See the paper He Who Pays The Piper Must Know The Tune and my review of the book Power and Prosperity for more arguments that led me to this belief.

To try to avoid perpetuating this problem, I try to focus on organizations whose results I can evaluate, but I'm still concerned about the subjective nature of my evaluations, and am trying to look for better metrics, hopefully something approaching the objectivity of the accounting system used to measure corporate profits.

Also, I try to focus on organizations that are small enough that they don't develop much of a self-perpetuating bureaucracy.

Here are the charities in which I have some hope:

• Electronic Frontier Foundation
  Defending freedom in cyberspace. If you're unfamiliar with the arguments it makes, you shouldn't assume you're informed enough to vote in the U.S.
• Methuselah Mouse Prize
  Dedicated to curing aging. Unlike most requests for health oriented research money, this is an information prize. It doesn't require the donor to have the expertize to distinguish in advance whether the research is promising (since even researchers have trouble knowing whether their research is valuable, it's unlikely that many donors are able to make even an educated guess at that). Instead, all a donor has to evaluate is whether the results it rewards are connected to something medically desirable. And a cure for aging would certainly be one of the more important medical advances that I can imagine, as it should reduce deaths from heart attacks and cancer to roughly the rates seen in 20 or 30 year olds.
  See the Strategies for Engineered Negligible Senescence web page for some serious discussion about the possibility of curing aging.
• Trickle Up
  Loaning small amounts of money to a large number of entrepeneurs who couldn't get a commercial loan because the transaction costs are large compared to the interest payments, and live in a poor enough community that they can't borrow $100 from friends or family to start their business.
• Foresight Institute
  Educating the world about the benefits and risks of accelerating technological change, particularly molecular nanotechnology.
• Transgender Law Center
  This group deals with less important problems than most charities I notice, but it deals with problems that relatively few people are aware of and where small amounts of effort make important differences to victims of the legal system's false assumptions.
  Typically I discount reports about a charity being valuable because the manner in which I get those reports is evidence that they are memes that propogate better than the reports I don't get, which is evidence that the charity gets more money than it would if we all carefully evaluated all charities. But I heard of the Transgender Law Center in an accidental manner (meeting one of the founders on a hike, and even then only hearing about the Transgender Law Center because someone with him mentioned his role in it) which has little apparent connection to the popularity of its meme. Of course, if you're using this report as evidence, you can't say the same about it. But I encourage you to apply the same reasoning to evidence you get.
• Singularity Institute
  This small group hopes to save the world from an AI whose goals don't coincide with those of human beings, by creating an AI or something roughly equivalent whose goals are demonstrably friendly to humans. But it's quite hard to decide whether this is an important charity in fairly urgent need of money, or whether it is a utopian pipe-dream.
  The track record for people who set out to save the world isn't very good, and there's no good way to check whether Eliezer is competent enough to be the first to create an AI. So there's no reason to think this project has a high probability of success, regardless of how much money it gets. But unlike the typical save-the-world dreamer, Eliezer abandons beliefs frequently enough that much of what he wrote a few years ago is obsolete, and his opinion of himself isn't nearly as inflated today as it was 5 years ago.
  But the concerns that the Singularity Institute is worried about are scary, and even if Eliezer is exaggerating the risks by an order of magnitude, raising his chances of success from 2% to 2.1% would be one of the best possible uses of money.
  I suspect most people treat AI as if the best analogy to use in starting to evaluate it is to treat it as a typical new piece of software or a typical new technology. The methods being used to produce AI, and the results produced by researchers who describe their field as AI, certainly suggest this is a plausible analogy. And the implication that it is safe to wait until the technology has been demonstrated to evaluate its risks has worked rather well in the past.
  But I fear that better analogies would be the times when a species has evolved information processing abilities that are qualitatively superior to anything that other species have. For instance, when homo sapiens acquired language abilities more powerful than what other apes had, this may have caused the extinction of Neanderthals, Mammoths, etc. And the many species of Ediacara that may have been the dominant animals 550 million years ago seemed to have disappeared rather abruptly when the first trilobites to evolve eyes starting tracking them down (see Andrew Parker's book In The Blink of an Eye for more on this). If these are good analogies for AI, then repeating strategies that have worked for us in the past will only work well if we become the first intelligences that can improve at Moore's Law rates or faster - something that sounds hard enough that I don't want to rely on being able to do it.

Last updated 2008-08-15.