Things I changed my mind about: Part III

What is this post about?

To keep a scout mindset, I try to keep track of beliefs I changed. This not only serves as a reminder that I regularly change beliefs but also reveals the underlying reason for them. Check out part I and part II to get an overview of previous changes and more context/background.

As in the previous edition, I will also write about things that I newly learned even when I didn’t have an explicit opposite opinion about them before.

Personal

Doubling down on AI safety

In 2016, when I first heard about AI safety, I thought “Oh shit. This sounds pretty important”. Already then, I thought it was important enough to focus my career on it. Over time, this belief has only increased in strength and recent achievements in large language models and various other applications of transformers have added an additional feeling of urgency.

By now, I mostly think of AI as a biased coin flip. If it goes well, AI might be able to solve a lot of problems for us, it could be used to eradicate diseases, make us more productive, assist with climate change and much more. If it goes wrong, AI can do a lot of damage possibly harming us (and other sentient beings) in very bad ways. Therefore, I want to work on reducing the probability of AI going bad as much as possible.

To prevent misunderstandings–I still think that other problems are very important and deserve more attention. Pandemics, global poverty, diseases, climate change, sexism, racism, etc. are all things I want to see solved as soon as possible. However, given the limited amount of time and resources I have I need to prioritize and AI safety seems bigger and more neglected than the other problems. I’m happy to discuss my reasoning in case you disagree.

Doubling down on New Liberalism

As a teenager, I used to like the narrative that “we shouldn’t be so narrow-minded about other cultures” and “other political systems can be good too”. While there might be a spark of truth to that, I now think liberal democracies got it mostly right. Liberal democracies that hold individual freedoms high and only curb them when necessary to protect others, e.g. in a pandemic, are probably the best political system we currently have. Similarly, social market economies are the best economic system we currently have. We can discuss the details like how direct the democracy should be and how much redistribution the society should have (see further below) but when compared to other historically tested systems, democracies with social market economies are just better.

This is obviously not a defense of everything the status quo has to offer–there are clearly a lot of injustices and inefficiencies in the current system that still need to be fixed. But there would be more problems in an alternative system such as nations that claimed to be communist or authoritarianism. We can have a long discussion about whether they were truly communist or not but I guess we can agree that whatever they implemented wasn’t optimal.

I want to be replaced

I sometimes catch myself clinging to my current identity even if it is objectively bad, e.g. I think “I’m just a person who is bad at X”. I think this is a self-fulfilling prophecy and it prevents us from improving faster. Therefore, I want to clearly commit to a mantra of not clinging to my current identity. I want to be replaced (see here for full version).

Exposition works really well

I was honestly very surprised by how good exposition works in EA and my academic community. By exposition I mean publishing blog posts, writing in different forums, posting stuff on Twitter and so on. I expected to get a bit of interaction and some casual conversations but it turned out to be much more. I regularly got helpful feedback on my posts from people I didn’t know before, some (probably most) important life opportunities only came due to the existence of my blog, some discussions on Twitter have led to research projects and some of our group posts on the EA forum have drawn the attention of some bigger names in the community.

I guess I previously overestimated how much work was necessary to create relevant output. It went from “literally unachievable by a mere mortal like me” to “5 hours of focused investment is more than enough to write a valuable post”. Furthermore, I think I underestimated how effectively messages spread around. I guess the Twitter algorithm does a lot of legwork here and recommendations spread fast in human networks. Somehow, my posts sometimes ended up with people who I definitely didn’t think of when writing them. This could also be a negative if you share a very controversial opinion.

Everyone is “just some dude”

Sorry for the gendered language but I didn’t find a good alternative word with the same vibe as “dude”. I had this realization in multiple situations, e.g. when supreme court justices show insanely low mathematical skills, when professors sometimes make very basic mistakes or when someone who I idolize for their intellect shows low social skills. Every time the realization hits me again—they are just some dude. They might be really good at some things but bad at others, they are victims of the same cognitive biases as other people and their communication isn’t magically perfect.

I think this makes it easier to approach the people you idolize and easier to understand when and why they mess up.

Mental health is a trade-off

I think it’s really great that people care about their mental health, that social awareness is rising and that employers start to understand that. Most people (>90%) are probably still under-focusing on mental health rather than over-focusing. However, I think some people overdo this narrative. Sometimes you hear things like “mental health always comes first” and I disagree. In my opinion, mental health interventions should be focused on preventing and treating downsides and less on trying to create large upsides. Most things in life are trade-offs and I think mental health belongs to them. In many cases, it might not be worth investing the time and energy to marginally improve your happiness, e.g. by spending thousands of hours meditating, because there might be better uses for your time.

To prevent misunderstandings, I want to emphasize that I endorse people going to therapy and engaging in preventative measures. But I think it is completely fine for someone to say “I work longer to get more done even if it makes me less happy” or “I don’t meditate because the upsides are too small for the time invested”. This may sound like a controversial opinion to some but I’d expect this to be the default opinion in society.

It’s really hard to build a good team but it’s easy to destroy it

Hiring / building a team is one thing that I drastically underestimated. Some reasons for that include

Combinatorial explosion: There are a lot of explicit or implicit conditions for a position. You want the person to be good at the task (which are often multiple conditions), you want them to fit the current team, you want them to have good social skills, and much more. With every additional criterium, your pool gets drastically smaller. So even if you have 100 applicants, there is a decent chance nobody will fit all criteria well enough.
It’s easy to destroy a good team: If the ratio of junior to senior people is too high, the senior people will be overworked and some might leave. This amplifies the problem and might lead to the collapse of the team. Furthermore, it often takes just one person to destroy a good team. This might be a person that consistently doesn’t comprehend at the same speed as the rest of the team, talks too much, holds inefficient meetings, etc. As soon as this happens the best people leave to form a new team.
It’s hard to keep up morale: Even if you found a good team, some things might go wrong, there might be infighting, there might be substantial disagreements or the vibe just feels off. A perfectly good team could be ruined by small things so it makes sense to invest substantially in good vibes in the team.
Management matters a lot: I underestimated the variance between managers. Previously, I thought that managers have some but not a large effect on a team’s output or morale. Now I think that good managers can drastically improve a team’s output and morale and that bad managers can single-handedly ruin a perfectly good team.

The consequences of this are

that I can totally understand why hiring processes take long and are stochastic,
I now see it as a good sign when companies want to make really sure that you are a good fit because it implies that the rest of the team is more likely to be good and
once I’m in a position where I hire myself, I will try to take care of these factors because of the large possible downsides of suboptimal hires.

Moving to the US

Two or three years ago, I couldn’t really imagine living in the USA. The country is politically polarized, crime rates are way higher than in Europe, health care bills can drown you in debt, race relations are bad, cities are built around cars and have bad bike infrastructure, and much more. However, I have heard many people who are very bullish about living in the US. For them, the Bay Area, Miami or New York are the best places on earth by far and I trust some of them enough to be at least curious. Furthermore, the EA and AI safety scene is very strong in the Bay Area.

All of this makes me curious about trying out living in the US at least for a while. I might end up not enjoying it but I at least want to give it a shot. I also think there is a decent chance that I will like it much more than expected.

Frontloading life

I have witnessed a lot of situations where to get something you have to already have it. To get into a Ph.D. program people want you to have extensive research experience, i.e. the thing you are supposed to get during a Ph.D. To get a grant, people want you to have results already, i.e. the thing you want to get with the grant. If you want to start a start-up people want to see a product, i.e. the thing you want to develop. If you apply for a job, people want you to be good at the job already. Of course, this doesn’t apply to everyone—many VCs or grantmakers are realistic about this. However, many times you will have to deal with people that have these kinds of expectations.

My experience is that it really pays off to find out what kind of expectations these are and try to build credible evidence for them, e.g. do a project on the side, read papers, work on open source code, etc. This usually comes with some financial or time investment upfront (thus frontloading life) but it can meaningfully increase your chances of success. If you are currently not in such a situation it helps to put some money on the side in case the preparation period takes longer.

There are more abstract notions here that a) doing things that make sense before you are explicitly asked to usually pays off and b) taking the right risks usually comes with high upsides—at least in expectation.

Most of this point can be summarized as “be more like Hermione Granger” and is not a major update for you but I find the framing helpful anyway.

Improvements and suffering

I found motivational statements a la “just put your head down and hustle” or “no pain, no gain” a bit amusing. I think that a lot of pain/annoyance is not necessary for improvement and could be prevented by better management or a different frame of mind. For example, learning a language with a book is less fun than with duolingo/lingvist or other apps but not necessarily more effective. In this case, the app is just the superior alternative.

I used to think that most improvements can be like this, e.g. there is an option that is superior in most aspects and doesn’t imply suffering/annoyance. However, I don’t think this is realistic.

If you want to get good at something it usually requires pushing your limits, investing a lot of time to understand it or just putting in some annoying work to get there. Expending resources on something that isn’t survival or reproduction is an evolutionary disadvantage so it only makes sense that we are hardwired to feel bad about it. Therefore, I expect that many things that will lead to improvements in the long term might come with some negative side-effects in the short term. I still think there are clearly superior strategies but I now expect even those to come with some pain that is worth trading off.

Gene editing

I was above average bullish on gene editing before researching it. However, since there is such strong opposition to gene editing in plants, animals and humans, I decided to dig in to understand their arguments and I was surprised by my findings.

The opposition’s arguments were way worse than I expected and often rely on misinformation and very weird trade-offs, e.g. talking about super mutation weeds even though there is no scientific evidence or theory of these super weeds’ existence or any of the suggested harms.

The proponents’ arguments, on the other hand, were even stronger than expected, e.g. by increasing yields of different plants by >3 times. In short, I find it very hard to argue against “we improve the good things and remove the bad things” and most contra-arguments are fixable by a better implementation. You can find my detailed post here.

Economic

GDP & economic growth matter

I often hear the sentiment that “GPD doesn’t matter after a certain point” or that “GDP doesn’t describe how people actually feel” and I agree that GDP is not perfect. It doesn’t capture everything we care about and it likely never will. However, I’m convinced that these criticism are small compared to the benefits of using GPD and economic growth as a metric. This is for multiple reasons.

GDP is the best available approximation to value and wealth we currently have. GDP essentially captures how much people value things in the economy by how much they are willing to pay for them. It’s not just some weird number, it really captures something we care about. It usually also captures public spending, so public goods are not excluded.
Economic growth is the prime example of a positive-sum game. It sounds too good to be true but everyone can get richer in real terms, i.e. controlling for inflation and purchasing power. Of course, that doesn’t mean that the wealth is correctly distributed around the globe but the fact that everyone gets richer is already mind-bogglingly great to me. Since positive-sum games are great, we should keep playing them unless the harms outweigh the benefits. Since I don’t think that e.g. economic growth and climate change prevention are mutually exclusive, I think we should strive to find ways that both reduce climate change and increase economic growth.
Growth still matters in rich countries: Some people say that poor countries should keep growing but rich countries should stop at the current level. I understand and agree that growth should not be the only metric we should look at. But all else equal, more growth is better. I think you could have made the same argument that countries are “rich enough” in 1980 and yet the vast majority of people in exactly these countries are much better off today than their 1980s equivalent. They are richer, have better access to healthcare, better education and better entertainment. I expect this trend to continue and think we have to tackle important problems like climate change in other ways than “just reduce economic growth”. I have elaborated my view in my post Against Degrowth, for massive Green Investments
GDP is power. Why can the West sanction Russia? Why do people want to move to the West? There are, of course, many answers to these questions but first and foremost because the West is rich. The EU’s economy will tank less than Russia’s and people want to come here because they want to participate in our economy (and we should let them in!).

Land value tax!!

I didn’t know what a land value tax (LVT) is but some of my friends who I trust to have good judgment were really excited about it. So when Astral Codex Ten hosted a piece on it, I wanted to understand what all the buzz is about—and I understand it now. In short, an LVT taxes a high proportion of the value of land. This simple idea has a bunch of really nice implications.

It forces you to use the land efficiently. If you want to keep your extra garden plot in a high-rent area, be prepared to pay a lot of money for the privilege. The space should be used to build houses in which many people live and thrive and your extra garden should be outside of the city where land is less valuable.
It doesn’t create negative incentives: Most taxes create negative incentives. An income tax reduces the value of work, a wealth tax reduces the value of wealth creation, a value-added tax reduces the value of production & consumption, etc. The LVT doesn’t have this problem. Therefore, if states had more income through an LVT, they would have to take less money from other taxes which would unburden taxpayers by paying less and by having better incentives.

There are, of course, more pros, cons and technical questions with the LVT that you can and should read up here. But I think the idea is a very clean way to solve some major social problems and we should at least give it a try.

Nuclear power

I was a fan of nuclear power. It’s carbon-neutral, it’s relatively safe and it provides a stable independent source of energy. However, when I looked into it in more detail (see my post with Samu for more), I changed my mind a bit. My main conclusions are

Nuclear is pretty safe and waste is not that much of a problem. It’s not perfect, but the public perception is overblown.
Nuclear is almost always better than fossil fuel alternatives. If states close a nuclear plant and then compensate for the loss with coal or gas, this is clearly bad.
However, nuclear is really expensive. There are people arguing that regulations are too high and that future innovation will reduce the price but even if all of this is true, solar+batteries seem to be cheaper. This trend is likely to continue because the price of batteries and solar are exponentially decreasing, while the gains in nuclear are likely not.

Urban planning

I used to think that housing/urban planning is boring and doesn’t really matter. Now I think it’s exciting and REALLY important. My two main influences were

Not just bikes: a Youtube channel that talks about urban planning in a very direct but terribly effective way. Some of his videos are along the lines of “This is how it is to live/walk/bike in a good city (Amsterdam). This is how it is to live/walk/bike in a bad city (random US city). See the difference?”. After such a video, it’s very obvious that some city designs are just superior in all aspects and that most countries should be more like the Netherlands—I have been orange-pilled.
The housing theory of everything. It mainly states that housing has a really big influence on a lot of relevant things. Expensive housing has social justice implications since poorer people can’t afford them, it has economic implications since a bigger part of the salary goes into rent which doesn’t directly produce anything, it has political implications since it gives more power to homeowners and much more.

This drastically changed the way I think about cities. Now, when I walk somewhere I can’t stop thinking stuff like “This traffic jam could be prevented by a better bicycle infrastructure”, “This house should be taller so more people can live closer to the city center” or “fucking zoning laws”.

Political

The military

In a perfect world, I would want people to focus on positive-sum interactions—helping others, doing science, increasing GDP, etc. Thus, I often thought military spending was unnecessarily high or that the military was getting too much attention.

However, there are a lot of people who have a zero-sum perspective on the world. They think in terms of nations or ethnicities rivaling each other and competing against each other (see e.g. Russia or China). Therefore, as long as sufficiently many people have this mindset, we have to prepare against the possible consequences of territorial expansions, cyber-attacks, etc. Thus, a functioning military is absolutely necessary and countries should be willing to spend lots of money on it.

This, of course, doesn’t mean that current militaries are optimal, they usually have problems with inefficient spending or bad norms.

There are two further observations.

Signal boosting one’s own pacifism by voting for defunding of the military or lobbying against international military cooperation might feel purifying but usually just leads to more harm. For example, not joining the fight against ISIS might feel like “not being responsible for bad things” but is still worse counterfactually for the people killed by ISIS.
The more you denounce the military the worse it becomes. If people present a picture of the military as a bunch of Nazis who primarily engage in war crimes the most qualified people will choose not to join. Thus, the pool of candidates gets worse and the probability of people with strong ideological reasons, e.g. the Nazis and maniacs, getting selected for lack of better candidates is higher, thus increasing the likelihood of Nazis and war crimes.

Both of these are prime examples of signaling that makes you feel better but is worse for the world.

Slightly more technocracy

I used to think that technocracy was a clearly superior form of government in nearly all aspects. It just seemed weird to me that the median opinion could be better than that of experts. And I still think this is somewhat true. However, I updated on a bunch of things

Experts are no saints. They have their own personal goals, they have biases and they like people in their ideological/ethnic/tribe. If a technocratic government is implemented just the right way, it will create better outcomes but it’s much less robust to disturbances and less self-correcting than democracies. Technocracies’ upsides can be higher but so can their downsides.
Good decisions are often meaningless without the buy-in of the population. A mask mandate or a lockdown only make sense when the people wear a mask or stay home. If they don’t understand or disagree, the perfect policy in a vacuum just doesn’t stick in the real world. I would expect democracies to have higher buy-in since the people feel a partial responsibility for whatever the government decides.
However, this doesn’t mean that democracies are perfect and I still much prefer a representative over a direct democracy. People systematically vote against their own long-term self-interest and e.g. questions of interest rates or immigration policy should just be left to the economists. Furthermore, people have a bad track record of caring about future people, e.g. in pandemics, or caring about people in other parts of the world, e.g. with foreign aid. Therefore, I still tend to like technocratic elements but I have toned down my perspective slightly. My rule of thumb is “the more complex and the further away from people’s direct lives a topic is, the bigger the proportion of expert influence vs. people’s influence”.

Underpopulation vs. overpopulation

In school, we were taught about population growth and resource scarcity along the lines of “if every person lived like the average Western person, we would require 3 worlds”. This instilled the belief in me that population growth is bad. Now, I think resource use is a problem but the focus should be on better distribution and more efficient production rather than reducing the number of people. For example, if more people would be vegetarian or vegan, there would be drastically less land use for agriculture and much more people could be fed than are alive today. If we sped up the deployment of renewables, energy scarcity wouldn’t be much of a problem either.

On the other hand, I think underpopulation is a real problem. This belief was influenced by One Billion Americans where Matthew Yglesias makes the case that more people means more growth means more progress and political power which is simple but true. Furthermore, the belief was influenced by the Our World in Data graph on fertility which seems to be shrinking much faster than I expected.

My high-level conclusion from all of this is two-fold: a) underpopulation is at least as concerning as overpopulation. b) More people are a relatively safe way to get to the future faster (more innovation, more growth, etc.). Whether we want that, I’m not entirely sure yet.

Policy design matters a lot

Finding a good policy goal is already hard. You need to balance interests, find compromises, and so forth. However, one part of the policy pipeline that I find highly underrated are the nitty-gritty details of its exact design. For example, whether a child benefit comes in form of a direct payment to your bank account or a tax deduction seems irrelevant on paper but results in two completely different realities. Wealthy, literate native speakers will get the money either way but people who are less fortunate or non-natives might not understand what they have to do or even know that the policy exists. Another example would be opt-in vs. opt-out solutions to e.g. organ donations. On paper, the people who want to donate their organs do it either way but in practice, the two systems result in <20% vs. >95% participation. This one seemingly irrelevant paragraph in the law translates to thousands of lives saved or not saved.

Getting the right goals is one thing but executing it correctly is a whole new problem that lawmakers seem to only slowly catch up on in many areas.

Relationships

The female orgasm is much better

From my limited number of sexual experiences, it seems like the female orgasm is longer and more intense than what I’m experiencing. All (but one) male and female friends I have talked to about this have similar intuitions. This doesn’t have any major implications—just that, maybe, dating utilitarians might be in women’s self-interest on the margins.

Talk about everything

I was always a big fan of talking about everything in a relationship but I have further double-downed on that after reading The pragmatist’s guide to relationships. In my opinion, both partners should always feel like they could talk about a hard question and expect to get help from their partner. If you ever feel like your partner “just wouldn’t understand” or “find it weird” that is a sign you should talk about your communication. The space of allowed topics should really range from “I think I have a crush on someone else” over “I might want to live in a different country” to “I want to try a very unconventional sexual thing”.

Of course, the way in which you communicate still matters a lot. You should want to communicate constructively, signal that you want to get to a good conclusion, and so forth. This stance does not give you permission to be mean or break communication norms “because you are allowed to say everything”.

My reasoning for this stance is as follows. If you intend to lead a long-term relationship, your partner will be the person you trust the most. If you don’t trust them to deal well with your most intimate thoughts, you might not want to spend a substantial part of your life with them. If you think the relationship is only temporary the stakes are lower and it’s easier to be fully honest.

More exploration to find a long-term partner

Also inspired by The pragmatist’s guide to relationships I now think that there should be much more exploration to find a long-term partner than is common. In the book, they suggest that going on more than 100 first dates is a good heuristic. Should I return to the dating market, I might try dating more people but it also sounds a bit stressful. These 100 people are, of course, not randomly selected but pre-selected by algorithms or a specific location.

Improving sex life

I think the default social secrecy about sex means that most people’s sex life could easily be improved.

Communicate a lot. Your partner can’t look into your head and preferences vary a lot between people. Thus, the fastest way to make them understand what you want is to tell them. This can include relatively detailed micromanagement from time to time, e.g. “more like this, faster, slower, etc.”. Detailed communication should also not be interpreted as removing your partner’s agency but as a sign of trust that pays off very quickly.
Talk to your friends about sex. You don’t need to provide a detailed account of what you are doing but sharing broad insights is already helpful. I think it’s kind of weird that we are happy to tell our friends how to get better at cooking, sports, video games, etc. but rarely about sex even though the latter probably has a bigger impact on our lives.
Use technology. A vibrator doesn’t replace you, it frees one hand.

One last note

If you want to get informed about new posts you can follow me on Twitter.

If you have any feedback regarding anything (i.e. layout or opinions) please tell me in a constructive manner via your preferred means of communication.