Why the expected numbers of farmed animals in the far future might be huge
Summary
This post is essentially a case that factory farming could be a longtermist cause area. (but not a case that the current portfolio of work on factory farming are good or relevant for the future)
Alternatively, you can see this post as a response to claims similar to this: “The scale of the problem of factory farming itself (ignoring potential long-run benefits of animal advocacy) is small compared to the scale of issues affecting future generations.”
Unlike previous attempts like this or this, this post does not base the case on arguments about the flow-through/knock-on effects of farmed animal advocacy (or other animal advocacy).
I used the expected number of humans by (Future of Humanity Institute 2021), as a basis of my guesstimates. This formed my view that the case for concerning about farmed animals in the far future is at least as strong as the case for concerning about humans in the far future.
The strongest argument against the importance of this post is that, in expectation, artificial sentience will far outnumber the number of farmed animals. The caveat is that this argument would also suggest that humans are not of significant concern in the longtermist worldview.
The only intervention I am proposing in light of this proposed longtermist cause area, is to do research to update the guesstimates I provided in this post.
P.S. (2022 Jan 14). Thanks to Rohin Shah pointing out that the numbers in this post depend heavily on the probability (or your credence) that the universe will be filled mainly with digital people (In his words. I prefer to say “digital beings” when I discuss this topic). This is because digital people will, presumably, have very few incentive to raise animals for food, or even other purposes.
P.P.S. Thinking about the last point. I can actually think of one reason why digital beings might raise some animals. If bringing about animals somehow is the best way to extract resources from some planets (like, eat and then decompose stuff), they digital people might still do it.
Why this post?
In 10+ (I believe 12-14, I regret not tracking them seriously from the start) discussions (mine vs others, and also discussions I see on EA forum, podcasts, forwarded emails, etc.) about longtermism, I often heard from other longtermists claim that factory farming is a “shortermist”/non-lontermist cause area. For example, 80000 hours said in their factory farming cause area profile: “The scale of the problem of factory farming itself (ignoring potential long-run benefits of animal advocacy) is small compared to the scale of issues affecting future generations.” The most often heard (in fact, I don’t remember a single case when it wasn’t) justification for this view is that, they believe, factory farming is grossly energy inefficient in comparison with its potential substitutes, especially cultivated meat, and therefore will likely (some say certainly) be replaced. I (and a few people in animal welfare and alternative protein spaces I talked to) am personally baffled by the level of certainty they tend to have on the claim, and also the lack of support further than the energy efficiency argument. Since then, I have found more and more evidence that this view is probably misinformed. And since the truth of this claim might have very significant implications for the longtermist picture (especially for people who focus more on S-risk then X-risk). I decided to write a post on it.
Applying the expected value theory, also to animal issues
Before going into the technical details of the topic. I want to discuss a bit about applying the expected value theory consistently. In quite a lot of discussions on the relation between animal welfare and longtermism, people made arguments like “I think a key objection for me is to the idea that wild animals will be included in space settlement in any significant numbers.“ Or “I’m pretty optimistic this won’t happen however. I think by default we should expect that the future (if we don’t die out), will be predominantly composed of humans and our (digital) descendants, rather than things that look like wild animals today.” I think this type of responses suffer from the same problem, that they are not strong enough to refute the case to regard some animal welfare issues as longtermist cause areas, because even if they are right to claim that future scenarios with a lot of animals are “unlikely”, the stakes might still be very high due to expected value calculations. As a comparison, 80000 hours made a similar argument for the importance of the long-term future even if one thinks humanity is “unlikely” to survive for long: “if you think there’s a 5% chance that civilisation survives 10 million generations till the end of the Earth, then (in expectation) there will be over 500,000 future generations.”
There is also another type of response which doesn’t suffer from the same problem because they feel “certain” about claims that reject animals being a concern for longtermism. This example is a very typical one: “By that stage of technological development we will surely be eating meat produced without a whole animal, if we eat meat at all.” In my opinion, these claims of asserted certainty are not justified, and if they are wrong, the stakes can be extremely high, for similar reasons as the above. This post will explain why the claim that we will not have factory farming in the long-term future is not justified.
Quoting 80000 hours, “if you’re uncertain whether the future will be big, then a top priority should be to figure out whether it will be — it would be the most important moral discovery you could make.” This is what I am trying to do with this post, in the context of factory farming.
The argument of overwhelming expected numbers of artificial sentience
The strongest argument against the importance of this post is that, in expectation, artificial sentience will far outnumber the number of farmed animals I personally find this argument compelling, but feel extremely unsure how we should deal with uncertainties like, i. whether there will be artificial sentience, or ii. whether they will have the capacity to suffer, etc. Also, a big caveat I have is that this argument would also suggest that humans are not of significant concern in the longtermist worldview. I believe this is a position a lot of longtermists might not want to take. I personally think it would be unproductive to advocate for less concern for humans in the long-term. But to reduce my cognitive dissonance, my choice is to have more concern for animals in the long-term future.
How much is at stake (in expectation)
To come up with guesstimates about the far future, we probably need to start with current numbers. To stay as conservative as I can (and therefore as unlikely to refute what I want to refute as I can). For all the source data that has a range (lower to higher estimates) in the number of animals, I automatically take the lowest.
Step 1: In recent years (2016 to now, different data sets used different reference years), there were 80B terrestrial vertebrate (my own calculations using a series of data from FAOSTAT), 92B farmed finfish (Open Philanthropy, 2019), 260B (260B − 530B) farmed shrimps (Rethink Priorities, 2020 unpublished), and 1000B (1000B − 1200B) farmed insects (Rethink Priorities, 2020) slaughtered each year. If we divide them into vertebrates and invertebrates, we get 172B and 1260B respectively. (Note: I ignored numbers of animals that are not farmed for meat, but their products, such as milk, eggs, fur, and leather, because their statistics are more messy, and adding them won’t change the total numbers by significant portions.)
Step 2: Then I propose to turn this into a ratio of farmed animals per human life. Since the data for the farmed animals were roughly taken at 2017, on average, I will take the 2017 human population figure, which was roughly 7.6B (UN 2017). Therefore, we can conclude that in 2017, humans slaughtered 22.6 vertebrates and 165.8 invertebrates per (human) capita.
Step 3: Finally, I propose to turn these numbers into the number of farmed animals slaughtered per human in our whole lives. According to the World Bank, the average life expectancy from birth in the world was 72.391 in 2017. This means that each current generation human caused 1636 (1600 hereafter) vertebrates and 12002 (12000 hereafter) invertebrates to be slaughtered in our lifetime. I called these numbers per human capita of farmed animals (PHCFA). We can then use these numbers to extrapolate how many farmed animals there will be in the future.
I propose to use the expected number of humans by (Future of Humanity Institute 2021),as a basis of my guesstimates. Extrapolating from their estimates for number of humans in the future, I came up with my guesstimates of number of farmed animals in the future, and this formed my view that the case for concerning about farmed animals in the far future is at least as strong as the case for concerning about humans in the far future. It is important to clarify that I didn’t form the view that farmed animals are in expectation the largest group of moral patients in the future, just that their expected numbers are high enough to be of concern under a longtermist perspective
(Future of Humanity Institute 2021) estimates that, in expected value terms, there will be 1.1 * 1014 expected human lives in the future. But the report also estimated, also in expected value terms, considering the chances of successful space settlement and other factors, the number of humans in the future outside of earth. The table below summarizes the estimates:
| Expected human lives in the future | |
| Earth only | 1.1 * 1014 |
| Solar system only | 5 * 1025 |
| Milky way only | 2.5 * 1034 |
| Affectable universe | 5 * 1045 |
Table 1: Estimates of expected lifespan and population size of humanity
We can now therefore estimate the number of farmed animals in the future if PHCFA stays at the 2017 level. For example, if you believe that we will never leave the earth, the conservative estimate would be 1.1 * 1014 * 1600 = 1.76 * 1017 for vertebrates, or 1.32 * 1018 for invertebrates. But clearly, we cannot simply assume that factory farming will stay at the 2017 level. I will therefore try to come up with the guesstimates of PHCFA (future), from PHCFA (2017).
Below was my mental process for coming up with the calculation, you can read step 1 and skip the rest and go directly to the table. If you are interested, or if the table didn’t make sense to you, my mental process was roughly like this:
Step 1: PHCFA (future) / PHCFA (2017) = coefficient (R)
Step 2: To simplify my calculation and presentation, I assumed that there is no probability distribution for different values of R, and that I have only one prediction for each scenario (e.g. invertebrate farming on earth only). I choose the value of R that is the most likely to me, and only assign a probability to that outcome. For example, invertebrate farming on earth only. I just assumed that there is only one outcome possible: R = 1 (the outcome that is the most likely in this scenario to me), and the probability of this happening is 30%.
Step 3: R (most likely) * Probability for R (most likely) = Expected value of R
Step 4: Expected number of farmed animals in the future = Expected value of R * expected human lives in the future
With that, my views are summarized in the following table:
| Animal | R (most likely) | Probability for R (most likely) | Expected value of R | ||
| Earth only | Vertebrates | 50% | 30% | 6% | |
| Invertebrates | 100% | 50% | 50% | ||
| Solar system only | Vertebrates | 10% | 10% | 1% | |
| Invertebrates | 50% | 20% | 10% | ||
| Milky way only | Vertebrates | 10% | 5% | 0.5% | |
| Invertebrates | 50% | 5% | 2.5% | ||
| Affectable universe | Vertebrates | 10% | 0.5% | 0.05% | |
| Invertebrates | 50% | 1% | 0.5% | ||
Table 2: My beliefs on R in different scenarios, and the resulting products (expected values of R)
| Farmed animals slaughtered per human 2017 | Expected ratio of factory farming compared to 2017 | Expected human lives in the future | Expected number of animals farmed and slaughtered in the future | ||
| Earth only | 1,600 (vertebrates) | 6% | 1.1 * 1014 | 1.0 * 1016 | |
12,000 (invertebrates) | 50% | 6.6 * 1017 | |||
| Solar system only | 1,600 (vertebrates) | 1% | 5 * 1025 | 8 * 1027 | |
12,000 (invertebrates) | 10% | 6 * 1028 | |||
| Milky way only | 1,600 (vertebrates) | 0.5% | 2.5 * 1034 | 2 * 1035 | |
12,000 (invertebrates) | 2.5% | 7.5 * 1036 | |||
| Affectable universe | 1,600 (vertebrates) | 0.05% | 5 * 1045 | 4 * 1045 | |
12,000 (invertebrates) | 0.5% | 3 * 1047 | |||
Table 3: Expected number of animals farmed and slaughtered in the future
Why we shouldn’t expect factory farming to be 100% eliminated
Now we get to the claim that needs the most justification. These considerations, together, formed my views summarized in the tables above.
I. Not all types of factory farming are equal
When the term factory farming is mentioned, the picture people have in mind or get when they google it are usually chickens, pigs, and cows. This is quite expectable as these are the animals that the Western farmed animal movement had focused on in the one to two decades or so. But this kind of picture might provide a few wrong impressions. First, these animals actually do not represent the majority of farmed animals even if we are counting just vertebrates, as the total number of fish is more than double than of the total number of these three animals combined (OPP 2019 compared with Statista 2022). The total number of animals farmed would be more overwhelmingly larger than these three animals if invertebrates are also included. Second, the biggest problems from the raising of these three animals might be lower for the raising of some fishes and invertebrate species. For example, the raising of cows is known to be one of the highest (with lambs and goats) in carbon and water footprint, and while the raising of chickens and pigs are less bad in this regard than the raising of cows, they are still higher than the raising of fishes and invertebrates. The raising of these three animals also produce more manure per weight of animal farmed than the majority of the raising of fish and invertebrates. Also, the raising of cows, chickens, and pigs all are believed to have epidemic or even pandemic potentials. One of the hypotheses of the Spanish Flu was that it came from pigs (Nelson 2018), avian influenza from chickens caused outbreaks among humans multiple times (Yao 2019), the raising of cows might present less risk in thai category but they are also known to have spread zoonotic diseases such as cryptosporidiosis, E. Coli, ringworm, salmonella, and tuberculosis. But no known zoonotic diseases with similarly devastating effects were identified in the farming of fish and invertebrates. Probably most importantly, the feed conversion ratio of some fish and invertebrates can beat the 1.7-2.0 for chickens, or even fall below 1.
The important takeaway from this section is that in evaluating the economic and environmental reasons to replace animal products, we shouldn’t think that animal products = cows + chickens + pigs, and certainly not just cows, which seems to be some people’s impression.
It is important to point out that plants and plant-based foods are generally even more efficient than the raising of fish and invertebrates as food. Soy for instance is superior in many ways. But that is not the case for all plant crops. For instance, the farming of coffee and chocolate are both more carbon intensive than the meat of pigs, chickens, and fish. My intuition is that a lot of humans might want to live in an universe where at least some supply of coffee and chocolate exists despite their inefficiency in producing per weight of food.
II. Factory farming isn’t only about producing human food
Factory farming is often described as the raising of animals for human food such as meat, eggs, and dairy. While this is a big part that factory farming does now, it’s not the complete picture, and we shouldn’t assume that this part will continue to dominate the whole picture of factory farming. Humans also raise animals for making food for animals, fashion, pigments, medical supplies, experimentation, animals as pets and ornaments. These productions also experience pressure to reduce costs and increase the number of animals kept per space, they are often very much like the factory farming for meat, eggs, and dairy (sometimes exactly the same animals).
I am particularly concerned with one non-food use of farmed animals: harvesting organs. Some biotech companies are genetically engineering pigs to make their organs more compatible with the human body. But some scientists are trying to fully grow human organs in animals. This research, if successful, will prima facie increase human welfare. But it could have huge implications for animals as it will increase the likelihood of humans keeping factory farming around, and the number of farmed animals we will raise. The analysis in this paragraph alone increases my assigned probability that there will be factory farming of vertebrates on earth by 2x, and and 5x for scenarios where we expand beyond the earth. Also, this technology could be a reason that my guesstimates are all grossly underestimated.
The implications of this section for the long term is that even if the surge of superior ways to produce protein eliminate the need to use factory farming to produce protein for human consumption, there are still other reasons why we might have factory farming to produce animals for other uses.
The following three paragraphs are pretty optional, but if you are interested in gaining a rough sense of the possible order of magnitudes of the numbers for these sources of factory farming, it might be useful to read some currently existing examples. Take ornamental fish. In 2018, China produced (which means they didn’t count the ones that died before reaching markets) 5.35B ornamental fish (there are no figures for the number of fish existing at a certain point), which represents roughly 4 fish per human in China in 2018. This number might still rise in China with increasing wealth. But in the long-term perspective, the ratio could be much lower as keeping these fish seems more difficult during times of catastrophes, space traveling, and on terraformed planets and space structures, than in contemporary China. But it seems safe to think that humans have a significant chance of (i.e. >20%) raising at least 1 ornamental fish per year in the long run. If that is the case, then the number of animals that have to be farmed means that factory farming would be pretty much the standard set up to go. In expected value terms, the number of ornamental fish raised per human each year can be seen as 0.2.
Consider the use of silk worms to produce silk too. China produced 86,500 tonnes of silk in 2018. According to three sources, different breeds of silk worms, in different seasons, produce 0.22g to 5g of silk per worm. If we take it to be 5g, there are then 17.3B silkworms killed (to make silk each silk worm has to be killed, by boiling) in 2018 in China. But we cannot assume every countries demand silk as much as China. According to multiple sources, China produces roughly 70% of the worlds’ silk, making the estimate of the number of silk worms 24.7B. That means humanity killed roughly 3 silk worms per capita in 2018 according to the absolute minimum estimate. Considering that insect farming takes little space, water, feed, and energy, and is hypothesized to be doable in other planets, and also that it tends to be done not because it is efficient (in fact, it is one of the most inefficient fibers to produce) but because of its perceived cultural and artistic value, it is reasonable to expect that humans will still kill at least 0.1 silkworms per capita each year, in expected value terms, in the long run. But the estimate could be way off, for example if the ratio of people who practice Chinese culture shrinks dramatically, or that the research using silkworms to produce strong functional silk like spider silk becomes commercially successful.
III. Technologies, including AI, will make factory farming more efficient
It might be natural to think of factory farming to have a constant state of technology while technologies on PB/CM will improve very quickly—this was my picture previously.
Many people know that a lot of universities and PB/CM companies are receiving a lot of funding to develop their technologies, products, and market. But the same goes for the factory farming industry. Not only are meat companies very rich and spend a lot of money on R&D. Academia also contributes to technologies used in factory farming. In China, U.S., Netherlands, U.K., India, Brazil, there are universities that have departments or faculties that serve the factory farming industry. There are even universities in China and Japan that are named after “animal husbandry”.
There is currently a new technological trend in the factory farming industry, called “precision livestock farming”. I am now working with Peter Singer researching this development, and its moral implications. Short description of this trend: It is the use of sensors connected to the same computer or server (IoT), big data analytics, more recently the use of AI related technologies such as robotics and machine learning. Using machine learning, patterns such as body temperature, movements, sounds the animals make, or even their faces, can all be interpreted by AI systems to predict and detect diseases, injuries, parasites, model the growth rates, detect pregnancies, etc. These techniques aim at reducing the amount of feed and water used, reduce fatality rate, and increase stocking density.
Another technology that is helping the factory farming industry is bioengineering. Selective breeding techniques can select traits that make animal product production more efficient. For example, the growth rate of broiler chickens had doubled since the 1950s. Currently, there seems to be efforts to try to incorporate AI in this process (1, 2, 3), which industry people claim will make selective breeding much more powerful and quicker..
The analysis in this section casts major doubts that one of PB, CM, 3d printed meat (or their egg and dairy versions) will be superior to all animal products in every possible way, including but not limited to pure energy conversion terms.
IV. Sometimes inefficient things won’t simply disappear
This is probably a weak reason, but might still be worth considering. There seems to be a lot of evidence for this in our current generation. Silk, as we mentioned earlier, is one of the most energy and time inefficient ways to produce fibers. But it kept growing when people got wealthier. We also mentioned earlier that chocolate and coffee have even higher carbon intensity than the meat of fish, chicken and cow. But will humans eliminate the farming of coffee and chocolate entirely? Maybe, but I wouldn’t assign a 100% credence in this belief. And it seems to me that we shouldn’t do that for factory farming either. Some might counter-argue that chocolate and coffee are not just food items but they create enjoyment, but so do animal products for some people.
V. Factory farming might spread to space—Some proponents of space colonization include factory farming in their plans
This is probably the most important point from I to V, particularly regarding factory farming spreading to space.
Gerard O’Neill, in his 1974 paper The Colonization of Space, said “the colonies should be able to support 143 people per hectare with a diet of 3000 calories, 52 grams of usable protein and 4.3 pounds of total food per person per day [ref 9]. Much of the protein would come from poultry and pork.” He even consulted soil scientist Richard Bradfield, who told him that he “has grown enough to feed 72 people per hectare by the techniques of double planting and multiple cropping, and with the use of cuttings for livestock feed.” in an experiment done in the Philippines. Another early major space advocate, planetary scientist Thomas Heppenheimer, wrote in his book Colonies in Space (1977) about the methods of raising fish, rabbits, sheep, and egg laying hens on O’Neill’s cylinder type of space colonies.
It could be argued that cultivated meat was only a concept and plant-based meat was still very rudimentary and distasteful at their time, so they could be missing better options for feeding people. But let’s look at the more recent discussions.
Deep Space Food Challenge, a competition held by NASA and Canadian Space Agency, awarded 2 projects on insect farming in the Canadian branch, and 2 projects on insect farming plus 1 project on fish farming in the U.S. branch. The European Space Agency has a project called Lunar Hatch, looking to bring fish eggs to the moon and conduct aquaculture there. In academia, geologist Kevin M. Cannon, a proponent of space resource extraction, has a website eatlikeamartian.org and a paper Feeding One Million People on Mars, in both of which he proposed to use insect farming (along with cellular agriculture) to feed people’s need for protein. On the commercial side, a joint venture of SpaceX and CERN was announced in November 2018 it was said that “the first fish farms in Mars will be a fact in 2029.” I also find it amusing that a start-up in aquaponics (systems that combine plant agriculture on water, and fish farming) whose current target use scenarios are on earth, is called Terra Mars.
What should we do?
Honestly, I haven’t formed any views on interventions that can tractably affect the lives of farmed animals in the far future. More research on PB/CM, and maybe shifting PB/CM’s focus to replacing aquatic animal farming and competing with insect farming could be the way to go, but I feel too uncertain to make it a recommendation. The only kind of work I feel reasonably confident to recommend EAs to do is to do more research on this potentially important issue, probably refine or even re-do my whole analyses done in this post. I would also like to see an analysis of how this consideration might change current prioritization.
Bonuses
You might skip this part if you don’t want to read something that I didn’t use to form my views summarized in the tables. But these are thoughts that I had which I found interesting.
Bonus 1: Humans might find new animals on other planets, and they might factory farm them
The difficulty of transporting, bioengineering, adapting, and raising earth originated animals in space are reasons to be doubtful about its feasibility. But if we find new animals on other planets, the story might be totally different. This might be a minor consideration currently since the chance seems very low with current evidence. But I think this is worth listing out at least, so that future conversations on this might continue if we will update on this.
Bonus 2: The separation between farmed animals and wild animals might blur over time
The current distinction between these two types of animals are very clear. But as technologies advance, humans might directly manage what happens in “nature”, possibly down to molecular levels. This might be more likely if the “nature” was simply created by humans, on terraformed planets and space structures.
If wild animals live net negative lives, creating “new nature” in space might not be very different from factory farming. We bring them to life to fulfill our preferences, we manage them, we cause them enormous suffering. Whether we eat the corpses of the animals who suffered a lifetime of misery doesn’t really matter to me.
Update: Someone flagged that one of the links was actually an April’s fool—the SpaceX fish farm on mars. But the other projects on space factory farming are real, and the poster of the April’s fool seems to think that the joke is actually feasible.
- Longtermism and animals: Resources + join our Discord community! by (31 Jan 2023 10:45 UTC; 102 points)
- Could it be a (bad) lock-in to replace factory farming with alternative protein? by (10 Sep 2022 16:24 UTC; 85 points)
- Longtermism and Animal Farming Trajectories by (27 Dec 2022 0:58 UTC; 51 points)
- 's comment on The Future Fund’s Project Ideas Competition by (2 Mar 2022 12:08 UTC; 46 points)
- Why I don’t prioritize consciousness research by (28 Oct 2023 8:07 UTC; 45 points)
- How much current animal suffering does longtermism let us ignore? by (21 Apr 2022 9:10 UTC; 40 points)
- EA Updates for April 2022 by (31 Mar 2022 16:43 UTC; 32 points)
- 's comment on Who is protecting animals in the long-term future? by (26 Mar 2022 7:59 UTC; 25 points)
- 's comment on Who is protecting animals in the long-term future? by (22 Mar 2022 0:23 UTC; 24 points)
- 's comment on [Link Post] If We Don’t End Factory Farming Soon, It Might Be Here Forever. by (7 Dec 2022 13:20 UTC; 17 points)
- 's comment on Project Idea: The cost of Coccidiosis on Chicken farming and if AI can help by (26 Sep 2022 21:43 UTC; 15 points)
- 's comment on What is Intersectionality Theory? What does it mean for EA? by (27 Mar 2022 23:53 UTC; 14 points)
- An aspirationally comprehensive typology of future locked-in scenarios by (3 Apr 2023 2:11 UTC; 12 points)
- Giving What We Can March 2022 Newsletter by (31 Mar 2022 15:03 UTC; 10 points)
- 's comment on A longtermist critique of “The expected value of extinction risk reduction is positive” by (14 Apr 2022 5:27 UTC; 7 points)
- 's comment on Who is protecting animals in the long-term future? by (23 Mar 2022 5:10 UTC; 3 points)
- 's comment on We Should Give Extinction Risk an Acronym by (19 Oct 2022 20:06 UTC; 2 points)
- 's comment on Reasons to have hope by (27 Apr 2023 16:13 UTC; 2 points)
- 's comment on Why we may expect our successors not to care about suffering by (12 Jul 2023 17:31 UTC; 1 point)
- 's comment on Why we may expect our successors not to care about suffering by (11 Jul 2023 13:06 UTC; 1 point)
Nice post! One intervention could be specifically arguing against including factory farming in space colonization plans. But methods and the tone for doing that might have to be very different from what animal advocates mostly use now. We’d have to be very careful and strategic about how we do that because doing it badly would make future efforts much more difficult. We could also fund and encourage research how to colonize space without animal farming.
If they get good at factory farming in space colonies, then it will be more difficult to switch to other options later. It’s like we are selecting which branch of a tech tree we will go for in a video game right now but it’s in real life.
Thank you Saulius! I basically agree with everything you said here. I would really hope some people from the space governance space can give us some insights here. Do you happen to know some of them?
Unfortunately, I don’t
Thanks for writing this post. I have similar concerns and am glad to see this composed. I particularly like the note about the initial design of space colonies. A couple things:
My sense is that the dominance of digital minds (which you mention as a possible issue) is actually the main reason many longtermists think factory farming is likely to be small relative to the size of the future. You’re right to note that this means future human welfare is also relatively unimportant, and my sense is that most would admit that. Humanity is instrumentally important, however, since it will create those digital minds. I do think it’s an issue that a lot of discussion of the future treats it as the future “of humanity” when that’s not really what it’s about. I suspect that part of this is just a matter of avoiding overly weird messaging.
It would be good to explore how your argument changes when you weight animals in different ways, e.g. by neuron count, since that [does appear to change things](https://forum.effectivealtruism.org/posts/NfkEqssr7qDazTquW/the-expected-value-of-extinction-risk-reduction-is-positive). I think we should probably take a variety of approaches and place some weight on each, although there’s a sort of Pascalian problem with considering the possibility that each animal mind has equal weight in that it feels somewhat plausible but also leads to wild and seemingly wrong conclusions (e.g. that it’s all about insect larvae). But in general, this seems like a central issue worth adjusting for.
Thank you for your comment!
Yes, I recognize that some longtermists bite the bullet and admit that humanity virtually only have instrumental values, but I am not sure if they are the majority, it seems like they are not. In any case, it seems to me that the vast majority of longtermists either think the focus should be humanity, or digital beings. Animals are almost always left out of the picture.
I think you are right that “part of this” is a strategy to avoid weird messaging, but I think most longtermists I discussed with do not think that humanity do not matter, probably especially with new longtermists. Also, naming of initiatives such as human compatible AI, value alignment, learning from humans, etc, makes me feel that these people genuinely care about the future of humanity.
And I am not even sure digital sentience is even possible, we haven’t even proven that it is possible, right? And I don’t even know how to think about the feasibility of digital sentience. Maybe you can introduce me to some readings?
I find the neuron count model implausible. 1. Human infants have more neurons than adult humans. 2. Some nonhuman animals have more neurons than humans (btw, I have some credence, albeit low, that some nonhuman animals have higher moral weights than humans, 1v1) . 3. Using the neuron count model would also create seemingly absurd prescriptions. The total number of nematode neurons exceed that of humans, which would prescribe a focus on nematodes more than humans, which would sound no less absurd than focusing all on insect larvae. (nonetheless, I don’t put 0 credence to these possibilities) 4. There are evidence that within humans, the capacity to suffer has great variety, down to the extreme which some humans barely ever feel pain or suffer, and there were no evidence these vast differences was because of neuron counts.
In any case, my aim for this post is literally to present the number of animals, while also made the case that I expect most of these animals to be either fish or insects. Essentially this leaves the readers to judge how they deduce the importance of farmed animals in their moral and cause prioritization.
Yes, all those first points make sense. I did want to just point to where I see the most likely cruxes.
Re: neuron count, the idea would be to use various transformations of neuron counts, or of a particular type of neuron. I think it’s a judgment call whether to leave it to the readers to judge; I would prefer giving what one thinks is the most plausible benchmark way of counting and then giving the tools to adjust from there, but your approach is sensible too.
Sorry that I missed your comment and therefore the late reply!
Thank you for sharing. Let me clarify your suggestion here, do you mean you suggest me to give my model of accounting for moral significance, rather than just writing about the number of beings involved?
Also, do you mind sharing your credence of the possibility of digital sentience?
Yes, that’s an accurate characterization of my suggestion. Re: digital sentience, intuitively something in the 80-90% range?
Another example for “V. Factory farming might spread to space—Some proponents of space colonization include factory farming in their plans”
″North Carolina State University is now taking applicants for its Nuggets on Mars program [ . . .] The multi–step, year–long program will involve teachers learning STEM principles behind poultry production and the unique challenges of raising chickens on Mars [. . .] At the end of the program, the participants will be tasked with putting together a unit for their class focused on developing ideas on ways to raise chickens on mars.” (Cottrell 2022)
I would sum this article up as “A speciesist society capable of tiling itself across the galaxy is a frightening one we should be actively working to avoid, and this conclusion is robust to a wide variety of future scenarios with respect to AGI, factory farming, wild animal suffering, and alien civilizations.”
This is kind of like my comment at the other post, but it’s what I could think of as feedback here.
--
I liked your point IV, that inefficiency might not go away. One reason it might not is because humans (even digital ones) would have something like free will, or caprice, or random preferences, in the same way that they do now. Human values may not behave according to our concept of “reasonable rational values” over time, as they evolve. In human history, there have been impulses toward the rational and the irrational. So they might for some reason prefer something like “authentic” beef from a real / biological cow (rather than digital-world simulated beef), or wish to make some kind of sacrifice of “atoms” for some weird far future religion or quasi-religion that evolves.
--
I don’t know if my view is a mainstream one in longtermism, but I tend to think that civilization is inherently prone to fragility, and that it is uncertain that we will ever have faster-than-light travel or communications. (I haven’t thought a lot about these things, so maybe someone can show me a better way to see this.) If we don’t have FTL, then the different planets we colonize will be far apart enough to develop divergent cultures, and generally be unable to be helped by others in case of trouble. Maybe the trouble would be something like an asteroid strike. Or maybe it would be an endogenous cultural problem, like a power struggle among digital humans rippling out into the operation of the colony.
If this “trouble” caused a breakdown in civilization on some remote planet, it might impair their ability to do high tech things (like produce cultured meat). If there is some risk of this happening, they would probably try to have some kind of backup system. The backup system could be flesh-and-blood humans (more resilient in a physical environment than digital beings, even ones wedded to advanced robotics), along with a natural ecosystem and some kind of agriculture. They would have to keep the backup ecosystem and humans going throughout their history, and then if “trouble” came, the backup ecosystem and society might take over. Maybe for a while, hoping to return to high-tech digital human society, or maybe permanently, if they feel like it.
At that point, it all depends on the culture of the backup society staying true to “no factory farming” as to whether they don’t redevelop factory farming. If they do redevelop factory farming, then that would be part of the far future’s “burden of suffering” (or whatever term is better than that).
I guess one way to prevent this kind of thing from happening (maybe what longtermists already suggest), is to simply assume that some planets will break down, and try to re-colonize them if that happens, instead of expecting them to be able to deal with their own problems.
I guess if there isn’t such a thing as FTL, our ability to colonize space will be greatly limited, and so the sheer quantity of suffering possible will be a lot lower (as well as whatever good sentience gets out of existence). But, say, we only colonize 100 planets over the remainder of our existence (under no-FTL), and 5% of them re-develop factory farming, that’s still five times as many as Earth today.
Thank you so much for this thoughtful, detailed, and concerning post.
Seems relevant: SpaceX: Can meat be grown in space?
A test to see if we can grow cultivated meat in space.
Wow thank you! Very relevant!
Suppose we transition to a universe filled primarily with digital people. Do you still believe that there are tons of farmed animals in the far future in expectation?
It depends on the probability one assigns to the scenario. If we assume we will 100% get that scenario, my upper estimates would shrink a lot, because presumably digital people would have little incentives to keep non-digital farmed animals. But unless the earth will also be replaced with primarily digital beings, my estimates for the expected number of farmed animals on earth in the far future might still roughly hold.
And it depends on what you mean by “primarily”. If that means some small portion of the universe will still be occupied by humans, the number of farmed animals could still be much bigger than “earth only” numbers.
Also, it depends on what you mean by “tons” here. If you meant many times more (i.e. 10,000x to 1 bil x) than the farmed animals that will exist in the next 100 years, I think the answer could still be yes with a far future universe filled primarily with digital beings. If you meant as many as the number of digital beings, then the answer is quite clearly no if we assume your scenario will happen (not suggesting that it will certainly happen), the ratio of non-digital farmed animals : digital beings would be a number very close to 0 , even though the absolute number of non-digital farmed animals is itself enormous.
Thanks! (I was mainly thinking about the ratio of digital people : animals, which is what matters for choosing between actions that help all digital people and actions that help all animals.)
I am glad I sort of answered your question!
It happens that I also worry about digital suffering, but I have two great uncertainties:
Whether artificial consciousness is possible.
If 1 is possible, whether these beings can have the capacity for positive and negative experiences.
My uncertainty in 1 is much greater, like maybe 100x to 2. I wonder what your credence in artificial sentience is? It would be very useful for me if you can share. Am I right about my guess that you think, even after adjusting for the probability of creating digital beings vs probability of space factory farming, you still think the expected number of digital beings is still greater (or the ratio of moral significance)?
(btw, you might have realised I said digital beings instead of people, I cannot think of reasons why there will be digital people but not digital animals, unless the word “people included them)
I’m pretty confident artificial consciousness is possible, though I haven’t looked into it much. This is primarily because it seems like consciousness will be a property of the cognition, and independent of the substrate running that cognition.
As an intuition pump, suppose we understand in great detail the exact equations governing the firing of synapses in the brain, and we then recreate my brain in software using these equations. I claim that, given an environment that mimics the real world (i.e. inputs to the optic nerve that are identical to what the retina would have received, similarly for the other senses, and outputs to all of the muscles, including the tongue (for speech)), that the resulting system would do exactly what I would do (including e.g. saying that I am conscious when asked). It seems very likely that this system too is conscious.
(I’m also confident that digital beings can have the capacity for positive and negative experiences.)
If you ask me about particular digital “beings” (e.g. AI systems, databases, Google search), then I become a lot more uncertain about (1) and (2).
Hi Fai,
This seems pretty unlikely, as animals would presumably be useful thanks to some biological process, but planet/asteroid mining is a mechanical process.
The answers to this post are also relevant to the inputs of your analysis. Carl Shulman says (I am not quoting the full answer):
Thomas Kwa says:
Hello again Fai,
On the topic of efficiency, I can see that:
However, I think this will also make the resources going to food production as a fraction of the world gross product go down over time, in which case changing the welfare per capita of factory-farmed animals will become increasingly cheap. Changing the number of animals would also become easier, but, if we stay on the Solar System, which I think is the plausible scenario for factory-farming continuing to matter, I guess human population and activities will eventually stabilise, and so would the population animals.
I have not thought much about this, but I guess interventions which make the lives of farmed/wild animals net positive longterm, like ones leading to expanding our moral circle[1], are more robustly beneficial than ones which aim to reduce the number of animals. Although this could be instrumentally good to mitigate economies of scale, I do not think cost is going to be a major determinant longterm.
Linking for reference. You know about it better than I do!
This is a very good point that I didn’t think about. It should be considered.
However, arguing that the badness of factory farming will stay is not the main task of this post. It’s mainly arguing that the number farmed animals in the long-term future might be high.
Also, on the welfare effect problem. I am not entirely sure it works the way you think it does. It might, but not certainly. Here are some counterarguments:
A short case study suggests that this didn’t happen, now. Eggs costs roughly 0.1% of humanity’s GDP. Cage-free (not free range) eggs costs roughly 20-40% to produce, which means roughly 0.02-0.04% of humanity’s GDP. But the world is not yet going 100% cage-free after two decades (or more) of campaigning, not even the developed countries who spend even lower % of their GDP on eggs. In fact, the developed countries have higher % of caged vs cage-free eggs. China is raising their % cage vs cage-free rapidly (and made it a national policy) as they rapidly decrease the % of GDP they spend on eggs.
I personally struggle to imagine how we can make farmed animals live net positive lives, I struggle even more to imagine how humanity would be willing to spend the cost make it happen to farmed animals (investing to do the research might happen as there will be people interested in making other animals happy, but I can’t imagine why factory farms will be willing to apply it, or that animal advocates would be willing to promote it)
We also need to think about the value conveyed and reinforced by factory farming staying. Even if the total amount of suffering decreases significantly, if humanity still continues to endorse a practice the harming (decreased by still harming) and killing of nonhuman animals for our trivial interests in the long-term future, it tells a very pessimistic story (and Bayesian update) about our ability to improve our values.
Nice piece, Fai!
FYI, I estimated the scale of the welfare of farmed and wild animals is 12.0 and 74.4 M times that of humans.
The feed conversion ratio cannot fall below 1, because mass cannot be created out of nothing. I guess values below 1 have not accounted for all feed (e.g. maybe food naturally present in the environment is not accounted for sometimes).
Thank you for your comment!
Mass cannot be created out of nothing, but feed conversion ratio, in weight terms, can fall below 1. This is because body tissues (mainly muscles) also consist of things other than protein + carbs, such as water. If the muscles have a high enough proportion of water, and the efficiency of raising gets efficient enough, FCR does indeed sometimes fall below 1.
But if feed conversion ratio is calculated using calories conversion or protein conversion, then it is always higher than 1.
Thanks for clarifying! I was thinking in terms of dry mass.
Thanks too, for clarifying what you thought!
I think in practice no one uses dry mass to calculate FCR, possibly for these reasons:
It makes the FCR higher, and therefore uglier (from the business perspective, talking about FCR is ugly to me no matter what)
They very rarely sell entirely dried meat, so the FCR doesn’t really tell the sellers how much meat they should expect to be able to produce by raising an animal.
I think farms/‘nature’/artificial environments in space habitats or planetary colonies are likely to be net-positive for animals larger than insects, because they would be taken better care of as a matter of course due to how scarce resources are. It is likely very difficult to spread unmanaged wilderness on an unterraformed planet, so it has to be managed and the animals treated well. Even moreso in an artificial habitat. This gives me hope for wild animal welfare. I invite argument otherwise, as reading this post was a small update against this position. This also hinges on terraforming being difficult, which could be wrong (life finds a way). Animals on a new planet might also just suffer more as they try to adapt.
-
My model is that cultivated meat will create a cultural inflection point that makes it easier to ban factory farming. Once factory farming is banned (roughly speaking), wild animal welfare will in turn be the next frontier.
I also don’t see organ transfers from animals as a significant factor, though it might be indicative of a trend of continued animal uses through other means. I have an image of my head of animals being bred for organ transfers being treated better, which could be wrong.
Unfortunately, I don’t think it is true that “they would be taken better care of as a matter of course due to how scarce resources are.” Our current model of animal agriculture is to operate the system incredibly efficiently overall but with no regard for the welfare of the individual animal. “Animal health” is a billion dollar industry constantly optimizing this. The animals are kept alive with drugs and careful diets but live short lives of physical and mental distress until their bodies quickly reach slaughter age. I imagine this system in space would be even more suffering-causing and unnatural than dark sheds. Chickens that died, for example, in a space CAFO would probably just be recycled to more chicken feed.
Hi Dony! Thank you for your comment!
I think I disagree with your view here. Let me explain why.
Consider these two objective functions:
Maximize the efficiency of raising tilapia (or any species of animals)
Minimize the chance that the tilapia raised live net negative lives
I think we shouldn’t expect that optimizing for 1 would always, robustly, ensure that 2 is also optimized at the same time. I think highly intelligent systems are quite likely to identify ways to optimize for 1 that do not optimize for 2 at all. In fact, we probably don’t need AI for that. I believe the sole reason that classical, non-intensive, family sized type of raising animals became factory farming exactly because the industry was interested in optimizing for 1, with no regard for 2. I understand that your argument is that, you believe, resource scarcity will make people take “good care” of the animals. But it seems clear to me that the reason they will take good care, is to optimize for 1 and only 1. Unless we can bioengineer a new species of animal of which optimizing for 1 always optimize for 2 at the same time, I am not convinced that we can confidently expect farmed animals in space live net positive lives.
Re: banning factory farming. I think an important consideration is whether this will be done before large scale space exploration and colonization. If factory farming already spread far and wide everywhere in the universe, the earth or nearby planets banning factory farming might not have major effects.