You (and @tylermjohn) might be interested in Diffractor's Unifying Bargaining sequence. The sequence focuses on transferable utility games being a better target than just bargaining games, with I believe Nash being a special-case for bargaining games. As well as talking about avoiding threats in bargaining and trying to further refine. I think the defaults won't matter too much. Do you have any writing on the moral parliament that talks about the defaults issue more?
I generally agree with this post that it is a good target. This is effectively equivalent to Coherent Extrapolated Volition[1], which has been proposed as a target for alignment. The difference is whether you consider extrapolated values by default or not.
Just to be clear, my intended belief there was a 5-15% of meat industry taking dominance for 2-3 decades after the introduction of an alternative that is very close or beats meat entirely. Though I do think it is plausible the alternatives come soon, AI technology is advancing rapidly; and I don't believe many people's models are properly factoring in AI technology of our current level being applied to more areas, much less what we'll have in the future after significant advancements. Of course, EA tends to be a lot better at that than other charities.
As an example: Theorem proving is bottlenecked by the annoying but solvable triplet of: data, money to train larger models, and companies focusing on it. Scaling the current methods would hit noticeable limits due to planning/search being hard, but would allow a lot of automation towards proving software correct. AlphaProof itself is then a step above the methods that came before it. This could provide a good amount of value in terms of ensuring important software is correct but is generally ignored or assumed to need massive breakthroughs.
I find it plausible that more systems in the vein of AlphaFold (protein prediction, most centrally relevant to meat) can be extended to other areas of chemistry with a significant amount of time & effort to collect data and design. There's big data collection problems here, we have a lot of data about food but it is more locked away inside companies and not as carefully researched to a low level as proteins. I know the theorem proving better than I do the AlphaFold area, but that gets across my general view of "many mental models assume too much like we are in 2018 but with single isolated notable advancements like AlphaFold/AlphaProof/ChatGPT rather than a field with much to explore via permutations of those core ideas".
I believe it is entirely feasible to get the taste right. However, I don't believe that is a major problem. Even in the worlds where it is very expensive to get the texture exactly right, we do what many cultures have done over time and between themselves: we modify it and get used to it. Foods and other less literal tastes being so varying between cultures and even age groups makes me optimistic that even failed replications of taste/texture could replace meat simply through a change in generation where children see it as merely another food option. Though, admittedly, we are evolved to eat meat. This likely makes us more particular, yet we also prepare other foods like vegetables in exotic manners.
I don't see why you think if it was about efficiency we would already have switched. I'm somewhat confused: Current met production currently seems efficient based on people's eating habits, desire for meat, expectations about what is healthy (various people don't trust vegetarian answers for good and bad reasons), and most importantly our tech level. Is your argument that the meat industry is getting enough subsidies that they aren't truly more efficient than current alternatives? And/or that the government isn't requiring to price in the externalities of their effects on the land or climate? If they are actually less cost-effective (in terms of food produced) without the effective subsidies, that would be interesting information to learn, but I'd be somewhat surprised. It would actually make me more optimistic about the state of alternatives to meat, though I also understand that it would be a mark against my theory. (Just to be clear, I think transitioning will still take time. If we had gotten an instant win of better/cheaper/healthier alternatives back in 2015 without a slow buildup, that would have helped massively and things would have scaled up. I would expect a massive amount more beyond burger and such in stores by now, but I'd still expect meat for a while yet! Unfortunately we're in the world where it has become a somewhat politically polarized topic, and it isn't an obvious win to many consumers, which slows things down even if we had a definite better alternative available.)
I'm skeptical the meat industry survives in the current form. It is possible they drag out their existence for a long while, but since I expect cheaper more plentiful food from the artificial sources, that leads to a great method to out-compete them.
Just to be clear, I'm not arguing against donating to animal welfare, but I do see donations in this area as mostly bringing the time the transition occurs closer in most possibilities. Still very much worthwhile! Cutting a decade into five years gives a lot of value, and even setting the stage for when great alternatives exist is valuable.
Throwing a number out as a weak model, I'd say that about 5-15% of worlds where the meat industry manipulates government to strongly rely on factory farming for a significantly long time. A decade? Two decades? Three? In the other possibilities, I expect factory farming to limp along but be shifted out at varying speeds. They would still try to stifle alternatives, but not be in a dominant position. I think food companies are already interested in alternative meat products, which means you don't have a full cartel. Even in the 5-15%, I expect the meat companies to inevitably adopt the technology themselves even if they've choked out all the competitors. Not having the competitors is still very much a problem given how long they would delay the change. (Climate change, for example, is a harder problem because it requires more coordination and most people can't just substitute electricity into their car. Food is a lot more of a substitutable good.)
Hm, the post is good, though I see this as a relatively weak statement. The amount of factory farming would be massively lower than now, presuming that my assumptions about people switching off of meat are true.
I also expect that if we get a big boost of technology (even better AI-driven protein/chemical synthesis or discovery), then I don't expect the argument that we'll still farm them for dyes and such to hold. As the years go by, it becomes ever more feasible to synthesize those useful dyes or materials directly. The outcome described there would still be a lot less (several orders of magnitude?) factory farming. I do think this point depends notably on how soon you think the technology will occur and solve a lot of the general problem (getting chemicals/proteins en masse). I find it plausible that it will come before we solve various meat alternatives (in the better or equivalent price/taste/cost/nutrition sense), but also find it mildly plausible that it takes a decade or two after.
Point three of the article that AI will make factory farming more efficient is true, but also I don't have a reason to believe the final conclusion. Big data analytics does not provide notable evidence to believe that factory farming will outcompete alternative methods in the long run—it is an argument that they aren't constant and so have a longer shelf-life than the naive extrapolation. Growing an animal simply requires a lot of work and energy in a specific form that I don't have any reason to believe alternative meats require as much. As for the bioengineering example, a similar argument implies. To me, this is like saying that a person down in a mine is always going to be more efficient and scalable than a digging machine.
Point four is one that I think fills out a difference in vision. They compare it to chocolate which hasn't been replaced. My model of the world is that over the next thirty or so years we have significant advancement in fields like chemical synthesis, reverse engineering proteins, and so on. I don't see any reason to believe chocolate won't be replaced! Many of the foods made from it alter the taste massively themselves, if food companies could replace chocolate with a significantly cheaper version then they would. I believe they would be slower about it, it presumably doesn't cost much and has been quite tuned.
It feels like there's a missing note in the post, that they think things will stay comparatively the same as 2020 level competitiveness even though there's very little reason to believe that. Why should I think that growing a full animal is remotely efficient for any of our needs? There are some good points or areas to think about, and the weakest claim that there might be a notable amount of factory farming is true, but I feel overall a bit baffled by the thrust of the article. There's a missing note to the post, a strong background assumption that technology will continue looking as it is circa 2020 with no major optimizations feasible. AI isn't an integral part of my point, though I believe if it succeeds as it has so far then the timescale for various synthesis methods and the like moves up quite a lot, so I don't view that quite as the distinction.
Increased plenty leading to increased morality, better societies leading to more focus on effective altruism (in the general sense, rather than this specific movement), more wealth to throw around, work at replacing meat with fake alternatives (which should speed up as technology increases too), advanced technology which makes the previous far quicker to iterate on.
I certainly expect in the short-term to see increasing factory farming, but I don't see the current economies as being a sign that it will continue indefinitely. China, for example, is rich but is also a very quickly growing economy from not-so-rich—they will scale up their meat production quickly to meet their population's needs and desires. This, to me, is like asking why they would build a bunch of okay concrete apartments rather than fancier apartment designs from 201x with design issues despite the nicer status—because they need a lot of living space/food, the existing methods of breeding a massive amount of animals work easily at scale, the technology for it isn't super widespread or optimized, and they aren't massively rich in that they can easily rebuild that entire industry. They prefer eating meat and so the market meets that demand, if they could have been served by just as good alternatives that were cheaper then they would eat those.
I don't see notable reasons to believe that farming a chicken will be the most efficient way of producing meat for the next fifty years given the rate of technology. (And that's on the tail-end, even without AI beyond the current, there's a lot of solutions to extract in fifty years.)
(Current alternatives have various issues. Not tasting as good. Expense. Political polarization. The first two are entirely solvable with enough effort or a generation which grew up with both. The third is a problem but is something that would be faced in any spending on animal welfare as well, and I don't consider it a fundamental political disagreement which I think indicates it has a good chance to dissolve.)
The argument then, is that the more quickly we have people working to increase the speed of economic growth and technological advancement, the quicker the inevitable replacement of factory farming occurs. If the technology comes a year earlier, then that is another year to drive down the prices and scale up so that third-world countries don't go through a buildup of factory farming as they become richer. Of course, this can easily be taken (and I do) as an argument that favors animal welfare investments of a certain kind—that of increasing focus on synthesizing food. But I can certainly see the logic that increasing technological progress & economic growth would encourage (not necessarily charitable) investment and interest.
If a human dies and we have a lot of humans very very similar to them, I think it is plausible that we've lost less. Still a negative, but not as much of one. (Which is one answer that I favor to the repugnant conclusion, you can't just add new people indefinitely). I also think this makes more sense for societies that can freely copy minds.
For animals my logic was much the same, but that there's less variation/uniqueness that is lost because (for example) chicken minds have less axes on which they vary notably.
Here's another argument:
I think the "one long life vs two half-long lives" is a good example, but that it matters how long they live. Better to have a parrot that lives for a year rather than two parrots that live for six months. The parrot has more opportunity to learn and build on what it has learned and gets more value out of living for longer. A chicken wouldn't have as much value because it has stronger limits on what it can learn, be curious about, enjoy, and so on. But a parrot that lives 50 years vs two that live 25? I would lean towards two.
I disagree about how much children miss from adult lives, though it depends on how young we're calling children. Children are certainly very valuable, but I do think they miss out on a lot of adult experiences. The problems they solve are less intricate, the understanding of complex joys is significantly weaker (a child playing with toys vs. reading a 150k word book), and so on. But I don't know where I'd do the tradeoff precisely. I think part of the value of children, beyond being a good in-of-themselves, is that they will grow up to be adults which have richer more vibrant and varied experiences.
However, I don't think that matters much here. I don't believe that the longevity we manage to acquire is merely one long life vs. two half-lives. It is more of a "one ten century long human life vs. (tens of? (hundreds of?)) thousands of various animals living a couple years more". I think the human has a lot of space to continue learning, growing, and experiencing that many animals unfortunately saturate. (Of course their happiness+lack-of-suffering+other positive emotions matters significantly as well)
Then there's also the factor that paired with the fantastical technology that would allow life extension, many other ills of humanity will be pushed back. If a person isn't interchangeable at all (plausible), then ensuring that they survive means they'll experience all these wonders. Rather than letting many animals live for a few more years in happiness (a good thing!), you get X amount of people who are able to go on to live in a world closer to a utopia.
As I said previously, I find the animal welfare being far more neglected and more important than current human welfare to be probably true. However, I think comparing Animal Welfare vs. Global Health ignores that EA has areas of thought which indicate that Global Health isn't considering certain factors, like longevity meaning more people get to live in a better and better world where we may have solved aging. Most charities are operating under a 'everything continues as normal' paradigm, which gives EA an advantage here.
Part of what makes me uncertain and which would make Animal Welfare more of an obvious choice is that Global Health might already be putting a lot into longevity. I suspect they aren't, given general ignorance of cryo, but they're tackling many things that correlate with it, which would still tilt the calculation in the favor of Animal Welfare.
Somewhat neutral, though I concur that animal welfare is more neglected and that a straightforward shortterm calculation is on the animal welfare side. However:
With AI, ensuring longevity for many people may be a better use, though I'm uncertain about the exact costs. Animals are more interchangeable than humans, and will die within not that long regardless, which means that ensuring more humans live longer lives is more valuable. The more people that live longer, the more that are able to participate in a possible longevity escape velocity. (Theoretically one could argue for lengthening the lives of many animals, but most animals are unlikely to live for long enough and will take longer to be given life extending medications)
This makes me view the calculus as significantly more balanced, or potentially weighted on global health significantly. If we can ensure that more people survive, then that ensures a lot of positive utility.
You (and @tylermjohn) might be interested in Diffractor's Unifying Bargaining sequence. The sequence focuses on transferable utility games being a better target than just bargaining games, with I believe Nash being a special-case for bargaining games. As well as talking about avoiding threats in bargaining and trying to further refine.
I think the defaults won't matter too much. Do you have any writing on the moral parliament that talks about the defaults issue more?
I generally agree with this post that it is a good target. This is effectively equivalent to Coherent Extrapolated Volition[1], which has been proposed as a target for alignment. The difference is whether you consider extrapolated values by default or not.
https://www.lesswrong.com/w/coherent-extrapolated-volition-alignment-target
Just to be clear, my intended belief there was a 5-15% of meat industry taking dominance for 2-3 decades after the introduction of an alternative that is very close or beats meat entirely.
Though I do think it is plausible the alternatives come soon, AI technology is advancing rapidly; and I don't believe many people's models are properly factoring in AI technology of our current level being applied to more areas, much less what we'll have in the future after significant advancements.
Of course, EA tends to be a lot better at that than other charities.
As an example: Theorem proving is bottlenecked by the annoying but solvable triplet of: data, money to train larger models, and companies focusing on it. Scaling the current methods would hit noticeable limits due to planning/search being hard, but would allow a lot of automation towards proving software correct. AlphaProof itself is then a step above the methods that came before it. This could provide a good amount of value in terms of ensuring important software is correct but is generally ignored or assumed to need massive breakthroughs.
I find it plausible that more systems in the vein of AlphaFold (protein prediction, most centrally relevant to meat) can be extended to other areas of chemistry with a significant amount of time & effort to collect data and design. There's big data collection problems here, we have a lot of data about food but it is more locked away inside companies and not as carefully researched to a low level as proteins.
I know the theorem proving better than I do the AlphaFold area, but that gets across my general view of "many mental models assume too much like we are in 2018 but with single isolated notable advancements like AlphaFold/AlphaProof/ChatGPT rather than a field with much to explore via permutations of those core ideas".
I believe it is entirely feasible to get the taste right. However, I don't believe that is a major problem. Even in the worlds where it is very expensive to get the texture exactly right, we do what many cultures have done over time and between themselves: we modify it and get used to it. Foods and other less literal tastes being so varying between cultures and even age groups makes me optimistic that even failed replications of taste/texture could replace meat simply through a change in generation where children see it as merely another food option.
Though, admittedly, we are evolved to eat meat. This likely makes us more particular, yet we also prepare other foods like vegetables in exotic manners.
I don't see why you think if it was about efficiency we would already have switched.
I'm somewhat confused: Current met production currently seems efficient based on people's eating habits, desire for meat, expectations about what is healthy (various people don't trust vegetarian answers for good and bad reasons), and most importantly our tech level.
Is your argument that the meat industry is getting enough subsidies that they aren't truly more efficient than current alternatives? And/or that the government isn't requiring to price in the externalities of their effects on the land or climate? If they are actually less cost-effective (in terms of food produced) without the effective subsidies, that would be interesting information to learn, but I'd be somewhat surprised. It would actually make me more optimistic about the state of alternatives to meat, though I also understand that it would be a mark against my theory.
(Just to be clear, I think transitioning will still take time. If we had gotten an instant win of better/cheaper/healthier alternatives back in 2015 without a slow buildup, that would have helped massively and things would have scaled up. I would expect a massive amount more beyond burger and such in stores by now, but I'd still expect meat for a while yet! Unfortunately we're in the world where it has become a somewhat politically polarized topic, and it isn't an obvious win to many consumers, which slows things down even if we had a definite better alternative available.)
I'm skeptical the meat industry survives in the current form. It is possible they drag out their existence for a long while, but since I expect cheaper more plentiful food from the artificial sources, that leads to a great method to out-compete them.
Just to be clear, I'm not arguing against donating to animal welfare, but I do see donations in this area as mostly bringing the time the transition occurs closer in most possibilities. Still very much worthwhile! Cutting a decade into five years gives a lot of value, and even setting the stage for when great alternatives exist is valuable.
Throwing a number out as a weak model, I'd say that about 5-15% of worlds where the meat industry manipulates government to strongly rely on factory farming for a significantly long time. A decade? Two decades? Three? In the other possibilities, I expect factory farming to limp along but be shifted out at varying speeds. They would still try to stifle alternatives, but not be in a dominant position.
I think food companies are already interested in alternative meat products, which means you don't have a full cartel. Even in the 5-15%, I expect the meat companies to inevitably adopt the technology themselves even if they've choked out all the competitors. Not having the competitors is still very much a problem given how long they would delay the change.
(Climate change, for example, is a harder problem because it requires more coordination and most people can't just substitute electricity into their car. Food is a lot more of a substitutable good.)
Hm, the post is good, though I see this as a relatively weak statement. The amount of factory farming would be massively lower than now, presuming that my assumptions about people switching off of meat are true.
I also expect that if we get a big boost of technology (even better AI-driven protein/chemical synthesis or discovery), then I don't expect the argument that we'll still farm them for dyes and such to hold. As the years go by, it becomes ever more feasible to synthesize those useful dyes or materials directly. The outcome described there would still be a lot less (several orders of magnitude?) factory farming. I do think this point depends notably on how soon you think the technology will occur and solve a lot of the general problem (getting chemicals/proteins en masse). I find it plausible that it will come before we solve various meat alternatives (in the better or equivalent price/taste/cost/nutrition sense), but also find it mildly plausible that it takes a decade or two after.
Point three of the article that AI will make factory farming more efficient is true, but also I don't have a reason to believe the final conclusion. Big data analytics does not provide notable evidence to believe that factory farming will outcompete alternative methods in the long run—it is an argument that they aren't constant and so have a longer shelf-life than the naive extrapolation. Growing an animal simply requires a lot of work and energy in a specific form that I don't have any reason to believe alternative meats require as much. As for the bioengineering example, a similar argument implies. To me, this is like saying that a person down in a mine is always going to be more efficient and scalable than a digging machine.
Point four is one that I think fills out a difference in vision. They compare it to chocolate which hasn't been replaced. My model of the world is that over the next thirty or so years we have significant advancement in fields like chemical synthesis, reverse engineering proteins, and so on. I don't see any reason to believe chocolate won't be replaced! Many of the foods made from it alter the taste massively themselves, if food companies could replace chocolate with a significantly cheaper version then they would. I believe they would be slower about it, it presumably doesn't cost much and has been quite tuned.
It feels like there's a missing note in the post, that they think things will stay comparatively the same as 2020 level competitiveness even though there's very little reason to believe that. Why should I think that growing a full animal is remotely efficient for any of our needs?
There are some good points or areas to think about, and the weakest claim that there might be a notable amount of factory farming is true, but I feel overall a bit baffled by the thrust of the article. There's a missing note to the post, a strong background assumption that technology will continue looking as it is circa 2020 with no major optimizations feasible. AI isn't an integral part of my point, though I believe if it succeeds as it has so far then the timescale for various synthesis methods and the like moves up quite a lot, so I don't view that quite as the distinction.
Increased plenty leading to increased morality, better societies leading to more focus on effective altruism (in the general sense, rather than this specific movement), more wealth to throw around, work at replacing meat with fake alternatives (which should speed up as technology increases too), advanced technology which makes the previous far quicker to iterate on.
I certainly expect in the short-term to see increasing factory farming, but I don't see the current economies as being a sign that it will continue indefinitely. China, for example, is rich but is also a very quickly growing economy from not-so-rich—they will scale up their meat production quickly to meet their population's needs and desires. This, to me, is like asking why they would build a bunch of okay concrete apartments rather than fancier apartment designs from 201x with design issues despite the nicer status—because they need a lot of living space/food, the existing methods of breeding a massive amount of animals work easily at scale, the technology for it isn't super widespread or optimized, and they aren't massively rich in that they can easily rebuild that entire industry. They prefer eating meat and so the market meets that demand, if they could have been served by just as good alternatives that were cheaper then they would eat those.
I don't see notable reasons to believe that farming a chicken will be the most efficient way of producing meat for the next fifty years given the rate of technology. (And that's on the tail-end, even without AI beyond the current, there's a lot of solutions to extract in fifty years.)
(Current alternatives have various issues. Not tasting as good. Expense. Political polarization. The first two are entirely solvable with enough effort or a generation which grew up with both. The third is a problem but is something that would be faced in any spending on animal welfare as well, and I don't consider it a fundamental political disagreement which I think indicates it has a good chance to dissolve.)
The argument then, is that the more quickly we have people working to increase the speed of economic growth and technological advancement, the quicker the inevitable replacement of factory farming occurs. If the technology comes a year earlier, then that is another year to drive down the prices and scale up so that third-world countries don't go through a buildup of factory farming as they become richer.
Of course, this can easily be taken (and I do) as an argument that favors animal welfare investments of a certain kind—that of increasing focus on synthesizing food. But I can certainly see the logic that increasing technological progress & economic growth would encourage (not necessarily charitable) investment and interest.
If a human dies and we have a lot of humans very very similar to them, I think it is plausible that we've lost less. Still a negative, but not as much of one. (Which is one answer that I favor to the repugnant conclusion, you can't just add new people indefinitely). I also think this makes more sense for societies that can freely copy minds.
For animals my logic was much the same, but that there's less variation/uniqueness that is lost because (for example) chicken minds have less axes on which they vary notably.
Here's another argument:
I think the "one long life vs two half-long lives" is a good example, but that it matters how long they live. Better to have a parrot that lives for a year rather than two parrots that live for six months. The parrot has more opportunity to learn and build on what it has learned and gets more value out of living for longer. A chicken wouldn't have as much value because it has stronger limits on what it can learn, be curious about, enjoy, and so on. But a parrot that lives 50 years vs two that live 25? I would lean towards two.
I disagree about how much children miss from adult lives, though it depends on how young we're calling children. Children are certainly very valuable, but I do think they miss out on a lot of adult experiences. The problems they solve are less intricate, the understanding of complex joys is significantly weaker (a child playing with toys vs. reading a 150k word book), and so on. But I don't know where I'd do the tradeoff precisely. I think part of the value of children, beyond being a good in-of-themselves, is that they will grow up to be adults which have richer more vibrant and varied experiences.
However, I don't think that matters much here. I don't believe that the longevity we manage to acquire is merely one long life vs. two half-lives. It is more of a "one ten century long human life vs. (tens of? (hundreds of?)) thousands of various animals living a couple years more". I think the human has a lot of space to continue learning, growing, and experiencing that many animals unfortunately saturate. (Of course their happiness+lack-of-suffering+other positive emotions matters significantly as well)
Then there's also the factor that paired with the fantastical technology that would allow life extension, many other ills of humanity will be pushed back. If a person isn't interchangeable at all (plausible), then ensuring that they survive means they'll experience all these wonders. Rather than letting many animals live for a few more years in happiness (a good thing!), you get X amount of people who are able to go on to live in a world closer to a utopia.
As I said previously, I find the animal welfare being far more neglected and more important than current human welfare to be probably true. However, I think comparing Animal Welfare vs. Global Health ignores that EA has areas of thought which indicate that Global Health isn't considering certain factors, like longevity meaning more people get to live in a better and better world where we may have solved aging. Most charities are operating under a 'everything continues as normal' paradigm, which gives EA an advantage here.
Part of what makes me uncertain and which would make Animal Welfare more of an obvious choice is that Global Health might already be putting a lot into longevity. I suspect they aren't, given general ignorance of cryo, but they're tackling many things that correlate with it, which would still tilt the calculation in the favor of Animal Welfare.
Somewhat neutral, though I concur that animal welfare is more neglected and that a straightforward shortterm calculation is on the animal welfare side. However:
With AI, ensuring longevity for many people may be a better use, though I'm uncertain about the exact costs. Animals are more interchangeable than humans, and will die within not that long regardless, which means that ensuring more humans live longer lives is more valuable. The more people that live longer, the more that are able to participate in a possible longevity escape velocity. (Theoretically one could argue for lengthening the lives of many animals, but most animals are unlikely to live for long enough and will take longer to be given life extending medications)
This makes me view the calculus as significantly more balanced, or potentially weighted on global health significantly. If we can ensure that more people survive, then that ensures a lot of positive utility.