Going to repost a comment I originally made on Noah Smith's blog, because it spells out part of the approach I believe we should be taking wrt China:
I've always gotten the vibe Chinese leadership in general fears a brain drain to the the West (especially US). So, when Washington's discriminatory policies make it harder for talented Chinese researchers (and well-educated people in general) to come to the United States, Xi and the Standing Committee aren't upset. In the least (their public rhetoric notwithstanding). Rather, they're doing high fives.
America should be doing everything in its power to attract the world's best and brightest: more student visas, more green cards, special visa programs for foreign graduates of US colleges, more H1s. All of it. And yes, that should include lots of China's best. STEAL THEIR TALENT. I guarantee many top Chinese brains will LEAP at the chance to trade in Chinese smog, crowds, low wages and totalitarian constraints for the freedom, clean air, potable water, wide open spaces, higher wages, and detached houses of Silicon Valley, Boston or Seattle. This is a complete no-brainer.
(Ok, so maybe not many will be able to buy a house in Palo Alto, at least right away! But it's getting pretty tough in Beijing or Hangzhou, too!)
PS — Completely non-scientific observation here, but, personal experience to me suggests China's educated classes are indeed getting a bit nervous, and there's a general increase in top level talent trying to get out. I've got five well-educated friends here who are finalizing exit strategies: three to Canada, one to the US, and one to Europe.
PPS — Preemptive reply to the inevitable objection based on national security: fears of Chinese espionage are overdone for a variety of reason; our people (FBI) in any event know how to handle this stuff, but if their budget needs to be increased, so be it; cutting ourselves off from an increasingly large portion of the planet's smartest folks isn't going to just help China: it'll also directly weaken US science.
Yes this is a great idea. I find it bitterly ironic that the "America is the greatest place that ever existed" folks are terrified about welcoming new talent, while the people who want to make it easy to come here are aligned with the "America is a racist, sexist hellhole" brigade.
You'd think that if you could get patriotism and immigration support aligned in one party they would be a juggernaut on that issue.
It's deeply ironic that the GOP isn't salivating at the possibility of bringing in a lot of new voters predisposed to hate socialism and big government.
You mean the big government and socialism that has taken them from a multi millenia history of famine and a multi century history of backwardness to the brink of industrial first world status?
That's a good point: some of the Chinese I know defend their government as generally beneficent, and, especially after COVID, they see their approach as better overall. Still, even for them, staying in the US is better for their science, and we should do everything in our power to keep it that way.
This supports a suspicion of mine that for all our flaws, no one has come up with a better model for international prominence/influence/even dominance than the US. The Axis powers fell before the united democracies (+ the USSR); the USSR alternative flopped; our fears of a dominant Japan were overblown -- and we just keep trudging along.
I suspect the same will be for China. Despite its laudable success in bringing hundreds of millions out of poverty, I don't think it has the tools or ideas to become a true global competitor to the US. Its demography is a killer, and unlike the US it doesn't have alternatives like immigration to counter it. It is growing its military power but doesn't have a clue how to pursue effective statecraft (its real superpower seems to be pushing more and more nations to ally against it). Its state-managed capitalism has helped it rise to a middle rank economically but doesn't seem to have any idea how to help it become a truly advanced economic power.
So whatever errors we're making and have made, it looks like the good lord keeps looking after us.(*)
Only objection I have there is that our population was growing through defeating Axis and USSR and Japan given spacial constraints never had a chance. China on the other hand is much bigger and without economic collapse there's no reason to think it wouldn't surpass the US. In other words ONE BILLION AMERICANS!!! :)
It seems obvious, but a simple way to reduce external threats is to become the greatest country in the world and invite as many people as possible to come live here
As a scientist in academia, I just want to say: THANK YOU!!! I have been banging this drum for a while now and I've often felt like the only one who gets it. We need to compete with China on talent. That means making it easier to do science here and to stay here. Current policies do the opposite.
I've worked with plenty of Chinese post-docs. And a lot of them did want to stay in the US. I had a conversation with one of them and he mentioned that he was part of a small ethnic minority in China. He seemed quite happy when he said that in the US it really didn't matter what ethnic group he was from. The US has a racism problem, but the Chinese government's treatment of its ethnic minorities in the past 5 years (or for that matter the past 50 years) is reprehensible and will probably only get worse. Science does best in open societies because science often reveals things that governments don't like. If I were an infectious disease scientist in China right now, I would want to get out soon because there is a good chance that the government would put pressure on infectious disease scientists.
I wholehearted agree. The US beat the English in the revolution with our English. We won the American civil war by beating their Americans with our Americans. We won the world wars by having our Germans, Italians, etc. beat their Germans, Italians, etc. We should look to defeat China (preferably in economic, not military war) by having our Chinese beat their Chinese - just need more Chinese to do it. Also should prepare for any Indian conflict by getting a bunch of them as well...
I mean, thats a nice story...But the higher the percentage of German ancestry in a county was correlated with voting against FDR and for anti WW2 candidates
Sure - but when it came time to fight, build, and grow - they did. In fact, one of our biggest shames was thinking that the Japanese part of Japanese-Americans was more important that the American part.
We see a similar thing today with absolute maniac Cubans in Florida holding the entire country's foreign policy with Cuba hostage. Immigrants often have insane politics when it comes to their country of origin
"We are not going to hit the 1.5-degree or even 2-degree target, but 2.7 is much better than 3, and 2.4 is much better than 2.7 — we need to do what we can."
A good post, but I'm going to quibble with this particular sentence, because I think it leads to bad analysis and bad policy recommendations. True, we most likely can't hit the 1.5-degree target, but we most certainly can hit the 2-degree target. In the 2018-IPCC report (the newest report has roughly the same findings, but I'm a little less familiar with it) the remaining carbon budget of the whole world for the 2C-target with at least 2/3 probability was estimated to be a little above 1000 GtCO2. For 50 % probability, the budget is 1500 GtCO2.
That's between 20 and 35 years of our current global emissions! The carbon budget for the 1.5-degrees target however was just 420 GtCO2, which is roughly ten years of emissions (from 2015!). If we simplify enormously and assume a linear reduction to net zero from today, that means we need to get half way to zero emissions in ten years and all the way to zero in 20 years to limit warming til 1.5-degrees. To limit warming to 2-degrees might require halving emissions in 20-30 years, and then getting to net zero in the next 20-30. Even if emissions plateau for a few years and then start going down, we still have a few decades. Getting to net zero in a few decades certainly going to be difficult, but it is not at all inconceivable.
The climate science fully supports the notion that every 0.1 degrees C counts, like the post says. So stabilising the warming at 1.9, 1.8 or even 1.7 degrees instead of 2 degrees matters a lot. This perspective indicates that speeding up the transition to lower emissions is hugely important. And yes, the international negotiations do matter. If Chinas emissions peak a few years earlier and then falls gradually instead of plateauing, that matters a lot. International pressure, negotiations, but also technological and financial developments do affect this.
The longer time horizon of the 2-degree goal also means that technologies that take longer to mature can play a larger role. In climate circles many people say that advanced nuclear is not important, because it won't be ready and deployed at scale by 2030. But it might be technologically ready by the late 2020s and deployed at scale gradually through the 2030s. Furthermore, if that provides energy abundance, that could make some of the hardest climate problems (such as industrial heat, or heavy transport and air travel using synthetic fuels or hydrogen-based fuels) easier to achieve. The same applies to other technologies.
The current focus on the 1.5-degree target means that we need drastic action right now, or we fail. I would certainly support drastic action right now, but that's not the only option. Accepting 2.7 or 2.4 degree warming as the best we can do, means we can just patiently wait for technologies to mature. Aiming for limiting to at most 2, means that we need to act now and that there are plenty of politically realistic things we can do.
How about, uh, the Republican Party radicalizing against democracy? Was a bit amazed to see out-there stuff like 'AI takeover' mentioned in the piece, but not the fact that one of our two major parties is turning into some kind of Peronist/North Korean-hybrid personality cult.
If we had a more logical electoral system, a lot of Republican craziness would be for naught. But we have a very specific failure point in the ridiculous Electoral College, where Constitutionally there is at least some play for states to send electors based on whatever criteria they please. I have seen a number of thoughtful commenters, including conservatives, note that Wisconsin, Pennsylvania, Arizona, Georgia and others could send the 'wrong' electors just based on allegations of voter fraud. Even if the court system sorted it out later, the loss of perceived political stability to the US would be permanent, permanent damage. We could divide American history into before and after. And that's without rioting, supporters clashing in the streets, shootings, etc.
BTW, Trumpist hysteria & death threats against Republican legislators who voted for the infrastructure bill kinda debunks MY's pieces about how the R's are becoming more politically moderate (while procedurally radical). I don't expect them to be loudly for it, but being (literally) violently against even bridges & roads is a pretty ominous sign
The hostility toward R legislators who voted for the infrastructure bill is precisely procedurally radical: they would have happily supported the same bill if Trump had proposed it. It's an example of denying the opposition party the right to govern.
First, I will put the obligatory disclaimer that I don't believe Trump's election lies.
That being said, I think the Dems election lies are only slightly less troubling (there was no voter suppression in 2018, or 2020, there were record turn outs).
Worse, is the HR1 bill that seems determined to make fraud easier.
Without faith in free and fair elections democracy can't work. We need faith in elections, and partisan's need to stop spreading lies and conspiracy theories.
Moreover, we are going to need to agree on a basic framework. That's going to be something like several weeks of mainly in person voting, but also requiring voter ID
Honestly I think this whole argument goes to Matt's point about arguing about symbolic issues. Dems talk much more about the symbolic election stuff when really what matters is gerrymandering. That said while democratic aren't perfectly honest about elections stuff, they are not fueling outright conspiracy and at times encouraging violence.
The first is that people believe the lies that hucksters tell and give these systems more control in safety-critical situations for which they cannot safely perform (see: everyone thinking Tesla’s autopilot features are actually sufficient to drive for them).
The second is that people use algorithms to bias-launder. Doing things like releasing resume-scanning systems that somehow always reject non-white applicants (likely because it’s training data was from compiled by racially-biased people) yet claiming that the system cannot be biased because it is a computer and cannot harbor personal racial hated.
The former will likely go away when more people have more hands-on experience and recognize how much crap like “autopilot” is janky as all hell. The later is a much bigger and more imminent threat. But what is not a threat is any kind of singularity/SkyNet situation where strong AI comes into being. When people were resigning in protest from Google/Alphabet’s AI ethics group, it wasn’t because the researchers were concerned about an evil strong AI that Google refused to stop creating. It was because of the bias-laundering issue.
At the risk of sounding like some kind of AI alarmist, I can't help but comment that to say that there are definitely only two threats from AI and that they are fairly pedestrian as far as threats go, strikes me as being extremely overconfident. I don't think you have to get fully onboard the godlike super-intelligence singularity train either before accepting there are other genuine reasons for concern (automation, surveillance technology, AI weapons systems, etc., just to name a few) if we look beyond the immediate timeframe of the next few years. I don't see reason for the skepticism about the fact that there is a threat itself, although as was pointed out in the post, the position with actual AI safety efforts is pretty bleak right now.
I'll believe AI is really a thing when it passes my own version of the Turing test: write its own really killer standup comedy routine and deliver it (well, from behind a screen of some type), and not have anyone realize it wasn't done by a human.
I think automation doesn't fit as an AI threat for the same reason that the robotic automation of the 80s wasn't a threat. There may be temporary short-term losses in areas that are automated, but just like in the 80s we find that those jobs move into other sectors and the overall increase in economic productivity is a net-positive. (Also, if the automation you're worried about is something like all the truckers being replaced with self-driving cars, that's a fairy tale. Level 5 driving automation effectively requires Strong AI, so we're gonna see SkyNet before we see that. And level 4 driving automation would require someone behind the wheel, so the jobs aren't going anywhere.)
As far as surveillance technology, I think the threat there is far more from the actual capture tech rather than any machine-learning models they have running behind it. A million ring cameras installed in every neighborhood are the potential problem. Whether there is a human combing through the footage or a script is less of a concern (especially because a human is a lot more likely to clip the footage and tweet out some random embarrassing moment caught on camera).
And in regards to AI weapon systems, I think that falls under the first threat: Safety-critical system being handled by computers that aren't up to the task. But I also think that weapon systems are less of a concern, because in general military officers are less likely to have the "every mistake is just helping them make the next iteration better" attitude you see from Tesla drivers commenting on every video of an autopilot crash.
So, I think you are kind of underrating the risk from misaligned AI. You don't need a strong or even a general AI to cause an alignment crisis. You just need two things:
1) An AI with the power to do things without a human intermediary
2) Some goal for the AI to maximize
It doesn't have to be smart, evil, or anything else. Because a human intermediary is extremely slow, AI will inevitably be given the ability to directly affect the real world (in fact, it already has in financial transactions). It doesn't have to be an existential threat to be very bad. Example: profit-maximizing corporate AI notices enforcement patterns for some major regulation, determines that if it breaks the law in a specific way it will most likely never be caught.
Secondarily, the other huge threat from AI (one that is far worse than bias), is what it will enable regular ol' humans to do. AI tools will give single individuals with sufficient computing power the ability to have incredible power. This is already what has happened with social media companies, who have gained the ability to tweak various parameters and cause social and cultural change on a whim. Now imagine if that same ability was applied to the real economy of actual atoms via things like automated factories.
You're making one massive error here, assuming that an AI will be able to spontaneously go outside of its parameters in order to continue its maximizing goal. An example I like to use is chess AI. There are now a lot of extremely excellent AI chess playing systems that are able to easily beat all human opponents. However, no matter how optimized a chess-playing AI is, (even if you take the chess-playing AI and pit it against another chess-playing AI and have them continue to optimize against each other), that AI is never going to try and get away with making illegal moves to win. Also, even the most perfectly-optimized chess-playing AI will freeze up and be unable to make a single move if you have it try to play a game of checkers.
And on your second point, there are already real-world examples of people taking the tools and mis-using them. But it is happening today, and the mechanism by which is it happening is through things like bias-laundering. (Also, worry about regular humans causing huge problems via factories is precisely why we have the majority of our current environmental regulations.)
I think you may misunderstanding the very important point being made here. It's not that AI will spontaneously gain control of things on their own, it's the people will give it too much power to begin with, and it will use that power is harmful and unexpected ways.
The biggest example of this is Facebook. A slew of powerful AI algorithms control what people see, driven only by the objective of increasing user "engagement." In practice, the AI basically figured out that the best way to do this is to present the most inflammatory content possible, without regard for accuracy or moderation.
This is by far the biggest risk of AI at the present. Both of your original examples are specific instances of the broader issue Kade presented.
Another way to think of it: A profit-maximizing trading algorithm is still just a trading algorithm. It is going to be limited to its initial parameters of being able to do trades in the specific market where it operates. A poorly-aligned trading AI may cause a flash-crash, but it isn't going to suddenly start issuing credit default swaps.
And be absolutely screwed when environments change, and if it’s up against a real trader on the other side of the book with a bloodlust, a mortgage to pay, and numbers to hit that is taking in every tick on the BBG terminal to his trained senses
Here’s my general take on this. There are two real innovations that are expanding AI usage.
-Increased data storage capacity
-Advances in convex optimization research.
Both of these innovations are slow and steady for years and alone will not bring about this grand alignment issue that kills all humans. They lead to more data, and faster cleaner error minimization on that data.
The illusion of revolution is due to the advent of accessible, easy to use software packages that can make an AI model with very little code or understanding of the problem you’re trying to solve. These are dangerous not because they’re intelligent, but because as Lance has said, they’re trusted, go into production, suck at it, and screw shit up.
This is a threat. And the black box that he mentions does allow companies to do dirty work “bias laundering” in the name of innovation that they just “didn’t realize was causing this problem cause it’s so advanced but we have to innovate!”
The solution to this isn’t a safe harbor for bias but a re-calibration of the word “innovation.” If Wal-Mart suddenly sold space ships and nuclear reactors you wouldn’t call a guy who made a nuclear reactor space ship in his backyard an “inventor”
I think that's a reasonable analysis of _today's_ machine learning, and it's all totally valid. If I'm trolling someone concerned about general intelligence I say things like "what if your machine learning algorithm is programmed to maximize engagement or viewing hours and takes the strategy of teaching people that the world is flat."
I think the concern of the AI researchers who are looking at long term existential threat is that we'll wake up one day and have a machine learning system where we can't say where the boundaries are or that it is siloed to its domain or that these clear bright line constraints on capability are enforced.
Humans screw up computer systems _all the time_. We try hard to air-gap systems and we don't always succeed - adversaries take advantage of that now. The researchers fear that an AI that looks for ways to not be contained will be a lot better at finding them than today's human hackers.
Zillow’s news cycle for ruining 1/4 of their company’s jobs by putting too much faith in a crap model is still ongoing and here we have a “but these models might take over humanity” take. I love MattY but this was absurd.
It's surprising to me that someone who wrote preciently about the structural threats to American democracy would not consider that to be one of the important issues facing us.
It's funny you should mention that, because as I was reading it, I couldn't help thinking that the themes - it matters more what you do than participating in international talking shops, American government is short-termist, there's no unity and so much division, it's hard to pass legislation to achieve necessary goals, our politics is not adapting to Great Power Competition - are all themes that get trotted out by people who think American democracy is kind of decadent and that the Chinese have a pretty good way of running things, all told.
Now to be clear, I don't think that *is* Matt's position. I just thought it was interesting that the themes seemed very familiar.
I mean I think that my living standards are lower than they would be if there were fewer homeless people living around me—to say nothing of *their* living standards—and I think a government could solve or mitigate this problem without recourse to authoritarianism.
Yeah I was gonna say, the living standards of slow boring readers is not greatly affected but it really does hurt Americans. Also, while I am sure to one day become part of the American Bourgousie I am currently an indebted college student who also would like to live in a city, and I would really like to have a better government(s) that didn't let housing costs explode because the current bougousie are mostly NIMBYs.
Re what matters, why are the Dems allotting $285 billion - the SECOND LARGEST PROGRAM in BBB - to helping out the well-to-do by lifting the $10,000 cap on state and local taxes (SALT)? I guess this shows that "working families" are less of a priority than they claim.
YES. This makes me SO MAD. I personally will benefit from it -- I did some back of the envelope math back in the year of Trump's tax cut, and found that on net it cost my family about four grand, primarily because of the SALT limitation.
The thing is, I was _completely fine with that_. While I live in an incredibly expensive market, and I live basically an upper-middle-class lifestyle, it is just an objective fact that we are rich. Our net worth cracks seven figures, and our income is around the second or third percentile. We aren't _quite_ The One Percent, but getting close. We should be paying more taxes.
The fact that even progressive Democrats like Katie Porter from SoCal felt they had to go all-in on a tax cut for people like me makes me just despair. I assume she felt like she had to, because so many self-styled SoCal "progressives" are infected with Jarvis brainworms, and believe BS lies like "Prop 13 helps low-income people to afford homeownership". With Democrats like these, who needs Republicans?!
Thanks for your perspective. I don't begrudge people their wealth, so I didn't intend to suggest that the well-to-do are unworthy. But, given a choice between making the child tax credit permanent and offering what I understand to be a temporary increase of the SALT cap, I think our country is better served by reducing child poverty.
Katie "a 1 million dollar house is modest in my district" Porter? I just can't believe these supposed progressives have no shame in their SALT advocacy.
I mean, a million-dollar house may well _be_ quite modest in her district. Certainly where I am it is. I live in a million dollar house. It was built in 1940, it's a 4 bed, 2 bath, single story, with a tiny yard. It's _significantly_ smaller than what I grew up in, with my dad working full-time and my mom having left the workforce to be a homemaker.
But, like, so what? The solution to housing affordability isn't to comfort the comfortable while afflicting the afflicted. We need to make housing more affordable _broadly_, which means reforms to allow a LOT more housing to get built. And quite often, you'll find that exactly the people in my situation -- comfortably situated, but discontent because it's so much HARDER to achieve that comfort than it was for their parents' generation -- just want to pull up the ladder behind them. I've had multiple people tell me "The Bay Area is full, we should stop building."
I agree 100%. Housing cost burden in coastal areas that have pulled up the ladder through regulation and zoning is a huge burden on the productivity of the economy as discussed at length here.
All that being said, the optics of saying a "million dollar house is modest" to a national audience is not great! A person owning said "modest" house in Irvine may not feel rich spending 40% of their take home pay on their mortgage, but by definition their family income puts them well into the 90th percentile. And the appreciation of that house in the last year has definitely made them a big chunk of paper wealth.
Democrats could do a lot to help that person (reduce housing cost through zoning reform, college costs/debt, improve early childhood / school quality, etc. etc.), but like, aren't progressives here to help kids escape poverty? Make a fairer and more just society? Not pad the pockets of the already well off in the wealthiest areas of the country?
Agreed, the optics of rich people saying "but I don't feel rich!" is terrible.
I believe it was Brad DeLong who once paraphrased a whine from somebody on this topic as, "Well of course, it is objectively a very large sum of money, but it is so much smaller _after I have spent it_."
My anger at this is never ending. The bill has ended up going from a lot of stuff I thought was good and exciting to one really good thing that doesn't help dems much politically (climate stuff), a bunch of temporary ~okay~ stuff that will just die in the 2023 majority-Republican house, and a massive awful thing in SALT. Honestly if SALT were part of the BBB of July, maybe I'd let it slide, but at this point it's a Climate bill with some temporary and moderately regressive tax changes.
The SALT limitation raised my federal taxes 60%, not the biggest deal in the world, but pretty eye-opening at the time. I'm sure it did the same for a lot of other taxpayers in high tax states. Not surprising their representatives are responding to complaints, just as they would to higher gas and food prices.
You and I agree. What I was surprised by was how big an impact the limit had on my tax bill. I thought I was under-taxed before, and not complaining now. But a simple change to the tax code that changes your bill 60% makes no sense.
Crumbs??? It's the second largest tax expenditure! I could support their raising the cap a bit, but, as the article shows, only the tippy-top of the income distribution sees a significant tax savings on this, and, for the multi-millionaires, is $15k that big of a deal? People who earn less than $366,000 save ZERO or at most $1,300 per year. Raise the cap to $15k or $20k, but $80k? - no way!
I mean - we may all be getting spun up for nothing - I'm really hoping this SALT crap will get thrown out because it's gross and not even remotely the most important part of the bill.
But if it does end up passing, someone will write "What's the matter with New Jersey" - the democratic party will be a machine to convert progressive outrage at injustice into tax cuts for wealthy property owners in blue states.
In a 2022 red wave, Malinowski, Sherrill and probably even Porter and Gottheimer are going to lose anyway, so spending the biggest part of BBB on SALT to give semi-republicans a tax credit is truly the worst possible policy choice democrats can envision.
I think MY's critique of BBB (tries to do everything poorly, does nothing well) was already pretty damning, but putting SALT deductions in it is gross.
I get where the Biden administration is coming from ("we're comfortable with this crap sandwich because we're not gonna get the votes otherwise" - slight paraphrase by me) even though it's stupid and bankrupt. I'm disappointed in Pelosi for not cracking the whip and killing this thing - it's a bad idea and it undermines democrats among the progressive base and swing state centrist democrats.
For a party that campaigned on raising taxes on the rich as a way of reducing inequality, the existence of the costly SALT proposal in the middle of an ambitious progressive bill sticks out like a sore, hypocritical thumb. The policy is no minor throw-in or pet project; its cost as a tax expenditure represents a significant portion of the overall bill and undercuts Biden’s goal of reducing inequality by shifting the tax burden from the poor and working class toward the wealthy. Jason Furman, a Harvard economist who served as the chair of President Barack Obama’s Council of Economic Advisers, remarked earlier this month that the proposed giveaway for wealthy taxpayers was “obscene.” Households making several million dollars a year “could get a $25,900 tax cut,” wrote Chuck Marr of the Center on Budget and Policy Priorities, a liberal think tank.
Also from that Atlantic article: "The scope of the SALT provision within the overall bill is also enormous; over the first five years of the bill, it is the single largest piece of the proposal, Furman told me, costing more money over that period than the Democrats’ spending on climate change, education, or health care."
Dumb hot take: we can solve the global warming risk and AI existential risk by doing a particularly terrible job at the super power war risk. We'll get into a land war with China over Taiwan and in the combat, the world's major semiconductor fabs will all be destroyed, absolutely devastating the global economy and reducing green house gasses in a super-depression, while also setting back AI research due to lack of computing power.
Pretty soon you'll be talking about the benefits of a nice cool nuclear winter to combat global warming..... Every cloud has a silver lining. Sometimes the lining is even radioactive.
Idea for new "Terminator" reboot -- a time traveler comes back to the present to stop the future AI takeover of Earth but his plan for doing so is to trigger a major US-China nuclear exchange.
"we ultimately need a vision for America that doesn’t rely on hoping the Chinese economy collapses if we want to remain the number one economy in the world."
China will be the world's number one economy unless it collapses. This is just math. China is already the biggest economy in purchasing power terms. If it can achieve the per capita GDP of Greece or Hungary, its economy would be twice as big as ours. Neither of those countries is well run. It's an easy mark to hit.
The good news is relative decline isnt that bad, especially if you chill out. The UK is a good illustration. It was the biggest Western economy in 1850. In the last 170 years, its standing has declined relative to almost every other Western country, yet median British incomes today are 10 times what they were in 1850. Britons are much better off despite all that relative decline.
The biggest peril of relative decline is killing people by overplaying one's hand. It's fine to invest in a navy and nuclear weapons, but Britain has fared poorly every time it has fought great power conflicts on land. Its position declined somewhat during World War One and catastrophically during World War Two. Britain never came close to its original war aim of preserving Poland's independence. It would have done much better to sit out World War Two and let the Nazis and communists duke it out. It could have protected Jewish lives much more effectively by opening India (and possibly the Dominions) to immigration than by fighting a war it had no way of winning. I see a similar dynamic with Taiwan and I dont want my son to die defending the "indepedence" of an island that was part of China for 200 years.
What you say is valid BUT the PRC's demographics (as I'm sure you're aware) are truly dire. It's entirely possible (even likely) US GDP could be growing faster than PRC GDP by the end of this decade. Heck, I wouldn't be shocked if US GDP growth beat's China's in 2021. IOW, China's economy will surely be quite a bit larger than America's in 2028 (in PPP terms, as you point out, it's already a good deal larger), but the US could quite conceivably regain the lead from China in the medium term.*
We just have to stop being stupid on immigration and one or two other issues. Maintaining a favorable demographic profile compared to the PRC should be the easiest of lifts for the United States of America. We don't have to swing for the fences (ie, One Billion Americans). We just have to refrain from making own goals like, erm, watching the UK and Canada beat us in in international college admissions.
*Having the world's largest national economy isn't everything, as you mention. But it's not nothing, either. In any event I fully expect India to emerge as number one in the second half of the century. I believe that, just like the original Cold War, a successful US policy vis a vis China means playing the long game: recognizing that the US has some significant advantages, and acting with prudence and patience. But my fear is US policy makers of both parties seem to think it's in the country's interests to have a major conflict as soon as possible.
Indeed, having a big economy may not mean that much. By some measures, China had the world's biggest economy in the 17th and 18th centuries, and even into the early 19th century (https://en.wikipedia.org/wiki/List_of_regions_by_past_GDP_(PPP)). Really, that just means they had a lot of people. They didn't have the surpluses that come from high GDP per capita to allow them to leverage that gross size into increasing growth or international prominence. Instead, they were just a big, weak and fragile "nation."
You are more pessimistic about Chinese economic growth than I. China’s GDP has exploded despite a stagnant population. It’s growth has been driven by importing technologies developed elsewhere to make its workers more productive. This is a time honored model. It’s what basically all of Latin America and eastern Europe have done. It’s what France and the low countries did in the mid 19th century with British technologies. It’s what India is doing today. This strategy will work until Chinese per capita GDP is a large fraction of the US or until upheaval stops growth.
This model hasn’t worked in Africa or much of the middle east because it requires political stability. The CCP appears pretty solid to me, and established Chinese dynasties have lasted 200-400 years. Is there any reason to think the CCP won’t last as long as the Ming dynasty?
Yeah, I'm definitely more of a pessimist than you re: China's growth prospects.
The thing is, China's population hasn't really *stagnated* through the post Deng/Opening Up years. Since 1980 PRC population has increased by more than 50%. That's about the same as the US. That rate of population growth has slowed, of course, but it's only in recent years that's it's begun to approach Russia/Japan/Italy territory. And sure enough, economic growth, too has slowed. If you believe official statistics, the country's economy has been expanding since 2010 at less than half the rate of the go-go years (in the 6% range) and there's wide skepticism these figures are even accurate. And indeed, China's workforce has begun to shrink in absolute terms. And the country's efforts to exhort women to have more babies would appear to have fallen on deaf ears.
I'm not suggesting China's about to suffer some massive crash, mind you. When your population is shrinking, living standards can still increase nicely even when you're only growing at a modest 2 or 3 percent annually. But yes, I do indeed think such numbers are in China's very near future. The country's technocrats get a lot of things right, but they're not magicians, and they're not going to enjoy developing world growth numbers with Japan's demographics. (I also think there's growing evidence Xi isn't particularly competent on economic matters, so there's that, too).
Since 1980, Chinese GDP has grown by -1200% while population has grown by 43%. Calling China’s population stagnant was inartful though, as you note, population growth has slowed markedly since 1990. Still, population growth is only very loosely correlated with GDP growth. Most GDP growth has had little to do with more people.
China’s urban population, which has surged, is likely a better predictor of GDP. Peasants contribute very little to GDP. Indeed US agriculture is still radically more efficient than China’s.
>>> I see a similar dynamic with Taiwan and I dont want my son to die defending the "indepedence" of an island that was part of China for 200 years.<<<
The world would be miraculously lucky if an actual shooting war between the US and the PRC were confined solely to the battlefield deaths of soldiers and did not encompass (as seems terrifyingly likely) the mass death of civilians caused by nuclear weapons.
Which is why I think Taiwan hawks are a huge threat to the world. I’m all for self determination and I wish the CCP were too. But I care much more about peace than political arrangements half a world away
The question is always - how likely are political arrangements half a world away likely to get closer and closer. If the US doesn't involve itself in WW1 or WW2 and Germany is able to gain hegemony in Europe, could we have kept its impact far way from us?
if germany had won world war one, they absolutely would have trades with us, just like they traded with China, Brazil and Argentina, who also remained neutral. Holland and Sweden also remained neutral. Germany traded with them too. American entry into the great war was a huge mistake, which helped to cause the second
Completely agree. I wish Taiwan well. But Taipei ain't worth Los Angeles. Also, it'd be nice if they managed to bump their defense spending above 2% of GDP.
If we accept that China is likely (let's say 50/50, but I have no idea) to invade Taiwan sometime in the next couple decades, and that war is likely to go nuclear - I'm reading your comment as it's 90% likely to be nuclear - than it seems like you're saying there's a 45% chance we die in a nuclear war. So I'm trying to figure out what the real probabilities are. If it's really 45% then I need to get a bunker.
it would only go nuclear if one side were so committed to not losing that it would rather die.
if imperial germany and france had both had huge arsenals in 19-7, that conflict might not have gone nuclear. the kaiser didnt want to destroy his country, nor did poincare. but hitler or stalin would definitely have let the birds fly rather than surrender
Right. If anything, nuclear weapons, I fear, are more likely to be used if the PRC were losing, than if they quickly triumphed. For the CPC, reunification is truly a red line. I'm pretty sure they believe, probably with reason, that losing to the United States over Taiwan would force regime change. And that's a recipe for desperate measures.
It's likely to go nuclear **IF** the US gets involved.
Obviously the PRC doesn't need/won't use nuclear weapons if it's just them vs. Taiwan.
As to how likely it is that nuclear weapons would be introduced if the US decided to go to war with China to prevent the annexation of what Beijing considers (not implausibly, I should add) to be a Chinese province, I hesitate to put a number on it. But even if it's only ten percent, that's ten percent too high for my taste. Let's put it this way: if the last couple of decades (911, Great Recession, covid) have taught us, anything it's that black swan events happen.
For the record I'm not suggesting the US, in conjunction with allies, shouldn't prepare a blistering response if Beijing makes a move on Taiwan. I just happen to believe that response shouldn't be a military one (other than, pre-invasion, selling weapons to Taipei). Ideally you want PRC leadership to calculate it's just not worth it.
I think if Great Britain sued for peace with Hitler, then there's no massive flow of supplies from the US to the USSR and, sooner or later, Hitler defeats the Russians. Probably much, much sooner.
Perhaps not as much "duking it out" as you imagine. It was a near run thing, but Great Britain absolutely made the right call for its long term interests.
Britain caught two breaks it had no right to expect. Germany invading Russia and Pearl Harbor. Without those, they would, at best, have had to build up the navy and fight a 40 year low intensity war while Europe did its own thing
Britain and Russia could have won without Pearl Harbor, American supplies were more important than American troops and they were going to the UK and USSR anyway. Barbarossa was critically important though.
Hard disagree. A successful invasion of Russia would hardly have been to Britain's advantage. Pearl Harbor turned out great for Britain but really only because of what *was* really Britain's luck break, which you did not mention: Hitler's foolish declaration of war on the US. That may have been the greatest own goal in world history.
Eh, it wasn't really that much of an own goal. Once the US was unambiguously at war with Japan, it really would not have been a stretch for the US to declare war on Germany shortly thereafter even in the absence of an express declaration of war by Germany. US and German naval vessels had already been shooting at each other for a few months before Pearl Harbor and the Tripartite Pact expressly committed Japan and Germany to mutually assist and support each other. Hitler would have needed to not just forgo declaring war on the US, but actively repudiate the Tripartite Pact and make some really conciliatory gestures toward the US that Germany basically couldn't afford to make (e.g., agreeing to no attacks on US-flagged merchant vessels bound for Britain) to avoid a US DoW.
The Tripartite Pact called for each to declare war on countries that attacked one of the signers. The US didn't attack Japan, but the opposite.
Possibly the US would have declared war on Germany. Possibly not. American feelings toward Germany were a lot murkier than they were toward Japan after Dec. 7, 1941.
Without pearl barbour, germany would never needed to choose between war with the US and keeping japan as an ally. Without the invasion of Russia, Britain would have plodded on alone, able to shelter behind the channel but compl troy unable to make a dent in fortress europe. At best it could have fought an Italian campaign a la Wellington agsinst Napoleon.
I do basically agree with this take. However, I think a lot of US fiscal dominance is also tied up in (perceived) military dominance, and just overall being The Superpower. That's why the dollar is the world's reserve currency, and so on- that's why we're able to run such massive deficits for our incredibly expensive social welfare programs.
If we (when we?) just allow Taiwan to be invaded, there's going to be a huge run on Treasuries & the dollar. Once we're not perceived as number one anymore, those deficits are going to catch up to us. This is not a pro-war with China take- that's insane- but just a much more pessimistic read on where the US will end up. I'd say there'll be at least a decade or two of serious financial problems, before maybe a rebound later on in the century?
(And, like, there's nothing that we can do to punish China for invading Taiwan? I agree that a full-on naval war in the South China Sea is insane, but we can't just sit at a distance, blockade their ports, and cut China off & let their economy collapse as punishment? Make Taiwan a Pyrrhic victory, I dunno)
Going to repost a comment I originally made on Noah Smith's blog, because it spells out part of the approach I believe we should be taking wrt China:
I've always gotten the vibe Chinese leadership in general fears a brain drain to the the West (especially US). So, when Washington's discriminatory policies make it harder for talented Chinese researchers (and well-educated people in general) to come to the United States, Xi and the Standing Committee aren't upset. In the least (their public rhetoric notwithstanding). Rather, they're doing high fives.
America should be doing everything in its power to attract the world's best and brightest: more student visas, more green cards, special visa programs for foreign graduates of US colleges, more H1s. All of it. And yes, that should include lots of China's best. STEAL THEIR TALENT. I guarantee many top Chinese brains will LEAP at the chance to trade in Chinese smog, crowds, low wages and totalitarian constraints for the freedom, clean air, potable water, wide open spaces, higher wages, and detached houses of Silicon Valley, Boston or Seattle. This is a complete no-brainer.
(Ok, so maybe not many will be able to buy a house in Palo Alto, at least right away! But it's getting pretty tough in Beijing or Hangzhou, too!)
PS — Completely non-scientific observation here, but, personal experience to me suggests China's educated classes are indeed getting a bit nervous, and there's a general increase in top level talent trying to get out. I've got five well-educated friends here who are finalizing exit strategies: three to Canada, one to the US, and one to Europe.
PPS — Preemptive reply to the inevitable objection based on national security: fears of Chinese espionage are overdone for a variety of reason; our people (FBI) in any event know how to handle this stuff, but if their budget needs to be increased, so be it; cutting ourselves off from an increasingly large portion of the planet's smartest folks isn't going to just help China: it'll also directly weaken US science.
Yes this is a great idea. I find it bitterly ironic that the "America is the greatest place that ever existed" folks are terrified about welcoming new talent, while the people who want to make it easy to come here are aligned with the "America is a racist, sexist hellhole" brigade.
You'd think that if you could get patriotism and immigration support aligned in one party they would be a juggernaut on that issue.
It's deeply ironic that the GOP isn't salivating at the possibility of bringing in a lot of new voters predisposed to hate socialism and big government.
You mean the big government and socialism that has taken them from a multi millenia history of famine and a multi century history of backwardness to the brink of industrial first world status?
That's a good point: some of the Chinese I know defend their government as generally beneficent, and, especially after COVID, they see their approach as better overall. Still, even for them, staying in the US is better for their science, and we should do everything in our power to keep it that way.
Stealing the best and the brightest is absolutely the US's lower-case-s superpower. FFS, do it more!
Really interesting.
This supports a suspicion of mine that for all our flaws, no one has come up with a better model for international prominence/influence/even dominance than the US. The Axis powers fell before the united democracies (+ the USSR); the USSR alternative flopped; our fears of a dominant Japan were overblown -- and we just keep trudging along.
I suspect the same will be for China. Despite its laudable success in bringing hundreds of millions out of poverty, I don't think it has the tools or ideas to become a true global competitor to the US. Its demography is a killer, and unlike the US it doesn't have alternatives like immigration to counter it. It is growing its military power but doesn't have a clue how to pursue effective statecraft (its real superpower seems to be pushing more and more nations to ally against it). Its state-managed capitalism has helped it rise to a middle rank economically but doesn't seem to have any idea how to help it become a truly advanced economic power.
So whatever errors we're making and have made, it looks like the good lord keeps looking after us.(*)
(*) Subject to change after November 2024.
It's hard to beat the US formula of large, open, and assimilative/syncretic.
China has us beat on large, but Xi seems to be taking them in the wrong direction on open, assimilative and syncretic.
That assumes some basics of good governance in place like rule of law; otherwise we'd be worried about India surpassing the US instead of China.
Only objection I have there is that our population was growing through defeating Axis and USSR and Japan given spacial constraints never had a chance. China on the other hand is much bigger and without economic collapse there's no reason to think it wouldn't surpass the US. In other words ONE BILLION AMERICANS!!! :)
It seems obvious, but a simple way to reduce external threats is to become the greatest country in the world and invite as many people as possible to come live here
As a scientist in academia, I just want to say: THANK YOU!!! I have been banging this drum for a while now and I've often felt like the only one who gets it. We need to compete with China on talent. That means making it easier to do science here and to stay here. Current policies do the opposite.
I've worked with plenty of Chinese post-docs. And a lot of them did want to stay in the US. I had a conversation with one of them and he mentioned that he was part of a small ethnic minority in China. He seemed quite happy when he said that in the US it really didn't matter what ethnic group he was from. The US has a racism problem, but the Chinese government's treatment of its ethnic minorities in the past 5 years (or for that matter the past 50 years) is reprehensible and will probably only get worse. Science does best in open societies because science often reveals things that governments don't like. If I were an infectious disease scientist in China right now, I would want to get out soon because there is a good chance that the government would put pressure on infectious disease scientists.
I wholehearted agree. The US beat the English in the revolution with our English. We won the American civil war by beating their Americans with our Americans. We won the world wars by having our Germans, Italians, etc. beat their Germans, Italians, etc. We should look to defeat China (preferably in economic, not military war) by having our Chinese beat their Chinese - just need more Chinese to do it. Also should prepare for any Indian conflict by getting a bunch of them as well...
I mean, thats a nice story...But the higher the percentage of German ancestry in a county was correlated with voting against FDR and for anti WW2 candidates
Sure - but when it came time to fight, build, and grow - they did. In fact, one of our biggest shames was thinking that the Japanese part of Japanese-Americans was more important that the American part.
We see a similar thing today with absolute maniac Cubans in Florida holding the entire country's foreign policy with Cuba hostage. Immigrants often have insane politics when it comes to their country of origin
That's not similar at all.
In the case of Cuba, the Cuban immigrants are extremely anti-Cuba.
In the German example, the German immigrants were soft pro-Germans.
I'm curious - why are your 5 friends leaving?
I'm also curious, if you're up for sharing, but how did you end up in China (Beijing?) and why are you staying?
"We are not going to hit the 1.5-degree or even 2-degree target, but 2.7 is much better than 3, and 2.4 is much better than 2.7 — we need to do what we can."
A good post, but I'm going to quibble with this particular sentence, because I think it leads to bad analysis and bad policy recommendations. True, we most likely can't hit the 1.5-degree target, but we most certainly can hit the 2-degree target. In the 2018-IPCC report (the newest report has roughly the same findings, but I'm a little less familiar with it) the remaining carbon budget of the whole world for the 2C-target with at least 2/3 probability was estimated to be a little above 1000 GtCO2. For 50 % probability, the budget is 1500 GtCO2.
That's between 20 and 35 years of our current global emissions! The carbon budget for the 1.5-degrees target however was just 420 GtCO2, which is roughly ten years of emissions (from 2015!). If we simplify enormously and assume a linear reduction to net zero from today, that means we need to get half way to zero emissions in ten years and all the way to zero in 20 years to limit warming til 1.5-degrees. To limit warming to 2-degrees might require halving emissions in 20-30 years, and then getting to net zero in the next 20-30. Even if emissions plateau for a few years and then start going down, we still have a few decades. Getting to net zero in a few decades certainly going to be difficult, but it is not at all inconceivable.
The climate science fully supports the notion that every 0.1 degrees C counts, like the post says. So stabilising the warming at 1.9, 1.8 or even 1.7 degrees instead of 2 degrees matters a lot. This perspective indicates that speeding up the transition to lower emissions is hugely important. And yes, the international negotiations do matter. If Chinas emissions peak a few years earlier and then falls gradually instead of plateauing, that matters a lot. International pressure, negotiations, but also technological and financial developments do affect this.
The longer time horizon of the 2-degree goal also means that technologies that take longer to mature can play a larger role. In climate circles many people say that advanced nuclear is not important, because it won't be ready and deployed at scale by 2030. But it might be technologically ready by the late 2020s and deployed at scale gradually through the 2030s. Furthermore, if that provides energy abundance, that could make some of the hardest climate problems (such as industrial heat, or heavy transport and air travel using synthetic fuels or hydrogen-based fuels) easier to achieve. The same applies to other technologies.
The current focus on the 1.5-degree target means that we need drastic action right now, or we fail. I would certainly support drastic action right now, but that's not the only option. Accepting 2.7 or 2.4 degree warming as the best we can do, means we can just patiently wait for technologies to mature. Aiming for limiting to at most 2, means that we need to act now and that there are plenty of politically realistic things we can do.
Great post IMO.
"Each story you read out of Xinjiang is more horrifying than the next."
Whew! Thank god it's getting better all the time. I was afraid you were going to say it was getting worse!
WHERE ARE THE EDITORS!!!!
How about, uh, the Republican Party radicalizing against democracy? Was a bit amazed to see out-there stuff like 'AI takeover' mentioned in the piece, but not the fact that one of our two major parties is turning into some kind of Peronist/North Korean-hybrid personality cult.
If we had a more logical electoral system, a lot of Republican craziness would be for naught. But we have a very specific failure point in the ridiculous Electoral College, where Constitutionally there is at least some play for states to send electors based on whatever criteria they please. I have seen a number of thoughtful commenters, including conservatives, note that Wisconsin, Pennsylvania, Arizona, Georgia and others could send the 'wrong' electors just based on allegations of voter fraud. Even if the court system sorted it out later, the loss of perceived political stability to the US would be permanent, permanent damage. We could divide American history into before and after. And that's without rioting, supporters clashing in the streets, shootings, etc.
BTW, Trumpist hysteria & death threats against Republican legislators who voted for the infrastructure bill kinda debunks MY's pieces about how the R's are becoming more politically moderate (while procedurally radical). I don't expect them to be loudly for it, but being (literally) violently against even bridges & roads is a pretty ominous sign
The hostility toward R legislators who voted for the infrastructure bill is precisely procedurally radical: they would have happily supported the same bill if Trump had proposed it. It's an example of denying the opposition party the right to govern.
I think it's second tier compared to near-existential threats - global warming, AI controlling humanity, and super power war.
Definitely at a global scale I agree, but at the country scale, democratic decay is quite possible and that is pretty impactful for the US.
First, I will put the obligatory disclaimer that I don't believe Trump's election lies.
That being said, I think the Dems election lies are only slightly less troubling (there was no voter suppression in 2018, or 2020, there were record turn outs).
Worse, is the HR1 bill that seems determined to make fraud easier.
Without faith in free and fair elections democracy can't work. We need faith in elections, and partisan's need to stop spreading lies and conspiracy theories.
Moreover, we are going to need to agree on a basic framework. That's going to be something like several weeks of mainly in person voting, but also requiring voter ID
Honestly I think this whole argument goes to Matt's point about arguing about symbolic issues. Dems talk much more about the symbolic election stuff when really what matters is gerrymandering. That said while democratic aren't perfectly honest about elections stuff, they are not fueling outright conspiracy and at times encouraging violence.
Stacy Abrams election lies were definitely outright conspiracy.
She just didn't have the platform Trump had
The are only two major threats from AI.
The first is that people believe the lies that hucksters tell and give these systems more control in safety-critical situations for which they cannot safely perform (see: everyone thinking Tesla’s autopilot features are actually sufficient to drive for them).
The second is that people use algorithms to bias-launder. Doing things like releasing resume-scanning systems that somehow always reject non-white applicants (likely because it’s training data was from compiled by racially-biased people) yet claiming that the system cannot be biased because it is a computer and cannot harbor personal racial hated.
The former will likely go away when more people have more hands-on experience and recognize how much crap like “autopilot” is janky as all hell. The later is a much bigger and more imminent threat. But what is not a threat is any kind of singularity/SkyNet situation where strong AI comes into being. When people were resigning in protest from Google/Alphabet’s AI ethics group, it wasn’t because the researchers were concerned about an evil strong AI that Google refused to stop creating. It was because of the bias-laundering issue.
At the risk of sounding like some kind of AI alarmist, I can't help but comment that to say that there are definitely only two threats from AI and that they are fairly pedestrian as far as threats go, strikes me as being extremely overconfident. I don't think you have to get fully onboard the godlike super-intelligence singularity train either before accepting there are other genuine reasons for concern (automation, surveillance technology, AI weapons systems, etc., just to name a few) if we look beyond the immediate timeframe of the next few years. I don't see reason for the skepticism about the fact that there is a threat itself, although as was pointed out in the post, the position with actual AI safety efforts is pretty bleak right now.
I'll believe AI is really a thing when it passes my own version of the Turing test: write its own really killer standup comedy routine and deliver it (well, from behind a screen of some type), and not have anyone realize it wasn't done by a human.
I think automation doesn't fit as an AI threat for the same reason that the robotic automation of the 80s wasn't a threat. There may be temporary short-term losses in areas that are automated, but just like in the 80s we find that those jobs move into other sectors and the overall increase in economic productivity is a net-positive. (Also, if the automation you're worried about is something like all the truckers being replaced with self-driving cars, that's a fairy tale. Level 5 driving automation effectively requires Strong AI, so we're gonna see SkyNet before we see that. And level 4 driving automation would require someone behind the wheel, so the jobs aren't going anywhere.)
As far as surveillance technology, I think the threat there is far more from the actual capture tech rather than any machine-learning models they have running behind it. A million ring cameras installed in every neighborhood are the potential problem. Whether there is a human combing through the footage or a script is less of a concern (especially because a human is a lot more likely to clip the footage and tweet out some random embarrassing moment caught on camera).
And in regards to AI weapon systems, I think that falls under the first threat: Safety-critical system being handled by computers that aren't up to the task. But I also think that weapon systems are less of a concern, because in general military officers are less likely to have the "every mistake is just helping them make the next iteration better" attitude you see from Tesla drivers commenting on every video of an autopilot crash.
So, I think you are kind of underrating the risk from misaligned AI. You don't need a strong or even a general AI to cause an alignment crisis. You just need two things:
1) An AI with the power to do things without a human intermediary
2) Some goal for the AI to maximize
It doesn't have to be smart, evil, or anything else. Because a human intermediary is extremely slow, AI will inevitably be given the ability to directly affect the real world (in fact, it already has in financial transactions). It doesn't have to be an existential threat to be very bad. Example: profit-maximizing corporate AI notices enforcement patterns for some major regulation, determines that if it breaks the law in a specific way it will most likely never be caught.
Secondarily, the other huge threat from AI (one that is far worse than bias), is what it will enable regular ol' humans to do. AI tools will give single individuals with sufficient computing power the ability to have incredible power. This is already what has happened with social media companies, who have gained the ability to tweak various parameters and cause social and cultural change on a whim. Now imagine if that same ability was applied to the real economy of actual atoms via things like automated factories.
You're making one massive error here, assuming that an AI will be able to spontaneously go outside of its parameters in order to continue its maximizing goal. An example I like to use is chess AI. There are now a lot of extremely excellent AI chess playing systems that are able to easily beat all human opponents. However, no matter how optimized a chess-playing AI is, (even if you take the chess-playing AI and pit it against another chess-playing AI and have them continue to optimize against each other), that AI is never going to try and get away with making illegal moves to win. Also, even the most perfectly-optimized chess-playing AI will freeze up and be unable to make a single move if you have it try to play a game of checkers.
And on your second point, there are already real-world examples of people taking the tools and mis-using them. But it is happening today, and the mechanism by which is it happening is through things like bias-laundering. (Also, worry about regular humans causing huge problems via factories is precisely why we have the majority of our current environmental regulations.)
I think you may misunderstanding the very important point being made here. It's not that AI will spontaneously gain control of things on their own, it's the people will give it too much power to begin with, and it will use that power is harmful and unexpected ways.
The biggest example of this is Facebook. A slew of powerful AI algorithms control what people see, driven only by the objective of increasing user "engagement." In practice, the AI basically figured out that the best way to do this is to present the most inflammatory content possible, without regard for accuracy or moderation.
This is by far the biggest risk of AI at the present. Both of your original examples are specific instances of the broader issue Kade presented.
Another way to think of it: A profit-maximizing trading algorithm is still just a trading algorithm. It is going to be limited to its initial parameters of being able to do trades in the specific market where it operates. A poorly-aligned trading AI may cause a flash-crash, but it isn't going to suddenly start issuing credit default swaps.
And be absolutely screwed when environments change, and if it’s up against a real trader on the other side of the book with a bloodlust, a mortgage to pay, and numbers to hit that is taking in every tick on the BBG terminal to his trained senses
Here’s my general take on this. There are two real innovations that are expanding AI usage.
-Increased data storage capacity
-Advances in convex optimization research.
Both of these innovations are slow and steady for years and alone will not bring about this grand alignment issue that kills all humans. They lead to more data, and faster cleaner error minimization on that data.
The illusion of revolution is due to the advent of accessible, easy to use software packages that can make an AI model with very little code or understanding of the problem you’re trying to solve. These are dangerous not because they’re intelligent, but because as Lance has said, they’re trusted, go into production, suck at it, and screw shit up.
This is a threat. And the black box that he mentions does allow companies to do dirty work “bias laundering” in the name of innovation that they just “didn’t realize was causing this problem cause it’s so advanced but we have to innovate!”
The solution to this isn’t a safe harbor for bias but a re-calibration of the word “innovation.” If Wal-Mart suddenly sold space ships and nuclear reactors you wouldn’t call a guy who made a nuclear reactor space ship in his backyard an “inventor”
Literally pounded the “comment” button to write basically this. You beat me to it. Good day sir.
I think that's a reasonable analysis of _today's_ machine learning, and it's all totally valid. If I'm trolling someone concerned about general intelligence I say things like "what if your machine learning algorithm is programmed to maximize engagement or viewing hours and takes the strategy of teaching people that the world is flat."
I think the concern of the AI researchers who are looking at long term existential threat is that we'll wake up one day and have a machine learning system where we can't say where the boundaries are or that it is siloed to its domain or that these clear bright line constraints on capability are enforced.
Humans screw up computer systems _all the time_. We try hard to air-gap systems and we don't always succeed - adversaries take advantage of that now. The researchers fear that an AI that looks for ways to not be contained will be a lot better at finding them than today's human hackers.
Zillow’s news cycle for ruining 1/4 of their company’s jobs by putting too much faith in a crap model is still ongoing and here we have a “but these models might take over humanity” take. I love MattY but this was absurd.
While I generally agree, let's not argue against the worst example. Zillow Offers was an incredibly weak shop.
It's surprising to me that someone who wrote preciently about the structural threats to American democracy would not consider that to be one of the important issues facing us.
It's funny you should mention that, because as I was reading it, I couldn't help thinking that the themes - it matters more what you do than participating in international talking shops, American government is short-termist, there's no unity and so much division, it's hard to pass legislation to achieve necessary goals, our politics is not adapting to Great Power Competition - are all themes that get trotted out by people who think American democracy is kind of decadent and that the Chinese have a pretty good way of running things, all told.
Now to be clear, I don't think that *is* Matt's position. I just thought it was interesting that the themes seemed very familiar.
I mean I think that my living standards are lower than they would be if there were fewer homeless people living around me—to say nothing of *their* living standards—and I think a government could solve or mitigate this problem without recourse to authoritarianism.
Yeah I was gonna say, the living standards of slow boring readers is not greatly affected but it really does hurt Americans. Also, while I am sure to one day become part of the American Bourgousie I am currently an indebted college student who also would like to live in a city, and I would really like to have a better government(s) that didn't let housing costs explode because the current bougousie are mostly NIMBYs.
Re what matters, why are the Dems allotting $285 billion - the SECOND LARGEST PROGRAM in BBB - to helping out the well-to-do by lifting the $10,000 cap on state and local taxes (SALT)? I guess this shows that "working families" are less of a priority than they claim.
https://www.washingtonpost.com/business/2021/11/16/second-biggest-program-democrats-budget-gives-billions-rich
YES. This makes me SO MAD. I personally will benefit from it -- I did some back of the envelope math back in the year of Trump's tax cut, and found that on net it cost my family about four grand, primarily because of the SALT limitation.
The thing is, I was _completely fine with that_. While I live in an incredibly expensive market, and I live basically an upper-middle-class lifestyle, it is just an objective fact that we are rich. Our net worth cracks seven figures, and our income is around the second or third percentile. We aren't _quite_ The One Percent, but getting close. We should be paying more taxes.
The fact that even progressive Democrats like Katie Porter from SoCal felt they had to go all-in on a tax cut for people like me makes me just despair. I assume she felt like she had to, because so many self-styled SoCal "progressives" are infected with Jarvis brainworms, and believe BS lies like "Prop 13 helps low-income people to afford homeownership". With Democrats like these, who needs Republicans?!
Thanks for your perspective. I don't begrudge people their wealth, so I didn't intend to suggest that the well-to-do are unworthy. But, given a choice between making the child tax credit permanent and offering what I understand to be a temporary increase of the SALT cap, I think our country is better served by reducing child poverty.
Katie "a 1 million dollar house is modest in my district" Porter? I just can't believe these supposed progressives have no shame in their SALT advocacy.
I mean, a million-dollar house may well _be_ quite modest in her district. Certainly where I am it is. I live in a million dollar house. It was built in 1940, it's a 4 bed, 2 bath, single story, with a tiny yard. It's _significantly_ smaller than what I grew up in, with my dad working full-time and my mom having left the workforce to be a homemaker.
But, like, so what? The solution to housing affordability isn't to comfort the comfortable while afflicting the afflicted. We need to make housing more affordable _broadly_, which means reforms to allow a LOT more housing to get built. And quite often, you'll find that exactly the people in my situation -- comfortably situated, but discontent because it's so much HARDER to achieve that comfort than it was for their parents' generation -- just want to pull up the ladder behind them. I've had multiple people tell me "The Bay Area is full, we should stop building."
I agree 100%. Housing cost burden in coastal areas that have pulled up the ladder through regulation and zoning is a huge burden on the productivity of the economy as discussed at length here.
All that being said, the optics of saying a "million dollar house is modest" to a national audience is not great! A person owning said "modest" house in Irvine may not feel rich spending 40% of their take home pay on their mortgage, but by definition their family income puts them well into the 90th percentile. And the appreciation of that house in the last year has definitely made them a big chunk of paper wealth.
Democrats could do a lot to help that person (reduce housing cost through zoning reform, college costs/debt, improve early childhood / school quality, etc. etc.), but like, aren't progressives here to help kids escape poverty? Make a fairer and more just society? Not pad the pockets of the already well off in the wealthiest areas of the country?
Agreed, the optics of rich people saying "but I don't feel rich!" is terrible.
I believe it was Brad DeLong who once paraphrased a whine from somebody on this topic as, "Well of course, it is objectively a very large sum of money, but it is so much smaller _after I have spent it_."
My anger at this is never ending. The bill has ended up going from a lot of stuff I thought was good and exciting to one really good thing that doesn't help dems much politically (climate stuff), a bunch of temporary ~okay~ stuff that will just die in the 2023 majority-Republican house, and a massive awful thing in SALT. Honestly if SALT were part of the BBB of July, maybe I'd let it slide, but at this point it's a Climate bill with some temporary and moderately regressive tax changes.
Agree - better to do a few things well and 'permanently' than a lot of things half-assed and temporary.
The SALT limitation raised my federal taxes 60%, not the biggest deal in the world, but pretty eye-opening at the time. I'm sure it did the same for a lot of other taxpayers in high tax states. Not surprising their representatives are responding to complaints, just as they would to higher gas and food prices.
Eye-popping, I'd say. As I wrote somewhere in this thread, I'd support doubling the SALT cap, just not raising it to 8 times its former limit.
You and I agree. What I was surprised by was how big an impact the limit had on my tax bill. I thought I was under-taxed before, and not complaining now. But a simple change to the tax code that changes your bill 60% makes no sense.
You need to throw some crumbs to your donor class to win elections. You can't take a principled stand on everything.
Crumbs??? It's the second largest tax expenditure! I could support their raising the cap a bit, but, as the article shows, only the tippy-top of the income distribution sees a significant tax savings on this, and, for the multi-millionaires, is $15k that big of a deal? People who earn less than $366,000 save ZERO or at most $1,300 per year. Raise the cap to $15k or $20k, but $80k? - no way!
I mean - we may all be getting spun up for nothing - I'm really hoping this SALT crap will get thrown out because it's gross and not even remotely the most important part of the bill.
But if it does end up passing, someone will write "What's the matter with New Jersey" - the democratic party will be a machine to convert progressive outrage at injustice into tax cuts for wealthy property owners in blue states.
Yeah I do have some hope for it getting thrown out considering Manchin is against it, and he's much more likely to tank the bill than Katie Porter.
In a 2022 red wave, Malinowski, Sherrill and probably even Porter and Gottheimer are going to lose anyway, so spending the biggest part of BBB on SALT to give semi-republicans a tax credit is truly the worst possible policy choice democrats can envision.
Wait is SALT exemption still part of BBB haggling? It hasn't been thrown out yet?
According to the WaPo article from Nov. 16, yes. A Forbes article today indicates a couple possibilities: https://www.forbes.com/sites/howardgleckman/2021/11/18/how-an-80000-salt-cap-stacks-up-against-a-full-deduction-for-those-making-400000-or-less/?sh=30b89e35704e
(Bashes head on desk repeatedly.)
I think MY's critique of BBB (tries to do everything poorly, does nothing well) was already pretty damning, but putting SALT deductions in it is gross.
I get where the Biden administration is coming from ("we're comfortable with this crap sandwich because we're not gonna get the votes otherwise" - slight paraphrase by me) even though it's stupid and bankrupt. I'm disappointed in Pelosi for not cracking the whip and killing this thing - it's a bad idea and it undermines democrats among the progressive base and swing state centrist democrats.
This in The Atlantic today:
For a party that campaigned on raising taxes on the rich as a way of reducing inequality, the existence of the costly SALT proposal in the middle of an ambitious progressive bill sticks out like a sore, hypocritical thumb. The policy is no minor throw-in or pet project; its cost as a tax expenditure represents a significant portion of the overall bill and undercuts Biden’s goal of reducing inequality by shifting the tax burden from the poor and working class toward the wealthy. Jason Furman, a Harvard economist who served as the chair of President Barack Obama’s Council of Economic Advisers, remarked earlier this month that the proposed giveaway for wealthy taxpayers was “obscene.” Households making several million dollars a year “could get a $25,900 tax cut,” wrote Chuck Marr of the Center on Budget and Policy Priorities, a liberal think tank.
https://www.theatlantic.com/politics/archive/2021/11/salt-tax-cut-democrats-build-back-better/620754/
Also from that Atlantic article: "The scope of the SALT provision within the overall bill is also enormous; over the first five years of the bill, it is the single largest piece of the proposal, Furman told me, costing more money over that period than the Democrats’ spending on climate change, education, or health care."
Yeah, MY's critique was spot on imho.
Dumb hot take: we can solve the global warming risk and AI existential risk by doing a particularly terrible job at the super power war risk. We'll get into a land war with China over Taiwan and in the combat, the world's major semiconductor fabs will all be destroyed, absolutely devastating the global economy and reducing green house gasses in a super-depression, while also setting back AI research due to lack of computing power.
Don't forget we also probably solve inequality. Historically, war has been a great leveler.
Pretty soon you'll be talking about the benefits of a nice cool nuclear winter to combat global warming..... Every cloud has a silver lining. Sometimes the lining is even radioactive.
Idea for new "Terminator" reboot -- a time traveler comes back to the present to stop the future AI takeover of Earth but his plan for doing so is to trigger a major US-China nuclear exchange.
> The easy-to-mine ores have all been mined.
Doesn't that mean all those materials are just sitting around on the surface now?
"we ultimately need a vision for America that doesn’t rely on hoping the Chinese economy collapses if we want to remain the number one economy in the world."
China will be the world's number one economy unless it collapses. This is just math. China is already the biggest economy in purchasing power terms. If it can achieve the per capita GDP of Greece or Hungary, its economy would be twice as big as ours. Neither of those countries is well run. It's an easy mark to hit.
The good news is relative decline isnt that bad, especially if you chill out. The UK is a good illustration. It was the biggest Western economy in 1850. In the last 170 years, its standing has declined relative to almost every other Western country, yet median British incomes today are 10 times what they were in 1850. Britons are much better off despite all that relative decline.
The biggest peril of relative decline is killing people by overplaying one's hand. It's fine to invest in a navy and nuclear weapons, but Britain has fared poorly every time it has fought great power conflicts on land. Its position declined somewhat during World War One and catastrophically during World War Two. Britain never came close to its original war aim of preserving Poland's independence. It would have done much better to sit out World War Two and let the Nazis and communists duke it out. It could have protected Jewish lives much more effectively by opening India (and possibly the Dominions) to immigration than by fighting a war it had no way of winning. I see a similar dynamic with Taiwan and I dont want my son to die defending the "indepedence" of an island that was part of China for 200 years.
What you say is valid BUT the PRC's demographics (as I'm sure you're aware) are truly dire. It's entirely possible (even likely) US GDP could be growing faster than PRC GDP by the end of this decade. Heck, I wouldn't be shocked if US GDP growth beat's China's in 2021. IOW, China's economy will surely be quite a bit larger than America's in 2028 (in PPP terms, as you point out, it's already a good deal larger), but the US could quite conceivably regain the lead from China in the medium term.*
We just have to stop being stupid on immigration and one or two other issues. Maintaining a favorable demographic profile compared to the PRC should be the easiest of lifts for the United States of America. We don't have to swing for the fences (ie, One Billion Americans). We just have to refrain from making own goals like, erm, watching the UK and Canada beat us in in international college admissions.
*Having the world's largest national economy isn't everything, as you mention. But it's not nothing, either. In any event I fully expect India to emerge as number one in the second half of the century. I believe that, just like the original Cold War, a successful US policy vis a vis China means playing the long game: recognizing that the US has some significant advantages, and acting with prudence and patience. But my fear is US policy makers of both parties seem to think it's in the country's interests to have a major conflict as soon as possible.
Indeed, having a big economy may not mean that much. By some measures, China had the world's biggest economy in the 17th and 18th centuries, and even into the early 19th century (https://en.wikipedia.org/wiki/List_of_regions_by_past_GDP_(PPP)). Really, that just means they had a lot of people. They didn't have the surpluses that come from high GDP per capita to allow them to leverage that gross size into increasing growth or international prominence. Instead, they were just a big, weak and fragile "nation."
You are more pessimistic about Chinese economic growth than I. China’s GDP has exploded despite a stagnant population. It’s growth has been driven by importing technologies developed elsewhere to make its workers more productive. This is a time honored model. It’s what basically all of Latin America and eastern Europe have done. It’s what France and the low countries did in the mid 19th century with British technologies. It’s what India is doing today. This strategy will work until Chinese per capita GDP is a large fraction of the US or until upheaval stops growth.
This model hasn’t worked in Africa or much of the middle east because it requires political stability. The CCP appears pretty solid to me, and established Chinese dynasties have lasted 200-400 years. Is there any reason to think the CCP won’t last as long as the Ming dynasty?
Yeah, I'm definitely more of a pessimist than you re: China's growth prospects.
The thing is, China's population hasn't really *stagnated* through the post Deng/Opening Up years. Since 1980 PRC population has increased by more than 50%. That's about the same as the US. That rate of population growth has slowed, of course, but it's only in recent years that's it's begun to approach Russia/Japan/Italy territory. And sure enough, economic growth, too has slowed. If you believe official statistics, the country's economy has been expanding since 2010 at less than half the rate of the go-go years (in the 6% range) and there's wide skepticism these figures are even accurate. And indeed, China's workforce has begun to shrink in absolute terms. And the country's efforts to exhort women to have more babies would appear to have fallen on deaf ears.
I'm not suggesting China's about to suffer some massive crash, mind you. When your population is shrinking, living standards can still increase nicely even when you're only growing at a modest 2 or 3 percent annually. But yes, I do indeed think such numbers are in China's very near future. The country's technocrats get a lot of things right, but they're not magicians, and they're not going to enjoy developing world growth numbers with Japan's demographics. (I also think there's growing evidence Xi isn't particularly competent on economic matters, so there's that, too).
Since 1980, Chinese GDP has grown by -1200% while population has grown by 43%. Calling China’s population stagnant was inartful though, as you note, population growth has slowed markedly since 1990. Still, population growth is only very loosely correlated with GDP growth. Most GDP growth has had little to do with more people.
China’s urban population, which has surged, is likely a better predictor of GDP. Peasants contribute very little to GDP. Indeed US agriculture is still radically more efficient than China’s.
>>> I see a similar dynamic with Taiwan and I dont want my son to die defending the "indepedence" of an island that was part of China for 200 years.<<<
The world would be miraculously lucky if an actual shooting war between the US and the PRC were confined solely to the battlefield deaths of soldiers and did not encompass (as seems terrifyingly likely) the mass death of civilians caused by nuclear weapons.
Which is why I think Taiwan hawks are a huge threat to the world. I’m all for self determination and I wish the CCP were too. But I care much more about peace than political arrangements half a world away
The question is always - how likely are political arrangements half a world away likely to get closer and closer. If the US doesn't involve itself in WW1 or WW2 and Germany is able to gain hegemony in Europe, could we have kept its impact far way from us?
if germany had won world war one, they absolutely would have trades with us, just like they traded with China, Brazil and Argentina, who also remained neutral. Holland and Sweden also remained neutral. Germany traded with them too. American entry into the great war was a huge mistake, which helped to cause the second
And the Nazis wouldn't have come to power.
Completely agree. I wish Taiwan well. But Taipei ain't worth Los Angeles. Also, it'd be nice if they managed to bump their defense spending above 2% of GDP.
Why is it likely to go nuclear?
If we accept that China is likely (let's say 50/50, but I have no idea) to invade Taiwan sometime in the next couple decades, and that war is likely to go nuclear - I'm reading your comment as it's 90% likely to be nuclear - than it seems like you're saying there's a 45% chance we die in a nuclear war. So I'm trying to figure out what the real probabilities are. If it's really 45% then I need to get a bunker.
it would only go nuclear if one side were so committed to not losing that it would rather die.
if imperial germany and france had both had huge arsenals in 19-7, that conflict might not have gone nuclear. the kaiser didnt want to destroy his country, nor did poincare. but hitler or stalin would definitely have let the birds fly rather than surrender
Right. If anything, nuclear weapons, I fear, are more likely to be used if the PRC were losing, than if they quickly triumphed. For the CPC, reunification is truly a red line. I'm pretty sure they believe, probably with reason, that losing to the United States over Taiwan would force regime change. And that's a recipe for desperate measures.
*1918
It's likely to go nuclear **IF** the US gets involved.
Obviously the PRC doesn't need/won't use nuclear weapons if it's just them vs. Taiwan.
As to how likely it is that nuclear weapons would be introduced if the US decided to go to war with China to prevent the annexation of what Beijing considers (not implausibly, I should add) to be a Chinese province, I hesitate to put a number on it. But even if it's only ten percent, that's ten percent too high for my taste. Let's put it this way: if the last couple of decades (911, Great Recession, covid) have taught us, anything it's that black swan events happen.
For the record I'm not suggesting the US, in conjunction with allies, shouldn't prepare a blistering response if Beijing makes a move on Taiwan. I just happen to believe that response shouldn't be a military one (other than, pre-invasion, selling weapons to Taipei). Ideally you want PRC leadership to calculate it's just not worth it.
I think if Great Britain sued for peace with Hitler, then there's no massive flow of supplies from the US to the USSR and, sooner or later, Hitler defeats the Russians. Probably much, much sooner.
Perhaps not as much "duking it out" as you imagine. It was a near run thing, but Great Britain absolutely made the right call for its long term interests.
Britain caught two breaks it had no right to expect. Germany invading Russia and Pearl Harbor. Without those, they would, at best, have had to build up the navy and fight a 40 year low intensity war while Europe did its own thing
Britain and Russia could have won without Pearl Harbor, American supplies were more important than American troops and they were going to the UK and USSR anyway. Barbarossa was critically important though.
Hard disagree. A successful invasion of Russia would hardly have been to Britain's advantage. Pearl Harbor turned out great for Britain but really only because of what *was* really Britain's luck break, which you did not mention: Hitler's foolish declaration of war on the US. That may have been the greatest own goal in world history.
Eh, it wasn't really that much of an own goal. Once the US was unambiguously at war with Japan, it really would not have been a stretch for the US to declare war on Germany shortly thereafter even in the absence of an express declaration of war by Germany. US and German naval vessels had already been shooting at each other for a few months before Pearl Harbor and the Tripartite Pact expressly committed Japan and Germany to mutually assist and support each other. Hitler would have needed to not just forgo declaring war on the US, but actively repudiate the Tripartite Pact and make some really conciliatory gestures toward the US that Germany basically couldn't afford to make (e.g., agreeing to no attacks on US-flagged merchant vessels bound for Britain) to avoid a US DoW.
The Tripartite Pact called for each to declare war on countries that attacked one of the signers. The US didn't attack Japan, but the opposite.
Possibly the US would have declared war on Germany. Possibly not. American feelings toward Germany were a lot murkier than they were toward Japan after Dec. 7, 1941.
Without pearl barbour, germany would never needed to choose between war with the US and keeping japan as an ally. Without the invasion of Russia, Britain would have plodded on alone, able to shelter behind the channel but compl troy unable to make a dent in fortress europe. At best it could have fought an Italian campaign a la Wellington agsinst Napoleon.
I do basically agree with this take. However, I think a lot of US fiscal dominance is also tied up in (perceived) military dominance, and just overall being The Superpower. That's why the dollar is the world's reserve currency, and so on- that's why we're able to run such massive deficits for our incredibly expensive social welfare programs.
If we (when we?) just allow Taiwan to be invaded, there's going to be a huge run on Treasuries & the dollar. Once we're not perceived as number one anymore, those deficits are going to catch up to us. This is not a pro-war with China take- that's insane- but just a much more pessimistic read on where the US will end up. I'd say there'll be at least a decade or two of serious financial problems, before maybe a rebound later on in the century?
(And, like, there's nothing that we can do to punish China for invading Taiwan? I agree that a full-on naval war in the South China Sea is insane, but we can't just sit at a distance, blockade their ports, and cut China off & let their economy collapse as punishment? Make Taiwan a Pyrrhic victory, I dunno)