Some of Ord’s estimates have data to support them – we know a little about past supervolcano eruptions and asteroid impacts, for example – but the threats we are The most feared are also the threats closest to pure speculation. That doesn’t mean the speculation is wrong; I think synthetic biological weapons are scary and the risks of artificial intelligence are worth taking seriously. But I think that false precision can make it feel unreal and mask the tremors of all that happened before quantification.
To me, this is the clearest link between the Bankman-Fried collapse and some elements of a culture of effective altruism: Cryptocurrencies are built on tying values and probabilities to assets and nominal currencies. What looks like a balance sheet from one angle proves to be nothing more than a set of arguments, assertions, and imaginary tests from another.
This problem exists elsewhere in capitalism, but it is concentrated in crystal form in the cryptocurrency market. Often, the only real value of a crypto asset is self-referential: The asset is valuable because it is valued. There is nothing behind it except that other people are willing to believe the story you are telling them. There is no army or automobile factory or even beloved artwork. Only codes and quantifiers. The numbers are showing undue solidity for an abstraction. I think effective altruists also tend to fascinate themselves in the same way.
We can do much better, and we should
I worry that this column will be seen as a reason to dismiss any concerns that may sound odd the first time they are heard. That’s not what I’m saying.
Effective altruists fought hard to convince people to worry more about artificial intelligence, and they were right to do so. I don’t think you can look at the exceptional performance of the latest artificial intelligence models — Meta create Cicero, an AI system that can manipulate and fool humans to achieve other goals, and OpenAI’s newest bot will advise you on how to create nuclear weapons if you ask smart enough — and don’t think it’s important to consider the consequences of much more powerful AI systems.
But I think too much of the energy and talent in effective altruism is drifting away from the compassionate rigor that originally distinguished it, and what the world still needs. Effective altruism is nearly as ineffective if it loses its original focus on improving the lives of those living today.
All of this brings me back to GiveWell. GiveWell was set up to evaluate and even present that proof, and it has done an excellent job. Its research is comprehensive, insightful and most importantly transparent. I disagree with every decision it makes — I think it sets the bar for cost-effectiveness too high and some charities have been dropped from its list, like Direct donation, is still mine – but it’s persistence that matters. I donate to GiveWell’s charities every year and while it’s not the entire amount I donate, it’s the part I feel most confident in. Donation organization I’m sure it’s a good feeling, and I hope you feel it too.