These kinds of talks are one of the primary ways people are introduced to Effective Altruism - it seems very useful to spend some time crafting them to produce the most effective talk we can. So it would be great if people could critique this talk. I'm sure Toby wouldn't mind. To get the ball rolling, here are some of my thoughts:
ps. Thank you for transcribing this!
"Is explaining QALYs going to turn people off?"
QALYs are a major component of this talk. You could make another talk without QALYs, but a lot of pieces of this one, like demonstrating the huge range of effectiveness in health interventions, depend on them.
"The presentation could be more engaging - less on each slide, more colorful"
I didn't get the sense that people in the room were tuning out or would have paid more attention with flashier slides. But this probably varies based on where you are.
"Thank you for transcribing this!"
I actually only cleaned up the transcription. The initial transcription was made by someone else because one of the people who wanted to watch the talk is deaf.
Jeff, did you do the transcription yourself? If so, massive kudos. (I really prefer reading to watching videos when it comes to this sort of thing, although I seem to be in the minority)
I think this talk does a surprisingly good job of walking through a lot of important (and potentially cold, dry) concepts in a way that makes sense.
Someone else did a transcription at google, I think because a deaf employee wanted to watch the talk. I cleaned it up, fixed typos, added links, screenshots, and paragraph breaks.
So If I donate 2 million USD over the course of my life and I expect to save around 2000 lives this clearly loses out to causing human uploading to come even 1 hour earlier.
Two problems:
I expect the cost of saving a life to go up in the future. So traditional charity will probably result in fewer lives saved. At what rate should I expect the cost of saving a life to change? Does Givewell attempt to graph this?
I have to figure out whether 2 million donated over the next several decades can plausibly cause uploading to happen sooner.
I have to figure out whether 2 million donated over the next several decades can plausibly cause uploading to happen sooner.
This is a problem I am extremely interested in. Please PM me and/or write a post if you come up with anything.
When uploading is first available, I expect it to be expensive, and for the available hardware to be limited in capacity. I assume $2M would make a noticeable impact on the uploading rate available. (Even if capacity grows rapidly due to singularity effects, I suspect this would be slow on the time scale of hours, unless some sort of manufacturing magic happens before uploading does.)
So you should trivially be able to bring about faster uploading if you're willing to set up a charity with long term planning skills to do so. How much uploading bandwidth do you expect $2M to buy?
When uploading is first available, I expect it to be expensive, and for the available hardware to be limited in capacity.
So do you expect uploading in the next several decades, or big slowdowns in the improvement of computation per dollar?
I expect most new things to first be available in small quantity at expensive prices and low quality, especially if you're literally talking about putting them to use in the first hours or days after they become possible. In fact, no non-trivial examples that involve any specialized hardware at all come to mind. Am I missing any major ones? Are there major reasons uploading would be different?
Why expect specialized hardware to be a bottleneck? Most software programs don't require very specialized hardware.
If computation per dollar continues on trend to far surpass minima for brain emulation, then I would expect final bottlenecks to be in brain-scanning or (more likely) understanding the neuroscience, not shaving cycles off the computation of Hodgkin-Huxley.
I can think of a whole lot of unlikely ways by which keeping your money and not donating it, or doing even worse things, increases the welfare of people around the world. For instance, maybe a society where people like myself are richer encourages people in other societies to improve their societies because they see how rich they could be, to a greater degree than if people in my own society are not as rich.
Or maybe when I keep my money rather than donating to charity, the average person in my society is incrementally richer, that incrementally reduces the population in my society (because of the demographic transition), and makes people in my society better off by an incremental amount that may still be quite a lot when summed over the whole society.
Maybe me being richer, and thus the United States being richer, incrementally increases the prestige of the US on an international level, incrementally discouraging North Korea from starting a war that kills millions of people.
Alternatively, maybe a war with North Korea is inevitable but if it starts sooner rather than later fewer people will be killed (since populations will be larger in the future) and the surviving descendants will have more man years of freedom. It is conceivable that I should prefer starting a war with North Korea to stopping malaria.
Maybe I could promote ideas and policies that make Bill Gates richer, on the grounds that making him richer also increases the amount he'll donate through the Bill Gates foundation. It is even conceivable that doing this is more efficient than just giving money to the Bill Gates foundation directly, or convincing other people to do so.
Have you tried to make rough order-of-magnitude estimates for the effectiveness of these approaches?
Small effects multiplied by large populations, or small probabilities multiplied by large benefits, are very difficult to estimate and any estimates I can come up with have such large error bars that they are useless.
I should add that I think some of the ideas in the original post have this problem too. For instance, how would you calculate the magnitude of the effect by which saving infant lives encourages the demographic transition and thus reduces overpopulation?
Video: (youtube)
Transcript: