I see Singularity Institute and Future of Humanity Institute as quite nicely complementary:

  • FHI is part of Oxford, and thus can add great credibility and funding to existential risk reduction. Resulting output: lots of peer-reviewed papers, books from OUP like Global Catastrophic Risks, conferences, media appearances, etc.

  • SI is independent and is less constrained by conservatism or the university system. Resulting output: Foundational works on Friendly AI (e.g., CFAI was just so much more advanced than anything else in machine ethics in 2001 it's not even funny), and the ability to do weird things that are nevertheless quite effective at creating tons of new people interested in rationality and existential risk reduction: (1) The Sequences, the best tool I know for creating rational transhumanists, (2) Harry Potter and the Methods of Rationality, now the most-reviewed fan fiction work of all time on FanFiction.net, (3) Less Wrong in general, a growing online and meatspace community of people who care about rationality and x-risk reduction, (4) Singularity Summit, a mainstream-aimed conference that nevertheless brings in people who end up making significant academic contributions like David Chalmers, along with a broader support base for transhumanism in general, (5) lots of behind-the-scenes work with Giving What We Can and 80,000 Hours that has encouraged the optimal philanthropy community in general to take existential risk reduction seriously.

  • FHI is in Britain, SI is in the USA.

  • FHI and SI have different but overlapping goals. FHI investigates a much broader range of transhumanist topics, whereas SI is more focused on AI risks and the rationality skills needed to think correctly about them.

I'm currently glad I work for the org with greater flexibility (SI), but I am quite happy with FHI's rather incredible productivity. I don't know any other philosophy research institute that has published so much original and important work in such a short time, ever.

edit 12/01/2011: Added point (5) in the list above, about optimal philanthropy.

Small correction: HP:MoR is the most reviewed Harry Potter fanfiction of all time. I did a quick search for Twilight fanfiction and found a story with 14,822 reviews compared with 14,710 reviews for HP:MoR. I wouldn't be surprised if that were the number to beat (twilight is a very popular fanfiction topic as far as I know) but due to the disorganization of fanfiction.net, it's hard to say.

SIAI vs. FHI achievements, 2008-2010

by Kaj_Sotala 4 min read25th Sep 201162 comments


After reading the FHI achievement report for 2008-2010, I thought it might be useful to compare their achievements to those of SIAI during the same time period. Since SIAI does not have an equivalent report, I've mostly pulled the data of their achievements from the SIAI blog.

My intention here is to help figure out which organization makes better use of my donations. For that purpose, I'm only looking at actual concrete outputs, and ignoring achievements such as successful fundraising drives or the hiring of extra staff.

For citation counts, I'm using Google Scholar data as-is. Note that this will include both self-cites and some cites from pages that really shouldn't be counted, since Google Scholar seems to be a bit liberal about what it includes in its database. I'm unsure as to whether or not the citation counts are very meaningful, since there hasn't been much time for anyone to cite papers published in 2010, say. But I'm including them anyway.

Future of Humanity Institute

Publications. The Achievement Report highlights three books and 22 journal articles. In addition, FHI staff has written 34 book chapters for academic volumes, including Companion to Philosophy of Technology; New Waves in Philosophy of Technology; Philosophy: Theoretical and Empirical Explorations; and Oxford Handbook of Neuroethics.

The three are the hardcover and paperback editions of Human Enhancement, as well as a paperback edition of Anthropic Bias: Observation Selection Effects in Science and Philosophy. Human Enhancement has been cited 22 times. Anthropic Bias was originally published in 2002, so I'm not including its citation count.

The highlighted 22 journal articles had been cited 59 times in total. The overwhelmingly most cited article was Cognitive Enhancement: Methods, Ethics, Regulatory Challenges in Science and Engineering Ethics, with 39 cites. The runner-up was Probing the Improbable: Methodological. Challenges for Risks with Low Probabilities and High Stake, with 5 cites. The remaining articles had 0-3 cites. But while Cognitive Enhancement is listed as a 2009 paper, it's worth noting that the first draft version of it was posted on Nick Bostrom's website back in 2006, and it has had time to accumulate cites since then. If we exclude it, FHI's 2008-2010 papers have been cited 20 times.

It's not listed in the Achievement Report, but I also want to include the 2008 Whole Brain Emulation Roadmap, which has been cited 15 times, bringing the total count (excluding Cognitive Enhancement) to 35.

Presentations. FHI members have given a total of 95 invited lectures and conference presentations.

Media appearances. Some 100 media appearances, including print, radio, and television appearances, since January 2009. These include BBC television, New Scientist, National Geographic, The Guardian, ITV, Bloomberg News, Discovery Channel, ABC, Radio Slovenia, Wired Magazine, BBC world service, Volkskrant (German newspaper), Utbildningsradion (Swedish national radio), Mehr News Agency (Iranian), Mladina Weekly (Slovenian magazine), Jyllands-Posten and Weekenavisen (Danish newspapers), Bayerisher Rundfunk (German radio), The History Channel, O Estado de São Paulo (Brazillian newspaper), Euronews, Kvallsposten (Swedish newspaper), City Helsinki (Finnish radio), Focus, Dutch Film and Television Academy, The Smart Manager (Indian magazine), Il Sole 24 Ore (Italian monthly), The Bulletin of the Atomic Sciences, Time Magazine, Astronomy Now, and Radio Bar-Kulan (Kenya).

Visitors. "The Institute receives many requests from students and scholars who wish to visit the Institute, only a few of which are accepted because of capacity limitations. The FHI has hosted a number of distinguished academic visitors over the past two years within its various areas of activity, such as Profs. David Chalmers, Michael Oppenheimer, and Thomas Homer-Dixon."

Policy advice. The Achievement Report highlights 23 groups or events which have received policy advice from either Nick Bostrom or Anders Sandberg. These include the World Economic Forum, the Public Services Offices of the Prime Minister's Office of Singapore, the UK Home Office, If (Stockholm insurance company), Jane Street Capital, IARPA (Intelligence Advanced Research Projects Activity) for US Government, The Swedish Institute for Infectious Disease Control, and setting up a research network, "A differential view of enhancement", within the Volkswagen Foundation.

Organized events. Three organized events. 1: Cognitive Enhancement Workshop. 2: Symposium on cognitive enhancement and related ethical and policy issues. 3: Uncertainty, Lags and Nonlinearity: Challenges to governance in a turbulent world.

Singularity Institute

Publications. The SIAI publications page has 15 papers from the 2008-2010 period, of which 11 are listed under "recent publications", 1 under "software", and 3 under "talks and working papers". Of these, Superintelligence does not imply benevolence has been cited once. The rest all have no citations.

The Sequences were written during this time period. They consist of about a million words, and might very well have a bigger impact than all the other FHI and SIAI articles together - though that's very hard to quantify.

Presentations and Media Appearances. The SIAI blog mentions a number of media appearances and presentations at various venues, but I don't have the energy to go through them all and count. From a quick eyeballing of the blog, though, SIAI has nowhere near as many presentations and media appearances as FHI.

Visitors. The Visiting Fellows page has a list of 27 Visiting Fellows from around the world, who attend or hold degrees from universities including Harvard, Stanford, Yale, Cambridge, Carnegie Mellon, Auckland University, Moscow Institute of Physics and Technology, and the University of California-Santa Barbara

Online communities and tools. Less Wrong was founded in 2009, and Google Analytics says that by the end of 2010, it had had over a million unique visitors.

Note that LW is an interesting case: as an FHI/SIAI collaboration, both organizations claim credit for it. However, since LW is to such a huge extent Eliezer's creation, and I'm not sure of what exactly the FHI contribution to LW is, I'm counting it as an SIAI and not a joint achievement.

SIAI also created the Uncertain Future, a tool for estimating the probability of AI.

Organized events. SIAI held Singularity Summits on all three years. The first Singularity Summit Australia was held in 2010. In 2008, SIAI co-sponsored the Converge unconference.

Ben Goertzel, acting as the SIAI Director of Research at the time, organized the 2008 and 2009 conferences on Artificial General Intelligence. He also co-organized a 2009 workshop on machine consciousness,

Artificial Intelligence projects. SIAI provided initial funding for the OpenCog project, as well as sponsoring Google Summer of Code events relating to the project in 2008 and 2009.


Based on this data, which organization is more deserving of my money? Hard to say, especially since SIAI has been changing a lot. The general AGI research, for instance, isn't really something that's being pursued anymore, and Ben Goertzel is no longer with the organization. Eliezer is no longer writing the sequences, which were possibly the biggest SIAI achievement of the whole 2008-2010 period.

Still, FHI's accomplishments seem a lot more impressive overall, suggesting that they might be a better target for the money. On the other hand, they are not as tightly focused on AI as SIAI is.

One imporant question is also the amount of funding the two organizations have had: accomplishing a lot is easier if you have more money. If an organization has thrice as much money, they should be expected to achieve thrice as much. SIAI's revenue was $426,000 in 2008 and $628,000 in 2009. FHI's funding was around $711,000 for 10/2008 - 10/2009. I don't know the 2010 figure for either organization. The FHI report also says the following:

To appreciate the significance of what has been accomplished, it should be kept in mind that the FHI has been understaffed for much of this period. One of our James Martin Research Fellows, Dr Rebecca Roache, has been on maternity leave for the past year. Our newest James Martin Research Fellow, Dr Eric Mandelbaum, who was recruited from an extremely strong field of over 170 applicants, has been in post for only two months. Thus, for half of the two-year period, FHI’s research staff has consisted of two persons, Professor Nick Bostrom and Dr Anders Sandberg.

SIAI - An Examination notes that in both 2008 and 2009, SIAI paid salaries to three people, so for a while at least, the amount of full-time staff in the two organizations was roughly comparable.