I remember reading, although I can't find the source, that cows have better living conditions in factory farms (compared to chickens) because stressed and unhappy cows produce less milk, etc.
With the case of chickens, we've managed to goodhart them harder and separate welfare from productivity in a way that we haven't been able to do with cows.
Ha, interesting! I’m happy to hear that because I originally wrote it with cows as the example. Then I learned that cows had big winter starvation issues in the Middle Ages, so I ended up leaning more towards their lives being net negative even then. But chickens didn’t have as much winter starvation because they need less food, so I rewrote it with chickens as the example. So it’s good to hear that there’s another way it was the better choice!
Excellent piece with a useful metaphor for opening up thinking about (potentially negative) possibility space. I wonder if amphetamines and peptides are already advanced tendrils of future organization trying to nibble away our biological limits/negotiating floor.
The total welfare for chickens point wrt to their domestication seems underargued to me. You've taken the wild animal suffering argument (deaths in the wild are painful) and yet not weighed it against the sum total of natural behaviours they may not get to express in captivity, the poor quality diets and lifestyles many would have, the cramped conditions etc. Anyone who has seen chickens in modern rural, poor countries will know their lives aren't always beds of roses. They can be ridden with parasites, confined to tiny pens, and even bald of many of their feathers.
By contrast, as with humans until just before the advent of modern medicine, things will usually kill you quickly in nature if you're not reasonably fit and healthy. There are exceptions in the wild, but there are equally many more exceptions of animals being ridden with parasites, having lingering disease and experiencing behavioural miseries in captivity of all kinds. The constant sufficient supply of food means miseries and painful health conditions that would have been deadly in nature are now just prolonged.
And to be clear I have read (and in the case of Pearce and Horta, met and discussed with in person) the various EA and EA-adjacent luminaries and forebears who have made these arguments, so I'm not ignorant of them.
But if you want your argument to be as strong as you're making it out to be, gesturing towards wild animals death isn't enough to establish domestication was a great improvement for chickens at first. You need to say what your account is of animal welfare, why you think behavioural factors are better in captivity/don't matter, why you think quality of death is a good proxy for total welfare, among other things
Very interesting, thanks a lot. Yeah I’d be very interested in getting more feedback from people “in the know” here. One thing that comes to mind though is that modern rural countries probably aren’t the exact same conditions as medieval times? (eg they may have enough food to not let them out of pens at all? and disease burden may be higher because of global connectivity?)
Claude says virtually all chickens were let out during the day in 1700, and 24/7 confinement basically didn’t exist as a concept. It also says farmer would usually kill visibly ill birds, so it might actually be the inverse, that diseased ones’ suffering was shorter than in the wild? However it also brings up many other factors https://claude.ai/share/4c5aa915-e89e-4543-a8af-1de538da1dd2
Anyway, as long as the lives don’t turn net negative, technically the analogy still goes through since that means total welfare is going up. but yeah, definitely seems like I should’ve been a bit more thorough with this
Skimmed it, looks nice. In footnote 1, there’s a bunch of other angles on this that I’ve come across, if you’re interested (although I hope in this post I am saying something novel on top of that).
I like this essay though I want to flag that a lot of biological evolution isn't by natural selection. Randomness like founder effects plays a large role in the final product, much more so than say other selection mechanisms people are somewhat familiar with, like artificial selection/breeding and gradient descent.
OK that makes sense to me but I need a concrete example. Would the point be that eg the population of initial AIs might be very small so there could be a huge "inefficiency" that persists because of founder effects?
re Goodharting.
I remember reading, although I can't find the source, that cows have better living conditions in factory farms (compared to chickens) because stressed and unhappy cows produce less milk, etc.
With the case of chickens, we've managed to goodhart them harder and separate welfare from productivity in a way that we haven't been able to do with cows.
Ha, interesting! I’m happy to hear that because I originally wrote it with cows as the example. Then I learned that cows had big winter starvation issues in the Middle Ages, so I ended up leaning more towards their lives being net negative even then. But chickens didn’t have as much winter starvation because they need less food, so I rewrote it with chickens as the example. So it’s good to hear that there’s another way it was the better choice!
Excellent piece with a useful metaphor for opening up thinking about (potentially negative) possibility space. I wonder if amphetamines and peptides are already advanced tendrils of future organization trying to nibble away our biological limits/negotiating floor.
thank you!
The total welfare for chickens point wrt to their domestication seems underargued to me. You've taken the wild animal suffering argument (deaths in the wild are painful) and yet not weighed it against the sum total of natural behaviours they may not get to express in captivity, the poor quality diets and lifestyles many would have, the cramped conditions etc. Anyone who has seen chickens in modern rural, poor countries will know their lives aren't always beds of roses. They can be ridden with parasites, confined to tiny pens, and even bald of many of their feathers.
By contrast, as with humans until just before the advent of modern medicine, things will usually kill you quickly in nature if you're not reasonably fit and healthy. There are exceptions in the wild, but there are equally many more exceptions of animals being ridden with parasites, having lingering disease and experiencing behavioural miseries in captivity of all kinds. The constant sufficient supply of food means miseries and painful health conditions that would have been deadly in nature are now just prolonged.
And to be clear I have read (and in the case of Pearce and Horta, met and discussed with in person) the various EA and EA-adjacent luminaries and forebears who have made these arguments, so I'm not ignorant of them.
But if you want your argument to be as strong as you're making it out to be, gesturing towards wild animals death isn't enough to establish domestication was a great improvement for chickens at first. You need to say what your account is of animal welfare, why you think behavioural factors are better in captivity/don't matter, why you think quality of death is a good proxy for total welfare, among other things
Very interesting, thanks a lot. Yeah I’d be very interested in getting more feedback from people “in the know” here. One thing that comes to mind though is that modern rural countries probably aren’t the exact same conditions as medieval times? (eg they may have enough food to not let them out of pens at all? and disease burden may be higher because of global connectivity?)
Do you have any sense or informed guess on what the median expert take here is? does it lean more towards domestication being good or bad?
Claude says virtually all chickens were let out during the day in 1700, and 24/7 confinement basically didn’t exist as a concept. It also says farmer would usually kill visibly ill birds, so it might actually be the inverse, that diseased ones’ suffering was shorter than in the wild? However it also brings up many other factors https://claude.ai/share/4c5aa915-e89e-4543-a8af-1de538da1dd2
Anyway, as long as the lives don’t turn net negative, technically the analogy still goes through since that means total welfare is going up. but yeah, definitely seems like I should’ve been a bit more thorough with this
Nice article!
Curious if you've read https://grantslatton.com/technocapital which is a different angle on some related dynamics
Thank you!
Skimmed it, looks nice. In footnote 1, there’s a bunch of other angles on this that I’ve come across, if you’re interested (although I hope in this post I am saying something novel on top of that).
I learned a lot about chickens. And humans.
I like this essay though I want to flag that a lot of biological evolution isn't by natural selection. Randomness like founder effects plays a large role in the final product, much more so than say other selection mechanisms people are somewhat familiar with, like artificial selection/breeding and gradient descent.
Thanks Linch!
This post is the one that helped the most with intuition for me: https://apomorphic.com/2019/12/12/evolution-is-sampling-error
looks really interesting, will have a read
OK that makes sense to me but I need a concrete example. Would the point be that eg the population of initial AIs might be very small so there could be a huge "inefficiency" that persists because of founder effects?