This is something I originally wanted to send to Jack_H, but I think it makes a potentially useful post. Jack asked me if there's anything I would add to Linch's RP work trial on aging. The post below is my answer. I haven't crafted this post extremely carefully and there might be a lot of important things missing (and some grammatical errors), but I figured it would be best to post it here rather than just in a FB chat.
The two areas Linch identified that I find most promising
I don't have in mind other large areas where to direct effort, but I have in mind a few specific points for how to direct research inside the areas he proposes:
The two areas I find most interesting are: "fundamental research on how to think about the effects of anti-aging" and "intervention research on how to do anti-aging".
Regarding the area "fundamental research on how to think about the effects of anti-aging"
To me, one of the most interesting questions to try to answer is similar to what I discussed with Heye in a call with him. Matthew Barnett summarizes it in "A mental shift among both elites and regular citizens about the best way to prepare for the future. Here, I imagine that politicians and other elites would regularly talk about the future thousands of years hence because it's reasonable that people will be around that long".
Basically, we need social science research to answer if longer lives change what people care about. Willbradshaw also wrote a comment in which, among other points, he explains why he thinks that anti-aging needs more social science research if we want to answer longetermism-relevant questions.
A very specific point that ought to be evaluated is this: if people will worry more about x-risk if they don't die from aging anymore, then this is a point in favor of trying to get anti-aging before technologies that pose an existential risk to humanity are developed (see Bostrom's arguments for differential tech development, etc.). Solving aging would align egoistic and altruistic instrumental values: x-risk might become the most significant risk of death even for the individual, not only for civilization.
But this kind of reasoning has a problem that I identified under Matthew Barnett's post, although people didn't seem to be receptive to it (maybe because it's incorrect? I don't know). The point is this: real world-interventions on anti-aging research mainly speed research up or slow it down. A difference of just a few years is very unlikely to influence differential tech development. Therefore even if we get an answer about such things (e.g. if people start caring about x-risk and the long-term future once they can live for thousands of years or more), it could be useless. But I might be wrong here, I haven't thought about it for long enough.
Longtermist interventions in the anti-aging space that don't suffer from that argument might be more promising. For example: preemptively influencing policymaking in the developed world to ensure that anti-aging research brings about a good future (mainly: make the transition to a post-aging world go smoothly). So we don't get locked in some kind of bad attractor state.
Regarding the area "intervention research on how to do anti-aging"
What I think would be useful is this:
- We already know that NIH and NIA spending on aging research is vastly inefficient, but we need a good writeup on this. Something that analyzes all of their grants as opposed to what would actually be helpful. The real field of aging research looks much smaller after we remove all the obviously misguided research.
- We need to identify more promising underfunded labs and potentially even labs that might be shut down for stupid reasons such as feuds between academics (point suggested to me by gavintaylor years ago).
- We need to have a clear "roadmap" of what basic research to finance if we want to finance it. I would build it like this: consider each of the hallmarks as a bottleneck until we have further info on what is most likely a bottleneck. Or if we already have a pretty good sense of what hallmarks constitute bottlenecks to putting aging under medical control, then consider only these hallmarks. In addition to hallmarks, we should include other areas that cut across them that are particularly important, evaluated separately. Every hallmark/area must be associated with neglectedness and tractability scores (scope is not relevant if we only use bottlenecks, but it is relevant for stuff that cuts across) and a complete list of groups that work on them. Lifespan.io's Rejuvenation Roadmap would help for this, but it's probably not complete.
But wait, there's more!
Finally, some work that needs to be done that cuts across all this (both longtermist considerations and other considerations): various guesstimate models evaluating various kinds of interventions based on different models of impact. The interventions might be basic research, trials, advocacy, policy change, social science research, etc.
EDIT: Another point I should write about here. This is an answer to this comment.
How would you answer the following arguments?
Existential risk reduction is much more important than life extension since it is possible to solve aging a few generations later, whereas humankinds potential, which could be enormous, is lost after an extinction event.
From a utilitarian perspective it does not matter if there are ten generations of people living 70 years or one generation of people living 700 years as long as they are happy. Therefore the moral value of life extension is neutral.
I am not wholly convinced of the second argument myself, but I do not see where exactly the logic goes wrong. Moreover, I want to play the devils advocate and I am curious for your answer.
My answer:
1. Yes, this is probably true. But see longtermist considerations of effects of anti-aging research. They might be in the same ballpark. Or not.
2. There are three ways in which the impact of anti-aging research is evaluated: DALYs averted and other short-term considerations, LEV being brought closer in time, and effects relevant to the long-term future. All three don't suffer from this objection.
EDIT 2: One other potentially promising idea, suggested to me by Matthew Barnett in private, is to use prediction platforms such as Metaculus to try to predict clinical trials outcomes (perhaps also on aging itself or important markers, rather than necessarily on trial endpoints?). One other way to use platforms such as Metaculus is to get a sense of what that community thinks regarding the long-term effects of anti-aging, before (and after) getting social science data.
