Where would we set up the next EA hubs?

By mariushobbhahn, MaxRa, Yannick_Muehlhaeuser, JasperGo @ 2022-03-16T13:37 (+55)

A question that we often asked ourselves is “where would be the best spot for a large EA hub?”. By this, we actually mean two questions. Firstly, “Where would be the optimal hub from first principles?” and secondly, “Which locations are currently neglected by the community?”

Recently, a lot of EAs suddenly live (temporarily) in the Bahamas. We think this mostly shows that there is a large number of EAs who have a high willingness to move as long as the outcome justifies it. However, we don’t think the Bahamas are a good long-term destination for an EA hub since they don’t fulfill many of the criteria below. Thus, we think it makes sense to (once again) think more deeply about the location question for EA hubs. 

TL;DR: just skim the headlines and suggestions.

Context: The topic of location has been discussed before (see e.g. this 2020 LW post or this MIRI post). We decided to write this post anyway because a) the topic is always relevant, b) previous discussions left out important locations (e.g. Asia, mainland Europe) and c) recent developments (Covid, Bahamas, new companies) changed some considerations. 

Of course, these are just preliminary thoughts and we’d be interested in your takes and, more importantly, location suggestions.

Criteria

Broadly sorted by how important we think they are and weighed by different use-cases, e.g. governance vs. industry vs. research. However, feel free to reorder them according to your preferences. 

Local Culture & Social Environment 

The entire point of having a hub is to be around like-minded people and benefit from the network effects. This becomes especially important when we want to collaborate with people outside of the EA community who do important work on similar problems, e.g. the Bay Area and AI research. 

Furthermore, the location should probably have a culture that is compatible with (or at least tolerant of) EA. This often entails an interest in intellectualism that is found in cities with a large university or the entrepreneurial spirit of start-up scenes. 

Quality of life

High quality of life usually depends on a lot of different factors such as the cost of living, how walkable a place is, the availability of health and child care, the quality of the internet connection, crime rates, the weather, the availability of housing, and much more. 

There are a number of indices that rank cities according to a weighted sum of many factors such as the mercer city ranking, the where-to-be-born index and the most livable cities overview. While there is some disagreement between different rankings, they often rate the current EA hubs outside of their top ten and their top cities usually have a smaller EA presence (e.g., Tokyo, Auckland, Adelaide, Vancouver, Vienna). Of course, EAs could just have very different preferences but it might still be a good starting point to think about these questions from first principles. 

The availability of healthcare differs strongly for trans people around the world. Since a non-trivial amount of the wider EA/rationality community is trans, it makes sense to look into this in more detail before building up the hub. 

Economic regulatory framework

There is a growing demand for megaprojects in EA. In many countries, the laws for NGOs are made for smaller institutions and do not really allow for these large-scale endeavors. Therefore, either private companies (e.g. FTX, Alvea(?)) or private-public hybrids (e.g. Anthropic, Redwood Research) might become more and more prominent in EA. Thus, the specific economic regulatory framework matters a lot. It is important to be able to set up these hybrid companies in the first place, to be able to hire international workers easily, to allow for international investors, to not be hindered by too much red tape, etc.

Our hunch is that there will be many such projects in the near future and thus this should be an important consideration when founding a new hub or growing an existing one. 

Governance institutions

Influencing governments is a rapidly growing priority within the EA community. Therefore, a potential hub should be close to institutions that matter. These can be governments, supranational bodies like the EU, or global institutions like the UN. 

Additionally, it might be important to be in places that have a relevant influence on the AI landscape such as the US and China. With more powerful AI systems, countries will have to create legislation and it is important that EAs play a part in the creation of such laws. 

Research environment

Good research requires good institutional support and an environment that supports researchers and allows them to focus solely on their mission. This environment is often influenced by local institutions such as universities or companies, e.g. it is easier to do good research in Oxbridge or DeepMind than a small underfunded institution. Concrete things to look out for include research spending as a percentage of GDP, ease of receiving grants, the potential for collaborations with high-impact non-EA researchers, or the conditions of concrete institutions (e.g. do they require you to publish X papers every year). 

However, we think the research environment is less important than the economic framework and governance since research seems to depend more on having the right people at the right place than the exact regulatory setup. Thus, it is more important to create a good environment for the researchers once there is a hub, rather than building the hub around some local research institution. 

Easy immigration & visa

EAs come from many different countries and often quickly need to get into another country to start a new position. The more complicated the immigration and visa process, the more we artificially reduce the talent pool. Immigration for high-skilled workers between western, rich nations is probably sufficiently easy (even though it could be better) but getting talent from the developing world (which contains the vast majority of people) to western countries is still an annoyingly complicated process. 

Timezone

There is a lot of cooperation between EA organizations around the globe and timezones matter more than one might expect. Since the two biggest hubs are either in the Bay Area time zone or London time zone, high overlap with those is likely desirable. 

Specifically, time zones beyond ~UTC+6 (East Asia, Australia, New Zealand) often have scheduling problems with many EA organizations. 

Language

Especially when trying to influence governments, language matters more than we like. Governments are often run in the country’s official language and policymakers might not like to interact with people that only speak English. 

Places where this is especially prominent include Japan, Italy, France, China, and many more. Therefore, for governance endeavors in these countries, it is especially important to have people who speak the local language very well and can build a trustworthy bridge to powerful institutions. 

Beyond the interaction with governments, living in a place where most people speak a language you know is also much more convenient for interaction with ordinary bureaucracy and health care and social activities like sport or other hobbies. 

Great Power War

With Putin’s war on Ukraine and China–Taiwan tensions, the probability of great power wars and nuclear war is non-negligible. Metaculus estimates 21 % chance for a third world war before 2050 and a 6 % chance for fatalities from a non-test nuclear detonation by 2024, and you can find a simple back-of-the-envelope calculation by Ben Todd on how endangered specific places are. 

A detailed look at the safety of different places can be found here. The consensus is that important international centers such as Washington DC, London, Paris, Berlin, etc., and (potential) conflict zones such as Eastern Europe and Taiwan are especially endangered. 

Places that are either mostly neutral such as Switzerland or far away from potential great power wars (like Latin America, Australia and Africa) likely have a low probability of being hit by nuclear weapons. 

Suggestions

From obvious to less obvious. Current hubs exist for a reason but they have some downsides such as high cost of living or being focused on a small number of Western countries.

• Bay Area: The upsides of the Bay area are manifold. The environment for startups is great, there is close proximity to many AI labs, the weather is usually pretty nice and there is a large international talent pool. 
  The downsides include a very high cost of living, non-walkable cities and potentially high cost of raising children.
• London: London has a great economic and cultural environment. Many EA and EA adjacent organizations such as 80K, CLR or DeepMind have their main offices in London and the EA scene is pretty large. 
  The downsides are mostly due to the high cost of living and potentially negative consequences of Brexit.
• Oxford/Cambridge: Both are known for their top universities. Furthermore, many EA organizations such as FHI or CEA are already there.
  Downsides include high cost of living, drastically increased costs for international students due to Brexit and generally really shitty salaries for researchers in the UK.
• Boston: We think that Boston might be the go-to biosecurity hub in the near future. With the lab of Kevin Esvelt at MIT and Telis/Alvea, there are two cutting-edge teams working hard to reduce the impact of biorisks. 
  Downsides include cold weather and the usual downsides of American cities.
• Washington: It’s where the US government and all the think tanks are. Unfortunately, there is not much else either.
• Toronto: The University of Toronto and associated institutes are among the best, if not the best place, for AI. Furthermore, Canada’s immigration system is very good, especially for high-skilled workers. Downsides include cold winters, high cost of living and not much to gain for non-AI people.
• Berlin: As the capital of Germany, Berlin is one of the best places to navigate European and obviously German politics. The city has a really high quality of life and is cheap for international standards. The biggest downside of Berlin is probably that for everything but influencing German politics there are better-specialized candidates.
• Paris: Behind Germany, France is the second most influential country in EU politics and Paris is much more important to France than Berlin to Germany. It’s probably still less relevant than Berlin or Brussels though. Downsides include the high cost of living.
• Prague: The case is made in a comment (see below). It has an existing very active community, it offers relatively high quality of living at low cost and there are some exciting ventures in the EA space.
• Brussels: It’s where the EU decides things and all the think tanks are. Similar to Washington, there’s not much else either.
• Zurich/Geneva: Switzerland has the highest standard of living worldwide, a good public school system and high wages. Geneva is interesting because of the UN office. Zurich is interesting for the high research quality of the ETH, especially in AI, and a host of companies such as Google. One downside is the high cost of living. Especially for governance-related tasks, Switzerland might not be very relevant because it doesn’t have a lot of influence on international politics.
• Amsterdam/The Hague/Copenhagen: The Netherlands and Denmark have a very high quality of life with walkable, bike-friendly cities and very good childcare and education. We think they would be very good locations for young families with remote jobs. Both of them have decent paths to immigration and have reasonable start-up laws. They are small countries so governance might not create much of a lever.
• Lisbon: Since the beginning of Covid, Lisbon has become one of the most desirable destinations for remote work. It’s part of the EU, has a low cost of living for a relatively high quality of life and good weather. It has good weather and better timezones for the US than central Europe. Lastly, it has very relaxed crypto laws which might be relevant for some EAs. Downsides include the limited importance of the Portuguese government and a lack of specialized EA reasons in general.
• Tel Aviv: Israel has a huge and still growing tech and start-up scene. Universities throughout Israel provide a large pool of talent. The EA scene is relatively new but booming. The weather and food seem pretty nice. 
  Like all tech hubs, costs of living are soaring and Israel has a higher chance of severe conflict than most other places on earth.
• Tokyo/Seoul: Over half of the global population lives in Asia and yet EA presence is very limited there. Tokyo and Seoul are both large capitals of important countries in the region. The foreign policy of South Korea and Japan is important to avoid a great power war in the Asia-Pacific. 
  It is likely necessary to speak the local languages to get anything done and Asian cities might not be for everyone.
• Singapore: Singapore has been called ‘the Switzerland of Asia’. Due to the government’s good relationship with both the US and China, we can expect Singapore to be a very important actor in Asian diplomacy. Our knowledge about Singapore is limited, but it might be easier to influence by competent outsiders than other Asian countries. Freedom of speech and sexuality are limited.
• Taiwan: Living in Taiwan can provide you with some understanding of the Chinese language and culture without the downsides of the mainland. Due to its special status, the Taiwanese government has very little influence on Asian geopolitics. On the other hand, influencing Taiwan’s defense policy could be quite impactful. Taiwan is the center of advanced semiconductor manufacturing. Due to the enormous importance of compute for AI, gaining influence in companies like TSMC could be impactful. The biggest downside is the risk associated with a Chinese invasion. Metaculus currently assigns a 59% probability of Chinese annexation of most of Taiwan by 2050.
• Beijing/Hong Kong/Shanghai: While there are some institutions and individuals working on China it still is too neglected within EA in our opinion (possibly EA research on China is just not public). Hong Kong and Shanghai might both be good cities to understand the country better while still having a somewhat Western lifestyle. Beijing is closer to power and has a more traditional Chinese lifestyle. However, all come with high risks. If the CCP doesn’t like what you’re saying you might run into considerable difficulties or just end up in prison.
• Mexico City/Sao Paolo: Mexico and Brazil are populous and reasonably important countries in global politics. Both already have a growing EA scene.
  We find it hard to estimate the economic and political conditions from the outside. Crime rates are high and the quality of life is worse than in most Western countries.
• Lagos: There are 1.2 Billion people in Africa (and growing rapidly) and Nigeria is its economic powerhouse. There is a growing tech scene and Lagos becomes more and more influential, especially in West-African politics. The EA group in Lagos is currently inactive and it would likely require more buildup. African politics are very important for global health and wellbeing, so it makes sense to have a hub there.
• Tübingen: Disclosure: two authors (Marius, Yannick) of this post are currently based in Tübingen. The city + surrounding region is trying to become the center for AI in mainland Europe. The ML/AI faculty of the University and Max-Planck Institutes is huge (already biggest in Europe by some metrics) and the city tries hard to build a start-up culture. However, Marius thinks that the current research focus is far away from AI safety and is frustrated with the narrow-mindedness of parts of the faculty when it comes to alignment (with some exceptions; PM me if interested). In many ways, Tübingen might just be the worse version of Zurich (from an alignment perspective) unless it sees a major shift in research focus over the coming years. On the other hand, it already is a major AI hub and a strong EA presence might thus be relevant in any case.

While our list of suggestions might seem long, there are a ton of major cities around the world that we have explicitly omitted from it. We think, for example, that many US cities wouldn't be ideal for a hub since the downsides of American city design outweigh the benefits unless there are important private or public actors to counter-balance it. 

First steps

If you are interested in setting up a new hub or want to contribute to an existing one we have some vague suggestions:

• Contact the local EA community (if they exist) and discuss your plans with them. They know more about the city, its institutions and culture.
• Talk to CEA or other community-building organizations. Grantmakers such as OpenPhil or the LTFF offer grants for community building.
• Set up an EA house, e.g. buy or rent a large flat or old Hotel.
• Set up a coworking space
• Make it really easy to visit/stay for a while. Make sure that people know of this option and potentially provide funds for visitors.