immigration may speed up the development of high risk technologies

Increased high skill immigration may increase risks from espionage

Developing top researchers who then return home speeds up arms races by reducing the gap between tech leaders and followers

“Western culture” would be “degraded” (nativist backlash)

Too many “eggheads” in one basket

Erosion of trust and good social norms

Peace prospects are not necessarily increased by inter-socialization

In an equilibrium with more mobility for the elites of different countries, the governance of some countries may worsen, leading to conflict.

Reaction to easier high skill immigrant citizenship in democracies could cause more authoritarian countries to further restrict their own citizens from leaving.

Assuming long term AI alignment problems are solved, but that deployed AI systems aren’t aligned with all humans, a world with multipolar outcomes may be preferable to unipolar ones, since human cooperation and mutual benefit will no longer be the way to wield the most power.

Most democracies already have favorable immigration policies for high skill immigrants, and even in the US there’s multiple visas high skilled persons can take advantage of.

High skill immigration is low leverage: Historically, the immigration shock after the collapse of the Soviet Union didn’t noticeably accelerate science in the US, so the impact is unclear.

Conclusion:

To summarize, we think the strongest future focused reasons to make citizenship easier for high skill/high potential immigrants to obtain are that it would likely increase the strategic advantage of systems that better approximate human values, give rational altruists increased advantage in governance, reduce the odds of catastrophic wars, and reduce the odds of unregulated work and arms races on dangerous technologies. The best potential counters seem to be that we should mitigate the risk of a totalitarian global regime, that the increased information flow that comes from collecting the best and brightest could lead to faster development of dangerous technologies if not controlled, and that via espionage and returning home, such progress on dangerous technologies could proliferate to the rest of the world. We think that these counters could potentially be addressed by increased efforts at strengthening institutions, funding differential technological progress in defensive technologies such as cybersecurity, and spreading ideas of safety engineering among those working in potentially dangerous fields. Since the benefits of increased high skill immigration appear difficult to gain without actually having it, and the downsides can be countered, a well calculated liberalization of high skill immigration is likely the best path forward.

There are many methods by which one could attempt to mitigate arms/tech race type dynamics which could be threatening to the far future of humanity. While exploring ways to achieve agreement and spread knowledge of risks, we should also seek ways to make the best of the situations we find ourselves in should global coordination fail. We provide improving high skill immigration policy as an example of a way in which a state actor could unilaterally make an arms race type of situation less harmful, while also acting in their own interest. There are likely other interventions which fall into a similar category that are worthy of investigation.

In a future post, we will discuss the current barriers to a better high skill immigration policy for the US, where it currently succeeds, how much room there is for improvement, and why such improvements may be tractable across party lines.


Notes: The general outlook of this post was partially informed by the following books:

8 comments

Comments sorted by top scores.

comment by ESRogs · 2019-01-14T19:51:07.559Z · LW(p) · GW(p)
Quan Xuesen

Qian, not Quan. Pronounced something like if you said "chee-ann" as one syllable.

comment by JamesFaville (elephantiskon) · 2017-12-06T06:23:55.044Z · LW(p) · GW(p)

I don't have much of a thoughtful opinion on the question at hand yet (though I have some questions below), but I wanted to express a deep appreciation for your use of detail elements: it really helps readability!

One concern I would want to see addressed is an estimation of negative effects of a "brain drain" on regional economies- if a focused high-skilled immigration policy has the potential to exacerbate global poverty, the argument that it has a positive impact on the far future needs to be very compelling. So would these economic costs be significant, or negligible? And would a more broadly permissive immigration policy have similar advantages? Also, given the scope of the issues at hand I would be very surprised if the advantages you ascribe to high-skilled immigration are all of roughly equal expected value: is there one which you think dominates the others? (Like reduced x-risk from AI?)

Replies from: Gentzel
comment by Gentzel · 2017-12-07T23:15:24.237Z · LW(p) · GW(p)

I suspect high skill immigration directly helps probably with other risks more than with AI due to the potential ease of espionage with software (though some huge data sets are impractical to steal). However, as risks from AI are likely more immanent, most of the net benefit will likely be concentrated with reductions in risk there, provided such changes are done carefully.

As for brain drain, it seems to be a net economic benefit to both sides, even if one side gets further ahead in the strategic sense: https://en.wikipedia.org/wiki/Human_capital_flight

Basically, smart people go places where they earn more, and send back larger remitances. Some plausibly good effect on home country institutions too: https://en.wikipedia.org/wiki/Human_capital_flight#Democracy,_human_rights_and_liberal_values

comment by Chris_Leong · 2017-12-08T03:55:38.539Z · LW(p) · GW(p)

I actually think that recruiting as many AI researchers as possible over to American could be a good way of reducing the chance of an arms race, though it also makes espionage easier too as you point out, so I'm not sure if it ends up being net positive or negative. Nonetheless, this is a fascinating proposal and one that I haven't ever heard before.

comment by Gentzel · 2017-12-07T22:49:30.814Z · LW(p) · GW(p)

The most up to date version of this post can be found here: https://theconsequentialist.wordpress.com/2017/12/05/strategic-high-skill-immigration/

comment by whpearson · 2017-12-06T13:26:42.571Z · LW(p) · GW(p)
In a singleton situation, those with power are more able to do what they wish and make the world how they think it should be, in a multipolar situation there is less room for maneuver and more incentive to put all resources toward outcompeting others to ensure survival or dominance: leaving far less potential surplus value for other humans.

We have lived with a multi-polar world for a long time and have puts tonnes of resources into things that are not survival or dominance. What do you think will change?

To me it seems that drawing all power to one entity robs the rest of autonomy and self-determination. There have been arguments that this autonomy is important for happiness. So if I had huge amounts of power I would seek to find a way to distribute power to give value to other humans. This means we are in a bit of a double bind. Either we stay in a multi-polar world where co-ordination is hard but people have autonomy, or we aim for a singleton world where lots of people might feel miserable and powerless, but well fed and safe.

Replies from: Gentzel
comment by Gentzel · 2017-12-07T23:00:15.563Z · LW(p) · GW(p)

We have lived in a multi-polar world where human alignment is a critical key to power: therefore in the most competitive systems, some humans got a lot of what they wanted. In the future with better AI, humans won't be doing as much of the problem solving, so keeping humans happy, motivating them, etc. might become less relevant to keeping strategic advantage. Why Nations Fail has a lot of examples on this sort of idea.

It's also true that states aren't unified rational actors, so this sort of analysis is more of a course grained description of what happens over time: in the long run, the most competitive systems win, but in the short run smaller coalition dynamics might prevent larger states from exploiting their position of advantage to the maximal degree.

As for happiness, autonomy doesn't require having all options, just some options. The US is simultaneously very strong, while also having lots of autonomy for its citizens. The US was less likely to respect the autonomy of other countries during the cold war when it percieved existential risks from Communism: centralized power can be compatible with increased autonomy, but you want the centralized power to be in a system which is less likely to abuse power (though all systems abuse power to some degree).

Replies from: whpearson
comment by whpearson · 2017-12-12T18:18:17.105Z · LW(p) · GW(p)
We have lived in a multi-polar world where human alignment is a critical key to power: therefore in the most competitive systems, some humans got a lot of what they wanted. In the future with better AI, humans won't be doing as much of the problem solving, so keeping humans happy, motivating them, etc. might become less relevant to keeping strategic advantage.

Then you have an alignment problem. AIs should be making decisions consistent with human values. If AIs are making the world a worse place just by their existence, then SOMETHING HAS GONE VERY WRONG.

in the long run, the most competitive systems win,

That is a truism. In evolutionary history, competitive does not mean necessarily mean the biggest or even smartest though.

As for happiness, autonomy doesn't require having all options, just some options. The US is simultaneously very strong, while also having lots of autonomy for its citizens. The US was less likely to respect the autonomy of other countries during the cold war when it percieved existential risks from Communism: centralized power can be compatible with increased autonomy, but you want the centralized power to be in a system which is less likely to abuse power (though all systems abuse power to some degree).

I may be wrong, but I expect a system that needs to maintain power without an external threat to be a lot more unforgiving on autonomy. It seems that every action that might lead to an increased chance of the sovreign losing DSA would have to be forbidden and cracked down upon. With a multipolar situation you don't rock the boat too much because your country, that you like, might lose to another.

Also with a sovreign I see no chance of fixing any abuse. In a multi-polar situation (especially if we go with merge with AI route), future people can choose to support less abusive power structures.