PLF doesn't seem like an AI technology. It's a sensor technology. The innovation is cheap and good sensors that can be integrated into agricultural workflows. You don't need to apply anything more than linear regressions or crude rules of thumb to the data from the sensors.
Of course, anything with a bit of math inside gets called "AI" nowadays since we're in that part of the hype cycle, so I understand why PLF is sold as "AI for farms".
Also, if the main effect of PLF is to make meat cheaper at the potential cost of animal welfare, then it's no different from factory farming itself. That's the tradeoff. Happier people who can enjoy more meat at lower prices compared to other farming methods, at the price of unhappier animals. So whatever moral arguments apply against or for factory farming apply automatically to PLF -- it's the same discussion.
I'm confused about some of the basic assumptions here; it seems you are saying that increasing the number of animals is bad for those animals. Would "overall animal welfare" be better off if all cows were immediately, painlessly killed, and no cows ever existed again?
Thanks for the engagement, this is an important question. I didn't dive too much into the underlying ethics here. Maybe I shouldn't have set this discussion aside.
The best case scenario, in my view, is that there are a lot of happy animals living lives worth living -- it's certainly possible, in principle, that we could reduce the factory-farmed animal population and increase the number of animals living happy lives.
Of course, in practice, as you rightly point out, we often have to choose between a factory-farmed animal existing and no animal at all. What to do then? Compared to the catastrophe that is factory farming, I indeed think it is very likely better for most factory farmed animals not to have been born at all.
Why? A great deal of factory-farmed animal lives are spent in at least moderate suffering. See https://welfarefootprint.org/compare/ for a really informative breakdown of how much suffering animals experience from moment to moment over the course of their lives, broken down by species, farming use, etc.. I think this research supports my view.
However looking at pain breakdowns like this doesn't get across just how bad the instances of expecially bad suffering can be. See e.g. here for a littany of some of the worst, most excruciating instances of suffering exacted on factory farmed animals: https://benthams.substack.com/p/factory-farming-delenda-est. In my view, some of these sufferings are so incomprehensably bad that even enduring them for a few minutes would make a life not worth living (especially in the context of a very short lifespan). Consider male chickens in egg factories, killed by maceration or suffocation within days of being born. It's hard for me to see how it is better for these animals to be lives these lives than not to.
I do feel less certain about this claim for cows than for chickens and smaller animals. However chickens and smaller animals are the vast majority of farmed animals, so when evaluating the idea that factory farmed animal lives are not worth living, this is the population that will dominate the calculus.
I see, perhaps I just fundamentally disagree then. This sort of "suffering is the worst possible thing" reasoning doesn't seem viable to me.
What about bears, for example. A single bear sure kills a lot of wild animals during its lifetime. It must inflict far more suffering than it enjoys life. It's very cruel and violent when you see how these predators hunt. Or eagles. Would you support the extinction of all bears and eagles, so that the wild deer and mice could live happier lives?
In the short term, I don't think we could successfully control nature with sufficient wisdom and skill that this would be a suffering-reducing action. And make a species extinct is especially risky, since it is mostly irreversible. We shouldn't be reckless.
And it's a relatively common idea in religious ethics. For example, Isaiah 11 has often been interpreted to suggest that, in a perfect state of affairs, animals would be non predatory.
• Most people who eat animal-sourced foods would agree that improving the welfare of livestock is a worthy goal. Many people would pay a higher price for eggs to assure more humane conditions for factory farming of poultry.
• Machine learning is a promising technology for improving the welfare of livestock.
• Whether machine learning is used to improve the welfare of livestock depends on how incentives are aligned. Laws/regulations, reliable labeling & marketing of humane conditions would all promote this point of view.
• It is not immediately obvious that your take on AI & animal welfare is relevant or useful, as you a priori object to the use of animal-sourced foods on an ethical basis.
• Humans evolved, over a period of millions of years, to eat fatty meat. Optimal health requires animal-sourced foods, as no available regimen of supplements presents a healthful substitute. Your failure to accept this is more of a religious or quasi-religious stance.
Thanks for engaging Steven. Responding to a few of your points here:
"Whether machine learning is used to improve the welfare of livestock depends on how incentives are aligned. Laws/regulations, reliable labeling & marketing of humane conditions would all promote this point of view."
Yes, exactly! Whether ML improves welfare depends on incentives. What I argue in this piece is that the incentives are such that average animal welfare may go up or down and total animal welfare will go down. That's also why I argue that we should use activism and regulation to shape these incentives for the better.
"you a priori object to the use of animal-sourced foods on an ethical basis"
I do object on an ethical basis, although not on an a priori one. My views about the badness of animal sourced food have to do with contingent facts about the welfare of the animals involved.
"Humans evolved, over a period of millions of years, to eat fatty meat. Optimal health requires animal-sourced foods, as no available regimen of supplements presents a healthful substitute."
Arguments about the healthfulness of vegan or vegetarian diets aside, everything in this post is compatible with reducitarianism or flexitarianism. So your point strikes me as a bit of a red herring.
Imagine a new breed of chicken with a level of consciousness reduced to that of a scallop, unable to experience pain. Such a designer chicken would not be viable as an independent entity but would be viable if controlled by AI.
There isn't any strong research result that supports the healthfulness of a 100% plant-based diets (including state-of-the-art supplements). Neither is there any strong research result that supports the unhealthfulness of eating animal-based foods. The diet that most Americans eat is unhealthful in large part because of inadequate intake of animal-based foods and excess intake of industrial plant-based foods, such as corn syrup, seed oils, and grains.
There are healthy indigenous diets that are reducitarian, such as the Kitavan diet of fish, tubers, coconut, and fresh fruit. There are also healthy indigenous diets that are hypercarnivore, such as the Inuit diet of meat (seal, caribou, muskox), birds & their eggs, and fish (sculpin, cod, lake trout). There is no indigenous diet that is 100% plant-based foods.
Neither is there any modern vegan diet that has been shown to be healthy.
PLF doesn't seem like an AI technology. It's a sensor technology. The innovation is cheap and good sensors that can be integrated into agricultural workflows. You don't need to apply anything more than linear regressions or crude rules of thumb to the data from the sensors.
Of course, anything with a bit of math inside gets called "AI" nowadays since we're in that part of the hype cycle, so I understand why PLF is sold as "AI for farms".
Also, if the main effect of PLF is to make meat cheaper at the potential cost of animal welfare, then it's no different from factory farming itself. That's the tradeoff. Happier people who can enjoy more meat at lower prices compared to other farming methods, at the price of unhappier animals. So whatever moral arguments apply against or for factory farming apply automatically to PLF -- it's the same discussion.
I'm confused about some of the basic assumptions here; it seems you are saying that increasing the number of animals is bad for those animals. Would "overall animal welfare" be better off if all cows were immediately, painlessly killed, and no cows ever existed again?
Thanks for the engagement, this is an important question. I didn't dive too much into the underlying ethics here. Maybe I shouldn't have set this discussion aside.
The best case scenario, in my view, is that there are a lot of happy animals living lives worth living -- it's certainly possible, in principle, that we could reduce the factory-farmed animal population and increase the number of animals living happy lives.
Of course, in practice, as you rightly point out, we often have to choose between a factory-farmed animal existing and no animal at all. What to do then? Compared to the catastrophe that is factory farming, I indeed think it is very likely better for most factory farmed animals not to have been born at all.
Why? A great deal of factory-farmed animal lives are spent in at least moderate suffering. See https://welfarefootprint.org/compare/ for a really informative breakdown of how much suffering animals experience from moment to moment over the course of their lives, broken down by species, farming use, etc.. I think this research supports my view.
However looking at pain breakdowns like this doesn't get across just how bad the instances of expecially bad suffering can be. See e.g. here for a littany of some of the worst, most excruciating instances of suffering exacted on factory farmed animals: https://benthams.substack.com/p/factory-farming-delenda-est. In my view, some of these sufferings are so incomprehensably bad that even enduring them for a few minutes would make a life not worth living (especially in the context of a very short lifespan). Consider male chickens in egg factories, killed by maceration or suffocation within days of being born. It's hard for me to see how it is better for these animals to be lives these lives than not to.
I do feel less certain about this claim for cows than for chickens and smaller animals. However chickens and smaller animals are the vast majority of farmed animals, so when evaluating the idea that factory farmed animal lives are not worth living, this is the population that will dominate the calculus.
I see, perhaps I just fundamentally disagree then. This sort of "suffering is the worst possible thing" reasoning doesn't seem viable to me.
What about bears, for example. A single bear sure kills a lot of wild animals during its lifetime. It must inflict far more suffering than it enjoys life. It's very cruel and violent when you see how these predators hunt. Or eagles. Would you support the extinction of all bears and eagles, so that the wild deer and mice could live happier lives?
In the short term, I don't think we could successfully control nature with sufficient wisdom and skill that this would be a suffering-reducing action. And make a species extinct is especially risky, since it is mostly irreversible. We shouldn't be reckless.
Still, I think we should work towards ways to reduce suffering in the wild, including from predation. This is a less freaky and unusual view than is ordinarily thought. Here are two mainstream philosophers who have endorsed this idea (https://www.newyorker.com/culture/the-weekend-essay/what-would-it-mean-to-treat-animals-fairly)
And it's a relatively common idea in religious ethics. For example, Isaiah 11 has often been interpreted to suggest that, in a perfect state of affairs, animals would be non predatory.
• Most people who eat animal-sourced foods would agree that improving the welfare of livestock is a worthy goal. Many people would pay a higher price for eggs to assure more humane conditions for factory farming of poultry.
• Machine learning is a promising technology for improving the welfare of livestock.
• Whether machine learning is used to improve the welfare of livestock depends on how incentives are aligned. Laws/regulations, reliable labeling & marketing of humane conditions would all promote this point of view.
• It is not immediately obvious that your take on AI & animal welfare is relevant or useful, as you a priori object to the use of animal-sourced foods on an ethical basis.
• Humans evolved, over a period of millions of years, to eat fatty meat. Optimal health requires animal-sourced foods, as no available regimen of supplements presents a healthful substitute. Your failure to accept this is more of a religious or quasi-religious stance.
Or so it seems to me.
Thanks for engaging Steven. Responding to a few of your points here:
"Whether machine learning is used to improve the welfare of livestock depends on how incentives are aligned. Laws/regulations, reliable labeling & marketing of humane conditions would all promote this point of view."
Yes, exactly! Whether ML improves welfare depends on incentives. What I argue in this piece is that the incentives are such that average animal welfare may go up or down and total animal welfare will go down. That's also why I argue that we should use activism and regulation to shape these incentives for the better.
"you a priori object to the use of animal-sourced foods on an ethical basis"
I do object on an ethical basis, although not on an a priori one. My views about the badness of animal sourced food have to do with contingent facts about the welfare of the animals involved.
"Humans evolved, over a period of millions of years, to eat fatty meat. Optimal health requires animal-sourced foods, as no available regimen of supplements presents a healthful substitute."
Arguments about the healthfulness of vegan or vegetarian diets aside, everything in this post is compatible with reducitarianism or flexitarianism. So your point strikes me as a bit of a red herring.
Imagine a new breed of chicken with a level of consciousness reduced to that of a scallop, unable to experience pain. Such a designer chicken would not be viable as an independent entity but would be viable if controlled by AI.
There isn't any strong research result that supports the healthfulness of a 100% plant-based diets (including state-of-the-art supplements). Neither is there any strong research result that supports the unhealthfulness of eating animal-based foods. The diet that most Americans eat is unhealthful in large part because of inadequate intake of animal-based foods and excess intake of industrial plant-based foods, such as corn syrup, seed oils, and grains.
There are healthy indigenous diets that are reducitarian, such as the Kitavan diet of fish, tubers, coconut, and fresh fruit. There are also healthy indigenous diets that are hypercarnivore, such as the Inuit diet of meat (seal, caribou, muskox), birds & their eggs, and fish (sculpin, cod, lake trout). There is no indigenous diet that is 100% plant-based foods.
Neither is there any modern vegan diet that has been shown to be healthy.