China's military AI capability is growing. National security Republicans need to make a choice.

Beijing (source)

Defense scholars agree: China's military technology is developing rapidly, with AI leading the way. As Elsa Kania makes clear in a new report, America's historic dominance in this arena is at risk.

This is so for several reasons. Most fundamentally, the dual-use nature of artificial intelligence—meaning it has both military and nonmilitary applications—ensures that knowledge and know-how diffuse rapidly throughout the globe. Private companies, academic research institutions and military organizations all compete and cooperate in advancing AI, which refers to a bundle of statistical techniques with varied applications. This contrasts with more traditional military innovation that focuses solely on defense and, at least initially, has much less integration with the broader economy. Achieving a large and durable advantage in military AI may be difficult for any country, given the freewheeling shape of private research efforts.

Structural issues aside, the US lags China in military AI development for three main reasons, two of which are self-inflicted.

First, AI leaders in the US (Google, Facebook etc.) are less keen on working with the military to develop weapons and defense applications compared to China and its 'military-civil fusion'. US companies are subject to different sorts of competitive pressures involving reputation and brand, and may differ in their internal cultures.

Second, US policymakers have failed to outline a comprehensive AI strategy appropriate to the scale of the challenge. Consequently, the total amount of public resources and attention devoted to advancing US military AI is insufficient to maintain technological dominance. Although the military itself is aware of the problem, politicians in both parties have not prioritized the issue nor have they outlined detailed plans for investing in AI-focused training, education and research.

Third, certain advantages the US has long enjoyed over China relating to economic dynamism and human capital are increasingly being undermined by the Republican Party's descent into racism, xenophobia and anti-immigrant hostility. Frontier AI research is migrating to other countries like Canada, while US colleges and businesses struggle to attract and retain foreign students in light of unwelcoming rhetoric and policy.

This last point is perhaps the most pivotal. It also contains deep ramifications for the future of the Republican coalition.


The Republican Party is changing

Republicans have long been defined by a "three-legged stool" alliance of social conservatives, business/free market interests and strong national defense proponents. For equally as long, pundits have observed various instabilities inherent in the setup, frequently predicting an impending coalitional crackup.

Most recently, smart observers have noted the striking and rapid disappearance of gay marriage opposition as the core unifying symbol of social conservatives. Under Trump, the coordinating role of religion appears to have been swapped-out with a toxic mix of barely-concealed racism, white ethnonationalism and anti-immigrant scapegoating.

Political scientist Lee Drutman has described a potential realignment of US politics whereby socially-liberal, urban, cosmopolitan business types join the Democratic Party and white, rural/suburban, socially-conservative 'hard hats' flee to the Republicans. This is indeed happening to some degree—the most salient cleavage in US politics is now between roughly-described 'cosmopolitans' and 'anti-cosmopolitans'.

Over the last few years, much of this analysis about the future of the Republican coalition has tended towards the social and economic dimensions. Overlooked, however, is the fate and agency of its national security leg.

This is likely due in part to the curious fact that while US politics overall is at a moment of peak partisan polarization, national security has for some time been remarkably non-polarized. Rhetoric aside, there has been a great deal of continuity spanning the administrations of Bush to Obama to Trump. Both Democrats and Republicans enthusiastically supported the Iraq war, continue to support vast military budgets, and have not substantively objected to the ongoing drone-fueled foreverwar. The permanent national security bureaucracy ('the blob') appears firmly entrenched, despite the media's breathless commentary over Trump's latest personnel choices. Indeed, foreign policy critics on the Democratic Party's left flank seem energized, perhaps indicating the end of peak non-polarization.


National security now requires cosmopolitanism

This brings us to a key dilemma facing the Republican Party: the demands of military AI place its national security wing on a collision course with its ascendent anti-immigrant, anti-cosmopolitan elements.

As advanced technologies like AI become relatively more crucial to US military power, maintaining global leadership in higher education, frontier innovation and elite talent also become relatively more crucial. How Republicans choose to navigate the increasingly unavoidable fact that US military dominance requires an embrace of immigration, diversity and social tolerance is a core question for the coming decade.

While speculation has centered on when and how US politics will reorient along social and economic lines, the prospect that national security Republicans will jump ship seems underappreciated. Alternatively, the 'strong on defense' camp could succeed in leveraging its clout within Republican politics to force productive change.

Perhaps the likeliest scenario involves neither option: Republicans could simply keep the US on a path towards eventual military decline. A definite analog exists in economic policy, where curtailing immigration will result in less growth and an older, whiter, smaller America. This is undoubtedly a tradeoff that the current Republican Party resoundingly endorses. Yet there is some reason to believe that military matters will play out differently.


China benefits from the status quo

Unlike slow-moving population demographics and macroeconomic trends, military power has a habit of changing quickly and spectacularly.

Whereas a small decline in US GDP growth might never jolt elements of the Republican coalition into action, a cinematic display of China's military advancement could galvanize US national security interests and lead to a more coherent AI strategy. China, of course, likely understands this situation and recognizes its incentive not to show off or make flashy displays of capability. By keeping its progress under the radar, China avoids creating any focusing events that might prompt US political action.

That a sizeable share of the US national security state has its home in the current Republican Party is a dream come true for China's AI ambitions. The US is still the global nexus of AI innovation and human capital. But as long as its defense interests remain aligned with anti-immigrant, anti-cosmopolitan politics, America will struggle to unleash its underlying strengths and keep pace with the rapidly changing technological landscape.

Tweet

Under Trump, presidential rhetoric has gotten less important

source: unknown (sorry)

Compared to his predecessors, the correlation between Trump's presidential rhetoric and government outcomes--both current and future--is dramatically lower.

This is true for two reasons. One, the signal-to-noise ratio of his communication is wildly lower: instead of talking and writing in a clear way, Trump's words have a large proportion of incoherent sentences and riffing; of the share that is coherent, a much lower proportion is about government outcomes directly.

His rhetoric contains a huge amount of commentary and reflection on media narratives, social conflict and status evaluation--all of which have a less direct causal relationship to tangible government outcomes. In other words, the absolute amount of important information contained in each presidential communication is far lower compared to past presidents (it might be fair to say that his total communication quantity is greater overall (i.e. he uses twitter differently), but this effect isn't enough to make up the difference--he isn't communicating that much more).

Secondly, of the (smaller) share of presidential rhetoric that is coherent and directly on-topic, the conversion rate of meaningful statements about government outcomes into actual government outcomes is lower.

Trump's constant lying, bullshitting, overconfidence and commitments to unrealistic outcomes degrade the value of his on-topic communication: compared to past presidents he's simply not as reliable a guide to what's actually going to happen (or is happening). Most politicians are careful about what they say, either because they have an inherent and healthy respect for language and social interaction, or because they know huge mismatches in rhetoric-vs-outcomes can be used against them as effective political weapons. It has become conventional wisdom to note that the range of likely government outcomes is far higher under Trump compared to past presidents; at least some of this is due to Trump's unwillingness (or inability) to treat rhetoric as a constraint.

In general, the main takeaway regarding Trump's on-topic presidential rhetoric is this: there's less of it, and it's less reliable.

Well so what? Why does presidential rhetoric matter at all? And what should we do if we do think there is a problem? Let's take each of these in turn.


I. 
Presidential rhetoric--what a president communicates to the public across any medium--matters for several reasons.

First, it matters for its own sake. Government has a socially symbolic component. Rightly or wrongly, people view elected officials as representing some metaphysical "will of the people", so when the highest-level politician speaks, in some sense a collective "America" is speaking. People get emotionally attached to politicians, and care about what they say irrespective of the downstream effects on public policy and culture. I've written previously about why I think these sorts of internal subjective psychological effects aren't a very good basis for ethics and advocacy, but they do undoubtedly matter to a lot of people.

Second, presidential rhetoric matters because it affects other things people care about--it has instrumental value. One example, related to the first point, might be social or cultural effects: Trump's words have inspired and antagonized, prompting actions by other people which in turn trigger their own effects. These sorts of causal relationships are often complicated and hard to prove, and are again highly subjective and value-laden.

More clear, however, is the simple informational utility of what presidents say. The federal government is massive and sprawling, so presidential rhetoric is a really convenient way of getting information. If you don't follow politics much, a president-focused narrative is the simplest and most natural focal point for achieving minimum political awareness*

To break down the informational aspect of presidential rhetoric somewhat further, it gives us data about current government outcomes and future government outcomes (a subset of which is character/personality information).

When a president provides information about things that have just happened, or are happening, there's generally not much wiggle-room between rhetoric and reality. Statements are usually able to be verified and evaluated for accuracy. In this category, Trump's mendacity is especially unusual and ridiculous.

Presidential communication about future government action--policies being pursued, broad goals, etc.--is by far the biggest source of value. Presidential politics mostly matters to the extent that the president's choices affect people's lives in tangible ways; presidential rhetoric is the best source of information about what those choices might be. When a president says they're going to do this or that, we increase the probability of those associated outcomes. When they say they absolutely won't do that or this, we decrease the associated probabilities.

This type of value--information about future government actions--is the reason why foreign governments and financial markets hang on the president's every word. The assumption that if a president says they'll pursue something they actually will is, generally speaking, accurate. In Trump's case, his rhetoric undoubtedly provides directional probability adjustment information--merely much less per communication relative to his predecessors.

A big subset of this dimension of value concerns how rhetoric can provide information about a president's analytical style, personality, character, cognitive profile, etc. Hearing presidents talk about stuff feeds data into everyone's implicit agent-modelling system, which is used to make predictions and update outcome probabilities. Trump's incoherent speaking style, bizarre feuds, and short sleeping schedule revealed by tweet timestamps provide information about his psychological health, for example.


II.
So what does this all mean? Trump's unique communication style has made presidential talk cheaper than ever in terms of its correspondence to real, meaningful policy outcomes. And yet it's driving political media narratives more than ever as well. In addition, activists and political opponents are harnessing the daily swirl of crazy, offensive and unusual statements to galvanize support. It's a weird dynamic.

My overall impression is that the media has to some degree lost its way and become marginally less focused on tangible government outcomes--upon which heavy presidential rhetoric reporting is largely justified. The media increasingly seems lost to a sea of personality-driven process coverage. Political media institutions have developed norms of coverage based on the longstanding tight relationship between presidential rhetoric and government outcomes. Trump's cheap talk has degraded that relationship, yet media institutions have so far failed to adjust accordingly.

I have no illusions that we will see a renormalization of media attention paid to public policy outcomes instead of process narrative. People have been griping about how the media "doesn't cover the issues" forever. But it is striking to see us moving so heavily in the opposite direction, driven primarily by changes on the government side instead of the journalism side.

Historically the problem has been: "given the quantity and reliability of presidential rhetoric about government outcomes, journalists aren't talking about government outcomes enough". But now, it seems increasingly: "given journalist incentives and norms in covering presidential rhetoric, the quantity and reliability of rhetoric about government outcomes is declining." In other words, the primary driver of the change in total quantity of coverage about government outcomes is being caused by changes in presidential rhetoric, not changes in journalism coverage.

I don't know what the solution is--certainly it's hard for individual defensively-profit-focused media institutions to unilaterally change their approach and not lose relevance.

I am reminded, however, of the original promise of data journalism. 2015, 2016 and 2017 so far have been big years for the traditional "find a source, get a quote" style of journalism. That's nice, but it does strongly reinforce an emphasis on personality-driven process soap opera narrative spinning. The structure of data journalism forces a certain skepticism of process narratives, necessarily rebalancing towards more coverage of tangible outcomes (that are measurable). I think we could use a little more of that these days.







*This is a somewhat meta point: people follow presidential rhetoric in part because everyone else does, not because they care about the information it may provide about future government outcomes. Many people care about being perceived by their peers as at least somewhat knowledgeable about federal politics, so media coverage about the president is the lowest-cost way to achieve that minimum awareness. If anyone tried to unilaterally stop paying attention to presidential rhetoric, they would bear a social cost without changing the prevailing equilibrium of [presidents are how we understand federal politics].

Tweet

Protesting politicians where they live: at the airport

President Trump's executive order essentially banning the border entry of non-citizens and green card-holders from certain Muslim-majority, geopolitically-weak countries has prompted a set of protests at an unusual location: airports.

This seems due to messy implementation causing some arriving travelers to get held in transit, unable to leave the airport. The suddenness of the rule change has also created a set of extremely vivid travel nightmare scenarios--people on vacation unable to return, employees and students stranded attending conferences and professional events, émigrés moving to the US now without a place to stay.

To say nothing of the long-term, structural effects of this policy on US security, standing and vibrancy, I think it's fair to say that coordinating protests around airports is a somewhat novel and interesting tactic.

"If they can't fly, no one can"

I.
Protests typically emerge in a somewhat complex, bottom-up manner--a mix of both explicit and implicit coordination. As a result, most protests happen in "logical" places that sound like places most protests would happen. City centers, town squares, police precinct buildings, Trump hotels and property. Sometimes, however, activists coordinate around unlikely locations; think about the recently-rediscovered innovation of seizing major highways.

The highway shutdown strategy is fascinating in part because it reveals how many US cities lack a truly universally-recognized "central plaza" that ordinarily serves as the dominant coordination point.

But it's also effective because it hurts random people who are just trying to go about their business. This may seem like a bug instead of a feature--those motorists aren't to blame for whatever injustice is at hand, after all. But in addition to raising the probability of making the news, randomly-targeted protests strike at everyone in a probabilistic sense. Interfluidity put it best:

In small matters, the fact that people will bear disproportionate costs to protest small ripoffs is essential to the integrity of everyday commerce. In larger affairs, the human propensity to altruistic punishment means we all bear costs of perceived injustice, we all have a stake in finding some mix of society and legitimating ideology under which outcomes are perceived as broadly right.

II.
So what about airports? Given the protestor goals, it seems to me that the attempted activist shutdown (or slowdown) of major airports has stumbled into a fairly smart approach. For many reasons, the psychology of air travel looms large in the public consciousness. Think about it--people are so weird about air travel! Everyone has their horror-story of missing flights, getting stuck, losing a passport, etc. etc.

The incredible vividness of the stories of people getting screwed by Trump's travel ban should easily trigger news consumers' flight anxiety and travel disaster porn mental subroutines. This seems likely to increase total empathy relative to the losers from other, more abstract policies, perhaps increasing the political pushback.

A second reason why protesting at airports might be an effective tactic is that federal politicians spend a massive amount of time in airports, and thus might be personally more exposed to disturbances targeted there. I am reminded of an old article by Alex Pareene about how a budget standoff quickly resolved after it started affecting air travel convenience:

“Shuttle flights between Washington and New York were running 60 to 90 minutes late,” the Times reports. Do you know who takes weekday shuttle flights between Washington and New York? People who think they are too important for the train, let alone the bus. People Congress listens to. (People Congress is, also.)

We'll see what happens.

Tweet

Free trade isn't about economics

Politics isn't about policy, so might free trade deals not really be about economics?

Daniel Drezner has an excellent post summarizing a recent burst of media attention concerning the benefits of free trade, and highlighting a bunch of NAFTA's non-economic dimensions. I agree with all of this, and it reminded me of several arguments I made repeatedly when discussing TPP with skeptics.

Firstly, in a narrow short-run economic sense many of these deals simply don't matter that much one way or another. I read the evidence as pointing to a moderate benefit for US GDP (helping especially poor people in the form of lower prices), but many either disagree or don't view these factors as decisive.*

The next big rationale for increasing trade and integration is essentially geostrategy. This is what Drezner describes regarding the US and Mexico, and what many people held up as an argument in favor of TPP (without TPP the Asia-Pacific region will integrate with China instead and institutionalize less-good rules). Another classic example is the European Union: trade integration was started after WWII explicitly to make future conflict unfathomable.

Those two categories of benefit encompass most of what makes free trade desirable. But I want to add an additional point that is more speculative.

A maze (source)

Fundamentally, there is a symbolic benefit in having our government validate and promote cosmopolitan ideals and openness. Maintaining the social narrative of human progress is important, and rejecting attempts to bring us closer together signals a backwards, defeatist attitude.

More concretely, I also think passing deals like TPP would make future trade deals more likely by 'reducing the temperature' of trade politics.

Deals these days are increasingly rare, and also increasingly massive. This has created a positive feedback loop that encourages even more interest groups to try and get in on the lobbying action (i.e. if we're only passing one trade deal every 15 years, you'd best have a say). One side effect of huge complex bloated trade agreements is that literally anyone can find something they disagree with, deepening the structural marketing bias in favor of the status quo.

But--you might say--wouldn't passing a massive kludgy trade deal like TPP positively reinforce the trend towards bigger and more bloated? To some degree probably. But forces would also operate in a different direction, towards smaller and more frequent deals. Passing anything--even TPP--would likely have a net effect of depressurizing the trade debate and shifting expectations for interest groups in a positive direction (showing that deals are possible and reducing the need for marginally-connected groups to get in on each one). In other words, the failure of trade deals does more to crank up the ratchet of political awfulness than does their passage.

I hope the public increasingly comes to recognize the economic, geostrategic and moral/symbolic benefits of free trade and further liberalisation. But the degree to which the debate itself has become mired in a structural trap is underappreciated, and suggests a rationale for supporting the passage of deals regardless of their specific content.




*To be clear: the boundaries of what defines "economic" vs "non-economic" isn't obvious; many of the non-economic benefits in fact operate via formal and informal institutions which shape economic behavior and outcomes (economics→institutions→economics). A classic case is Russ Roberts' argument that free trade can have powerful generational economic benefits, even if it hurts some people alive today.

Tweet

Information Nexus

1. Matthew Kahn on climate change adaptation

2. Very good Megan McArdle on long-term health care

3. Trailer for "The New Fire" nice title

4. That article claiming life expectancy was limited to 115 years is wrong

5. Interesting short documentary about ethnic diversity in Japan (Brazilian, specifically)

6. Tons of low-hanging fruit in aviation policy

7. Cyberpunk af irl

8. Clever infill development in Salt Lake City

Tweet

NASA and Earth science

Newt Gingrich has spoken aggressively about the moral imperative of space colonization

I recently had a good Facebook discussion about the possibility of NASA scrapping its Earth science research under Trump. I take a somewhat contrarian attitude on this and feel genuinely conflicted about whether or not it's a good idea. California Governor Jerry Brown's¹ recent pledge to maintain climate research continuity in the face of federal cuts reminded me of the exchange, and I want to post my thoughts here. Below are edited excerpts of my comments:

I'm still genuinely uncertain about whether space exploration is more important than climate change. They are obviously both super-duper important. In my mind, the desirability of Trump's plan ultimately boils down to a comparison of the magnitudes of various good and bad aspects, as opposed to a discrete conceptual/logical argument. Complicated!

Most obviously, if the cost savings from scrapping Earth science research aren't ploughed into space-focused projects, it's probably an awful outcome. NASA deserves far more total funding. For argument's sake, my list below mostly assumes a constant level of funding--changes in composition likely pale in comparison to changes in absolute size.

That said, if NASA were to scrap Earth science I'd expect to see all of the following effects in various degrees:

Bad

• Government Earth science research using space data collection gets gutted
• Climate change people in NASA and elsewhere get demoralized
• NASA loses relevance/status by not being a major player in climate change science

Good

• More funding/emphasis on space exploration
• NASA becomes more bureaucratically effective by eliminating a conflicting focus (Make NASA great again?)
• NASA increases relevance/status by eliminating a really partisan chunk, maybe inducing Republicans to support a beneficial proxy war of space achievement with China
• Maaaaybe this galvanizes environmentalists and results in some beneficial activism/politics
• And maaaybe it causes other environmentalist/science groups to compensate and invest in space capacity (public or private) & reconceptualize space as a useful method (instead of an "issue" or "policy area"). Example: Bill McKibben pushes for a satellite network.

The last point is perhaps the most compelling. By housing all government space stuff within NASA, you essentially pigeonhole space as "that thing NASA does". But a true proactive and forward-thinking attitude understands that space is merely a natural evolution of goal-seeking entities everywhere; space has great utility. 

Various US military institutions already have a large foothold in space: they justify this on the basis of an existing mandate. There is no reason why NOAA or EPA can't follow the same path and develop their own space expertise. By refocusing NASA on outward-looking space topics, perhaps you'll force a clarification of the value of inward-looking space topics. This idea no doubt makes some Earth scientists nervous. But I am optimistic that the value of space technology is clear enough as to assure its ongoing funding.

¹ I am on record that Jerry Brown is Good. Evidence here, here, here, here

Tweet

Maybe Trump is right to skip daily intelligence briefings

tfw too intelligent for daily intelligence briefings

President-elect Trump has received a fair amount of criticism lately for his announcement that he will forego the usual daily intelligence briefing once in office, instead delegating the task to his VP Mike Pence. Trump's stated rationale is that he's "like, a really smart person", and listening to the same basic information day after day is a waste of scarce time.

This is certainly a break with past administrations, but is it really such an Awful move that we should get worked up over it? I don't think so, and here's why.

The vast majority of the time, daily news events are unimportant. Most things that receive reporting attention never go on to matter in a huge way. If you only checked in with current events once a week instead of once a day (or multiple times a day), you probably wouldn't miss much. Classified intelligence is obviously not the same as a daily newspaper, but it inevitably shares this signal-to-noise problem.

Figuring out in advance what's important and what's not consumes the collective daily brainpower of millions of journalists and analysts--and yet there's still no reliable alternative to simply "wait and see". Institutions like daily newspapers and daily briefings are essentially mechanisms to cope with this limitation: by regularizing search around an arbitrary criteria ("things that happened yesterday"), you hopefully get a set of unbiased data that is very likely to include facts that will go on to become important. Even though you don't know which facts are useful, you can be confident that you've captured useful facts.

So what does this have to do with Trump? My theory is that he's deeply internalized this signal/noise problem and understands how it can interact with negativity bias to produce bad decisionmaking outcomes. Obviously.


Skipping daily briefings is meta-rational

It's been widely reported that Trump's beliefs and preferences are highly dependent on whatever information he's consumed most recently. It's also self-evident that Trump is not psychologically immune to negativity bias--that is, he tends to prioritize negative information over positive information of an equivalent magnitude. Being asymmetrically sensitive to recent negative data in a job where you're exposed to massive amounts of negative data that are also unimportant seems like a recipe for overreaction.

By having VP Pence do the daily briefings, Trump is merely establishing another gatekeeper institution to tighten his information filter; it's a quantitative shift instead of a qualitative one. President Obama gets a briefing once a day instead of two, three, ten times a day for fundamentally the same reason: tradeoffs.

In a certain sense, daily intelligence briefings function like insurance: pay a little cost every day in terms of time and attention, so that when crises happens you kinda sorta know what's going on. As the filter tightens (fewer briefings, shorter etc), you'll save on daily costs but must devote more time getting up-to-speed on the actual crisis days.

Whatever you think about the magnitudes of these costs and benefits, it's not obviously true that a 15-minute daily briefing totally nails the optimal balance of baseline knowledge and time costs. It's complicated! Maybe for Trump the benefits of reducing overreaction risk are worth the loss of control. Maybe he's not so good at maintaining baseline knowledge but is stronger at rapid assimilation (i.e. the conversion of time costs to knowledge are more efficiently allocated towards day-of cramming vs. anticipatory study).

Perhaps Trump has concluded that the Presidential Daily Briefing is more about media play and political risk-hedging than real decisionmaking utility: if something awful happens, the president wants to be able to say they were prepared and not caught off-guard.

I don't know if this shift is a good idea or not, but given Trump's historically atypical cognitive profile, it seems only natural that some administrative structures should change to better-suit his needs and optimize decisionmaking outcomes.

Tweet

Hanson on long-run survival

Economist and futurist Robin Hanson makes an important point that seems widely understood yet rarely discussed:

It is easy to assume that what is good for you is good overall. If you are an artist, you may assume the world is better when consumers more art. If you are a scientist, you may assume the world is better if it gives more attention and funding to science. Similarly, it is easy to assume that the world gets better if more of us get more of what we want, and thus move higher into Maslow’s Hierarchy. 

But I worry: as we attend more to higher needs, we may grow and innovate less regarding lower needs. Can the universe really get filled by creatures focused mainly on self-actualization? Why should they risk or tolerate disruptions from innovations that advance low needs if they don’t care much for that stuff? And many today see their higher needs as conflicting with more capacity to fill low needs. For example, many see more physical capacities as coming at the expense of less nature, weaker indigenous cultures, larger more soul-crushing organizations, more dehumanizing capitalism, etc. Rich nations today do seem to have weaker growth in raw physical capacities because of such issues.

On a species-wide level, we are not at all close to achieving the sort of long-run existence that everyone should wish for. Our position is precarious and entirely dependent on the integrity and productivity of one planet. Nudging our collective resources towards a greater emphasis on space exploration and technological innovation is a really good idea. To the extent that reducing existential risks like climate change, pandemic and asteroids help to maximize the probability of long-run survival in space, we should do more. To the extent that economic growth and poverty reduction increases the number of researchers working on these problems, it's valuable. To the extent that institutions, norms and beliefs geared around individual self-actualization increase technological progress and ensure the structural stability necessary to continue the work, we must defend their integrity.

Tweet

Information Nexus

1. Two very good pieces on the cost disease here and here

2. Accurate take on Atlanta's suburban poverty and mass transit nightmare

3. A good summary of what political pundits have been arguing about over the last 6 months plus this detailed analysis of Tom Hanks

4. Hillary will expand our space potential

5. Interesting long interview about climate change (via)

Tweet

Self-driving tugs will change airports

Source: planitplus.net

As I've mentioned before, self-driving cars have been analyzed and forecasted to death, which is why I'm always on the lookout for curious or niche angles to the technology. TechCrunch today reports on a new company that seeks to deploy self-driving pallets in factory settings:

They’re not exactly Teslas or Google self-driving cars. Instead, Clearpath makes vehicles that autonomously move boxes and pallets around factories, warehouses and distribution centers.

When it comes to self-driving car development, reducing complexity is key. Google and company are making huge strides in developing passenger vehicles that can operate in all-purpose, chaotic street environments, but a simpler, cheaper development path runs through places with highly-controlled environments: factories, warehouses, and airports.

Fitting self-driving cars into narrow tasks within existing industrial processes is a no-brainer, especially considering that employees can be trained to safely interact with them (as opposed to random pedestrians and other motorists). Deployment of these technologies is already expanding rapidly.

The article does mention applications for airport shuttles, but I think the real opportunity in aviation is for ramp operations tugs.


Baggage handlers drive a lot

At a typical airport, the 'ramp agent' job involves a variety of tasks, many of which involve mechanistic driving that is tailor-made for substitution by robot. In small airports, ramp agents may informally switch between dozens of tasks in the course of a single "turn" (aircraft arrival to aircraft departure). In larger hubs, scale efficiencies kick in and tasks become highly specialized, with a single employee spending their entire shift doing one thing.

When an aircraft arrives, ramp agents unload bags into empty carts, which are then theoretically carried off to their proper destinations. Most bags will be deposited at a centralized conveyor belt, with some being transported directly to another airplane (for those quick transfers). At least 50-60% of the labor-hours involved with this arrival phase involve driving from place to place along highly-controlled routes. This is perfect for robot tugs. Employees will load up a cart with arrived baggage, type in the desired locations, and off it goes.

After an aircraft's arrival, it prepares to depart. Some people work in centralized 'bag rooms', where they load passenger luggage and cargo onto carts. At least 20-30% of these labor-hours are devoted solely to driving luggage from the bag room to departing aircraft. These driving routes are also highly controlled, so it's not at all difficult to imagine this task being totally automated. Bag room employees load up a robot tug, type in the gate number, and off it goes.

Another set of tasks involves loading bags into the cargo bins of the aircraft themselves. The driving here tends to be short and chaotic, and might include things like positioning bag carts and other equipment. Most of these driving tasks are not routinized enough to accept robots anytime soon, but even still perhaps 5-10% of the labor-hours could be automated. After successfully loading luggage into the aircraft bins, robot tugs would remove empty carts and return them to the bag room, freeing up valuable employee time to assist with gate-check luggage and perform final safety and security checks.

The final big task is pushing a departing aircraft off of the gate and into the taxiway area. The ability to automate this with robot tugs is highly conditional on the physical design of airports themselves; most newer airports are spacious enough to eliminate complex maneuvers and collision risk, while those located in dense cities tend to be size-constrained and architecturally dysfunctional. Even still, the economic benefit to automating here transcends simple labor cost savings: in most airports, tugs are disconnected after pushback, with aircraft using their own engines to power and navigate the taxiways en route to the runway. Using jet fuel in this way costs billions upon billions of dollars every year, and if the US were ever to implement a carbon pricing scheme it would almost certainly speed up the deployment of automated (or pilot-controlled) pushback tugs by tipping the cost-benefit logic of self-powered taxiing.

In general, I would estimate that even minor cost reductions and technology improvements in self-driving tugs would expose probably 20-30% of total ramp agent labor-hours to immediate substitution. This is not revolutionary, of course, and will probably happen gradually without explicit labor force reductions (unless bag-stacking robot arms are added to the mix...).

Perhaps the productivity gains will result in higher wages for these jobs, which overwhelmingly draw from low-skill prime-age men, a demographic group that has struggled mightily in recent decades. Certainly airports will remain huge sources of decent service-sector employment for quite some time.

Airlines are highly regulated and notoriously penny-pinching, so it's difficult so say if the adoption of self-driving tugs will be constrained or stymied in some way. But the economic logic seems unstoppable.

Although the US is probably on the cusp of an unavoidable wave of airport refurbishments (some major terminal expansions are already underway), if a second round of massive infrastructure building moves to production before self-driving tug systems become viable, post hoc integration might create additional disincentives to adoption. Newer airports in Eurasia and Africa might well end up being built from the ground-up with self-driving tugs in mind.

Tweet

Information Nexus

1. Chuck Marohn on the surprising new White House land-use toolkit and a critical follow-up

2. Transhumanist Party presidential candidate Zoltan Istvan has officially become a write-in effort

3. Excellent tutorial on Bayes' Theorem

4. The FCC is seeking to regulate TV apps why?

5. This graph of music industry revenue sources is great

6. Why are so many hard scientists into extreme endurance sports?

Tweet

The Atlanta Beltline's affordable housing mission creep

The Beltline, a rails-to-trails network of bike paths that will eventually grow to encircle all of central Atlanta, is a glorious thing. The city of Atlanta is entirely devoid of physical geography--no ocean, no lakes, no mountains, no rivers--and as such is starved for natural car-free pedestrian corridors.

Its built environment was developed over decades to be car-centric and spread-out, and only recently has the broader national trend towards urban, walkable cities reached Atlanta and heightened the political urgency surrounding policy and infrastructure changes.

Part of the Beltline's success lies in its incrementalism: instead of trying to build the entire network as a finished project, it started in sections and with few bells and whistles. Over time, its wild success has provided essential feedback to developers and planners, and you now see tweaks to the original vision such as lighting, access points evolved from desire paths, and targeted business investment.

But the very fact that the Beltline continues to adapt and change means that its management is being constantly barraged with choices and demands from various stakeholders.

Recent controversy over affordable housing along the Beltline illustrates this dilemma well, and reveals a worrying trend towards an ever-expanding set of social considerations, which subtly undermines the core strength of the bike path itself.

Next City has a decent rundown of the situation, which seems to involve a healthy dose of office politics, but is fundamentally driven by the notion that The Beltline organization should play a large role in pushing affordable housing development in concert with construction of the sidewalk itself.

From a purely explanatory perspective, it makes total sense that some key founders' have personal commitments to affordable housing and other social justice causes, and see the Beltline as a clever way to further these goals. After all, nobody really expected the project to explode the way it did, and its success has certainly had a positive impact on many social justice measures.

But just because the Beltline has acquired a social justicey identity and brand doesn't mean it should adopt an explicit social justice mandate in its funding and management priorities. To mix metaphors horribly, Atlanta should be cautious when milking the goose that lays the golden egg.

This is considered revolutionary in Atlanta Source: www.wsj.com

Although home prices in the Atlanta metro region are lower than comparable cities around the country, its walkable urban core is rapidly becoming unaffordable. Like most cities in the US, densification and housing construction in desirable areas is constrained due to an overabundance of land-use regulations like zoning, parking minimums, setback requirements, and many others.

That said, Atlanta has added a decent amount of compact units in recent years, especially along the current and future Beltline route, where demand is booming. For a certain set of urbanists this building frenzy presages a classic gentrification dilemma, where poor people get displaced, retail shifts upscale, and the Beltline's amenity value flows primarily to a rich set of yuppies, locking out the original residents.

This scenario is certainly plausible. The question, however, is what to do about it.

Some urbanists see the Beltline boom and worry that its benefits won't accrue to poor people. Their solution is to push artificially cheap housing to ensure some minimally equitable distribution of access. Even if this results in a bit less overall housing supply--a deadweight loss--the social justice benefits and virtues of mixed-income neighborhoods make it an acceptable cost to bear. I draw a slightly different conclusion.

When I look at the Beltline and the rising house prices along it, I think to myself, "let's build more Beltline!"

Poor people in Atlanta are disproportionately harmed by the near complete lack of walkability and non-car transportation options, and the Beltline is a small step towards correcting this sorry state of affairs. Throwing a few affordable housing units up along the only car-free corridor in the city is wildly incommensurate with the scale of the problem, and by discouraging market-rate construction in literally the hottest neighborhoods it might even reduce elite political support for expanding Atlanta's network of bike paths.

Even if walkability isn't one's primary focus, affordable housing along the Beltline seems like an odd strategy. By maximizing market-rate construction along the Beltline, Atlanta would probably reduce price pressure in other, slightly-less-desirable neighborhoods that would otherwise be at risk of gentrification.

In addition, the abstract moral logic of affordable housing along the Beltline is somewhat weak. For most goods, we generally accept that markets are the best and most efficient way of allocating resources, provided the participants have a minimal amount of money and bargaining power. Decisions about price-quality tradeoffs are deeply personal and are usually best made with a reliance on local knowledge. Affordable housing, by regulating and/or subsidizing developers and construction institutions, throws this wisdom out the window.

Instead of heavily regulating and distorting the housing market in an attempt to give poor people what we think they might want, why not just give poor people money and let them decide where to live and what amenities they're willing to pay for? Maybe some would choose to live in expensive condos along the Beltline, but I sort of doubt it. What does that say about the real preferences of poor people?

The Beltline is literally the most awesome urban amenity in Atlanta, in part because of its absurd scarcity. Doesn't it seem totally correct that houses along it should be crazy expensive? Why in the world would you want to build affordable housing there, of all places? Considering the opportunity cost compared to market-rate housing, building along the Beltline is probably the least efficient, most costly place to put it. Amenities like the Beltline are great, and it's totally understandable that higher-class paternalists want poor people to live the dream. But you know what's also great, especially if you're poor? Money. Amenities like the Beltline can't lift someone out of poverty. You can't choose to save up the Beltline for a few years in order to pay for college, or a medical operation, or an upgrade to your small business facade. You can't skimp on the Beltline in order to splurge on a nice meal or a new computer. Forcing poor people to accept public assistance via the convoluted mechanism of housing regulation and subsidies fundamentally reduces their options and freedom to make hard choices.

Of course there is probably some intrinsic and instrumental value in having mixed-income neighborhoods. But forcing this outcome via government regulation is not a free lunch, and I'm thus far unconvinced that those benefits outweigh the overall costs. The Beltline organization should stick to its core mission and work tirelessly to expand non-car transportation options throughout the city. Affordable housing is a political, ethical, and policy minefield, and I fear that adding this constraint to its development ethos will undermine the Beltline's simplicity, incrementalism and consensus-oriented approach.

Tweet

Musk: colonizing Mars will require both idealism and realism

At an annual gathering of the International Astronautical Conference, SpaceX CEO Elon Musk described in detail his plan to develop a cheap transportation system to enable sustainable colonization of Mars.

Although the engineering aspects are likely to garner the most attention, followed by the ethical motivations of existential risk mitigation and exploration virtue, I found the event's economic and business framing to be the most compelling part. If you have the time, it's worth watching the event in full (starts at t=22m):

In general, space stuff is dominated by a small group of intensely committed fans, mostly drawn from math, science and engineering backgrounds. Although an increasing number of space enthusiasts are arriving via philosophical justifications about existential risk and species resilience, the field is still mostly focused on the technical aspects of rocket design.

That certainly makes sense: there are hard scientific and technical hurdles that must be overcome to truly achieve sustainable space living. Hardcore space geeks, whether technical or philosophical, play a crucially important role in keeping the momentum going. Musk explicitly described how space progress had stagnated prior to SpaceX filling the void, a feat made possible only because of Musk's ethical fanaticism. He expressed wariness over personal participation in space travel, for fear that his death would result in a corrupted SpaceX in hock to profit-maximizing investors.

So clearly the intensity and idealism associated with space progress is a necessary and special part of maintaining our current trajectory.

But the overarching theme to Musk's talk wasn't about engineering or philosophy. It was about economics and business.

Musk understands that to overcome the engineering challenges and accomplish ethical goals, you need resources--lots of them. This means talking about business logic: consumer demand, cost of production, price elasticity, investment and payoff schedules, diverse revenue streams, access to talent, integration with government infrastructure, etc.

These aspects of space often get lost in popular media coverage, in part because of the strongly idealistic and inaccessible cultural norms of most space people. To really take off, space companies need to attract and incorporate stakeholders who are totally uninterested in all the geeky intellectual shit. Market forces largely allocate resources and shift incentives, and so they both constrain the development of space economies and enable them.

Musk is helping to reintroduce space as merely another high-tech growth industry, as opposed to a geeky futurist curiosity. Paradoxically, by stressing the practical business realities of space colonization, he makes the enterprise seem more realistic, grounded and achievable while also revealing how frustratingly slow, incremental and difficult it will be.

Tweet

Self-driving cars and the auto service industry

On the heels of a big new regulatory rollout concerning the development and deployment of self-driving cars, Conor Sen makes some good observations about the technology's likely effects on urban life:

The final battle, should the first two problems be solved, will be about how to repurpose street parking and garages, especially in single-family residential neighborhoods. With street parking, some will argue that streets should be narrowed and sidewalks and paths widened, allowing for a better pedestrian experience and more cycling. Others will argue for increasing the percentage of land that can be built upon, at which point the nimby activists will rise up to shout "not in my backyard."

Self-driving cars present a fascinating and diverse set of possibilities, and informed thinkers have sliced and diced the issue from nearly every angle: urban planning and design; public safety; social justice (poverty, age, disability); lifestyle/time management; real income effects; business creation; labor market effects.

Discussion about labor market disruption caused by self-driving cars is politically salient, but often fairly cursory: truckers, bus drivers, delivery drivers and cabbies are screwed, as well as many industrial workers in factories and airports. Technological unemployment of this sort is historically mundane--productivity basically grows by substituting labor for capital, and presumably the losers from self-driving cars will on-net be offset by the (numerically greater) winners. Society will be better off.

But self-driving cars worry some people because the pace of change is likely to be rapid, thus limiting the ability for commercial drivers to re-skill in other fields. Labor effects will also disproportionately hit prime-age men, who are already experiencing record unemployment and detachment from the labor market. The contours of self-driving car labor effects align perfectly with the prevailing narrative about rich-world economic growth: high-skill, intellectual work is rewarded, while jobs requiring minimal education (like driving) are being eaten by AI.

As a speculative exercise, let's think about the possible second-order effects of self-driving car labor disruption on the (large) auto service industry. Innovation scholar Brent Skorup asks the question:

Seems possible that driverless cars, since they'll often refuel when passenger-less, will resurrect the gas station attendant position.

— Brent Skorup (@bskorup) September 19, 2016

He later speculates that all car service functions--fuel, cleaning, mechanical--might consolidate into a "one stop shop" business model.

This does seem inevitable if self-driving cars enable a shift from individual vehicle ownership to a 'service and flow' economy. It also seems inevitable that even if total auto service employment increases (possible if low costs bring low prices, leading to a demand-response resulting in higher overall car volume), the industry will change in important ways. Whether these changes end up being simply disruptive--or truly disastrous--for current workers will largely depend on the choices of car manufacturers and policymakers.

It seems clear that the new auto service industry has the potential to become highly automated and skill-intensive. This would involve innovations like automated fuel delivery systems, for example, and probably would be awful for current low-skill industry workers. The 'fuel attendant' jobs of the future might very well involve sitting behind a computer and managing robots.

On the other hand, low-skill auto service jobs might survive if industry consolidation leads to productivity increases in labor-intensive tasks that are sufficient to outcompete capital substitution. In other words, if Tesla or Uber opens up a huge maintenance center serving an entire city or region, the scale efficiencies might just make the old-fashioned 'gas station attendant' jobs economically rational again.

Paradoxically, if the industry doesn't consolidate and remains oriented around small service shops it might be even worse for low-skill workers, as they won't be able to leverage scale efficiencies to make their labor competitive in the face of robotic servicing.

The cultural power of cars in the US is mighty, however, and it's highly plausible that small business workers will be motivated to upskill in order to survive. Politicians might also be inclined to pass regulation limiting carmakers' ability to set up huge service centers. Additionally, if regulation and manufacturer choices do prevent huge consolidation, it's possible that service automation is too expensive and/or limited for small businesses to deploy. In this scenario, life would be pretty similar to the status quo, with low-skill workers able to survive with minimal upskilling.

Tweet

Liberaltarianism--it's awesome!

For quite some time now, an increasing number of policy experts and thinkers have been quietly converging on a loose set of attitudes about the role of governments and markets in society. I call it 'liberaltarianism', but 'left-libertarian', 'radical centrism', and policy analysis emphasizing 'kludgeocracy', 'political decay', 'upward redistribution' and 'rentier capitalism' are all basically different frames for the same approach: free markets combined with a generous welfare state.

The places that hew most closely to liberaltarianism are (paradoxically) Scandinavian countries typically viewed as big-government socialist havens. In fact they consistently top rankings of market freedom, even among solidly libertarian organizations.

The general political-philosophical orientation has mercifully avoided being labelled and categorized (I'm probably not helping), and at this point there's not really a clear definition or set of specific policy prescriptions associated with it. Some might falsely view it as a re-branding of Clinton-era New Democrat 'triangulation', but in many ways liberaltarianism is the precise opposite.

Where centrist Democrats pursued business-friendly policies like semi-privatization and narrow tax credits (designed to reduce the nominal size of government), liberaltarians recognize that the ultimate result of these 'submerged state' policies often increases market distortions and opportunities for rent-seeking with a swamp of regulations and loopholes. The current Democratic Party has improved in some ways, but seems increasingly enthusiastic about the direct regulation of markets, which is perhaps even worse.

Still, the intellectual bankruptcy of the Republican Party--lost to business interests and white ethnonationalists--means that in the near term any political traction liberaltarian ideas receive will probably come from leftists and maybe a few pragmatic libertarians.

To help flesh-out the liberaltarian framework as I understand it, here are a few links from authors I consider to be pushing policy in a positive direction:

-------

Will Wilkinson was one of the earliest exponents of liberaltarianism, often critiquing mainstream libertarians for eschewing social justice topics and the importance of redistribution in maintaining political support for free markets. Initially a blogger at The Economist, he now directs policy at the Niskanen Center and writes occasionally for Vox. Here is his Libertarian Case for Bernie Sanders, and here is an excellent Vox column articulating some key aspects of his idiosyncratic policy vision:

Free markets require the presence of good regulation, which define and protect property rights and facilitate market processes through the consistent application of clear law, and an absence of bad regulation, which interferes with productive economic activity. A government can tax and spend very little — yet still stomp all over markets. Conversely, a government can withdraw lots of money from the economy through taxes, but still totally nail the optimal balance of good and bad regulation. 

Whether a country’s market economy is free — open, competitive, and relatively unmolested by government — is more a question of regulation than a question of taxation and redistribution. It’s not primarily about how "big" its government is.

Scott Alexander is a freakishly talented internet blogger person, with roots in Less Wrong and Bay Area techno-libertarian culture. He is very explicit about his 'left-libertariansm', which he has articulated at a slow drip over the years. This is from a post on the subject:

What if we abandon our tribe’s custom of conflating free market values and unconcern about social welfare? 

Right now some people label themselves “capitalists”. They support free markets and oppose the social safety net. Other people call themselves “socialists”. They oppose free markets and support the social safety net. But there are two more possibilities to fill in there. 

Some people might oppose both free markets and a social safety net. I don’t know if there’s a name for this philosophy, but it sounds kind of like fascism – government-controlled corporations running the economy for the good of the strong. 

Others might support both free markets and a social safety net. You could call them “welfare capitalists”. I ran a Google search and some of them seem to call themselves “bleeding heart libertarians“. I would call them “correct”.

Steven Teles is a political scientist who has done much to further a certain type of institutional analysis that brings old-school public choice theory into the modern age. His reframed 'kludgeocracy' concept has made certain libertarian critiques of government more palatable to cultural leftists. This is from his classic Kludgeocracy in America essay:

Housing is perhaps the most striking and perverse example of this pattern of government growth through seemingly non-governmental means. The 30-year, fixed interest-rate mortgage exists on a mass scale only in the United States, and only because of massive distortions of the free market by government-sponsored entities like Fannie Mae. Added on top of that are the deduction of mortgage interest from taxable income — the third-largest exclusion in the tax code — and the delay in capital-gains taxation on home sales when another home is purchased. Taken together, the tax code and government-sponsored enterprises amount to a massive housing-welfare state. And although it delivers benefits to many citizens, this set of programs is fundamentally regressive — vastly favoring people in the highest tax brackets and artificially increasing the prices of homes, thus increasing barriers for first-time home buyers.

His essay The Scourge of Upward Redistribution is also recommended.


Francis Fukuyama is another political scientist whose focus has gradually drifted towards an analysis of US policy dysfunction that closely mirrors the libertarian critique of government, yet remains broadly amenable to welfare redistribution. This is an emerging theme among liberaltarians: decoupling the best libertarian critiques of government from other issue domains like welfare and fiscal policy, monetary policy and foreign policy.

Here is an excerpt from Fukuyama's essay The Decay of American Political Institutions:

The decay in the quality of American government has to do directly with the American penchant for a state of “courts and parties”, which has returned to center stage in the past fifty years. The courts and legislature have increasingly usurped many of the proper functions of the executive, making the operation of the government as a whole both incoherent and inefficient. The steadily increasing judicialization of functions that in other developed democracies are handled by administrative bureaucracies has led to an explosion of costly litigation, slow decision-making and highly inconsistent enforcement of laws. The courts, instead of being constraints on government, have become alternative instruments for the expansion of government. Ironically, out of a fear of empowering “big government”, the United States has ended up with a government that is very large, but that is actually less accountable because it is largely in the hands of unelected courts.

Ashwin Parameswaran blogs at Macroresilience, and has written about liberaltarianism with great clarity. A few choice quotes from his essay Radical Centrism: Uniting the Radical Right and the Radical Left (separated by paragraphs)

Neoliberal crony capitalism is driven by a grand coalition between the pragmatic centre-left and the pragmatic centre-right. Crony capitalist policies are always justified as the pragmatic solution. The range of policy options is narrowed down to a pragmatic compromise that maximises the rent that can be extracted by special interests. Instead of the government providing essential services such as healthcare and law and order, we get oligopolistic private healthcare and privatised prisons. Instead of a vibrant and competitive private sector with free entry and exit of firms we get heavily regulated and licensed industries, too-big-to-fail banks and corporate bailouts.

The Core Strategy of Pragmatic Crony Capitalism: Increase The Scope and Reduce the Scale of Government.

 The essence of a radical centrist approach is government provision of essential goods and services and a minimal-intervention, free enterprise environment for everything else.

 A bimodal strategy of combining a conservative core with an aggressive periphery is common across complex adaptive systems in many different domains.

Radical centrism involves a strengthening of the safety net for individuals combined with a dramatic increase in the competitive pressures exerted on incumbent firms. Today, we bail out banks because a banking collapse threatens the integrity of the financial system. We bail out incumbent firms because firm failure leaves the unemployed without even catastrophic health insurance. The principle of radical centrism aims to build a firewall that protects the common man from the worst impact of economic disturbances while simultaneously increasing the threat of failure at firm level. The presence of the ‘public option’ and a robust safety net is precisely what empowers us to allow incumbent firms to fail.

Tweet