Asymmetries, AI and Animal Advocacy

Kevin Xia 🔸16 May 2025 6:16 UTC
69 points
6 comments5 min readEA link

Thanks to Jeff, Max Taylor, Joanna Michalska, Albert Didriksen and Koen van Pelt for feedback on this post. All mistakes are my own. This post does not necessarily reflect the views of my employer.

AI development is happening fast and the implications could be numerous and fundamentally transformative for the animal movement. There is a lot to be won, but a lot at risk as well.[1] But I don’t think one needs to be convinced of short AGI timelines to pay close attention to what is going on in AI. You just have to be convinced that as a movement, we are (a) dramatically under-invested in AI for animals, given the likelihood of things changing quite radically, (b) under-prepared for how things may change eventually (whether this is in 3 years or 30) and/​or (c) that there is little to lose by thinking ambitiously about, and safeguarding against, the possibility of short AGI timelines. This piece intends to give a framework for thinking about the opportunities that may arise in our attempt to leverage AI to dramatically benefit animals.

What’s the problem?

It seems that many interventions at the intersection of AI and helping animals can be applied quite symmetrically;[2] that is, animal advocates and industrial animal agriculture can use very similar interventions in very similar ways, potentially cancelling each other out. For example:

  • AI-driven research could dramatically reduce alternative protein costs—but the same technology will likely help animal agriculture cut down expenses for conventional animal products.

  • AI could supercharge productivity for animal advocacy nonprofits—but identical tools will simultaneously boost efficiency across the industrial farming complex.

  • AI enables hyper-targeted, individualized mass outreach for animal advocates—but may equally enable animal agriculture marketers to craft hyper-targeted, individualized advertisements.

  • AI could help research strategies for effective animal advocacy—but could also help the industry to research and develop sophisticated counter-measures against advocacy efforts.

This symmetry problem creates a particularly interesting challenge: how can we leverage AI to create meaningful, lasting change for animals that cannot be neutralized by mirrored/​parallel developments in the opposing sector? To be clear, it will be important for our movement to work on these symmetrical interventions, ensuring that we don’t fall behind too much. Sure, getting alternative proteins cheaper may not be the holy grail given that animal products will drop in price as well—but if we don’t match that progress, we will fall much further behind. So in order to prevent things from getting worse, each of these interventions is still likely to be significant.[3] But in order to fully take advantage of AI development, it may be useful to identify and exploit fundamental asymmetries in our favor.

What are asymmetries?

I think of asymmetries as something like a comparative advantage that one group has over the other and thus of asymmetrical interventions as opportunities that arise only for them. There are many asymmetries that swing in favor of industrial animal agriculture (e.g., money, political networks, status quo), so over the past few weeks, I have attempted to find those that favor animal advocates. While I am (hopefully) far from having all or even the most important ones, I believe it is important to think through the intersection of AI and animals through this lens and hope readers and advocates will explore these further and find additional ones. Here are some ideas and corresponding example interventions.

  • Motivational asymmetry: Ultimately, we (animal advocates) care about reducing animal suffering/​care more about reducing animal suffering than consumers care about eating harm-causing animal products or than the industry cares about harming animals.

    • For example, we could imagine a transformed mass entertainment and pleasure-abundance world, where everyone has all their needs met. In this world, we may be less likely to be distracted. If we can then change the industry “behind the scenes” without cutting into consumer preferences or profit incentives/​without “bothering” anyone, we may have an upper hand.

  • Truth asymmetry: Ultimately, we are “more” right about the harm done to animals on various grounds: the moral alignment with consumers (animals are treated worse than consumers would like), the legality of some/​much of the cruelty involved in animal farming and, for the moral realists among us, the morality of harming non-human animals.[4]

    • For example, we could imagine maximally persuasive marketing campaigns or AI lawyers on either side of the debate and may have reason to hope that they would ultimately swing in favor of animal welfare efforts.

  • Efficiency asymmetry: Ultimately, alternative proteins are likely inherently more efficient conversion systems than animal farming.

    • For example, AI may disproportionately optimize alternative proteins and accelerate their economic advantage by bringing costs lower than is possible for the equivalent animal product (notwithstanding other influences on purchasing decisions).

  • Agility asymmetry: Alternative protein startups and advocacy organizations can pivot and adapt to new AI capabilities more quickly than entrenched animal agriculture systems with massive physical infrastructure investments.

    • For example, in accelerated AI development scenarios, political institutions may be more likely to regulate and slow down many efficiency-improving applications in animal farming, such as mass automation of jobs, due to its impact on employment. The same may not be true for ways to improve alternative proteins, potentially giving us a window of opportunity where alternatives can drastically undercut prices of conventional animal products, inducing important social norms, enabling narrative changes and maybe even re-investments from conventional animal farming to alternative proteins.

  • Ceiling asymmetry: As the animal advocacy movement, we have more to gain with AI technologies, given that we are currently less able to deploy the best possible ways we run interventions due to resource constraints.

    • For example, the industry may already be able to pay for the best human marketers, lobbyists and company leaders. AI could enable animal advocacy organizations to catch up by leveling the playing field (for example, see automation through deskilling).

  • Cooperation asymmetry: Animal advocacy organizations tend to be more collaborative and less competitive with one another, while industrial animal agriculture operates in competitive market structures that may inhibit industry-wide coordination on AI adoption.

    • I did not think of any feasible intervention associated with this asymmetry, but broadly there seem to be quite a few disadvantages of competitive markets that we could exploit, such as race-to-the-bottom dynamics (companies undercutting one another in price and thus profits) and fragmentation (companies being less united and coordinated).

And I am sure there are many more.[5] These asymmetries aren’t merely theoretical—they represent potential leverage points where the strategic application of AI technologies could create disproportionate advantages for animal advocacy that industrial animal agriculture cannot easily counter.

An alternative way of framing

Finding and leveraging asymmetries is not easy, and fighting the numerous asymmetries facing us may be even harder. But the asymmetry framing may miss crucial nuance and we could benefit from reframing our advocacy: We are not necessarily in a pro-animal vs. anti-animal dynamic. We are pro-animal, but they are just pro-money, pro-influence, pro-workers or pro-feeding-the-world.[6] If, then, we can find convergence in making the economic case for abandoning factory farming, we may not need to circumvent these asymmetries.

This reframing shifts our approach from confrontational to collaborative, potentially opening doors to partnerships with forward-thinking industry players who recognize the economic writing on the wall. In fact, AI could even be instrumental in developing and communicating compelling economic models that demonstrate how transitioning away from animal agriculture represents not just an ethical imperative but a financial opportunity.

Conclusion

In order to not fall behind, we need to leverage AI wherever we can, but in order to drastically take advantage of what AI may have to offer, it may be helpful to think in terms of asymmetries. After all, the risks and opportunities involved are enormous and it doesn’t hurt to think about our comparative advantages anyway. Feel free to reach out if you have further ideas about asymmetries and corresponding interventions, I would love to hear them. Some more things you can do:

  1. ^

    I personally work under the (outrageous) ambition of ending factory farming in a decade or so but with the fear of it being even more deeply entrenched, locked in and scaled up 3-5x in size.

  2. ^

    Think dual-use concerns, but for industrial animal agriculture vs. animal advocacy efforts.

  3. ^

    In fact, we don’t have any strong reason to expect that our best interventions will not be symmetrical. We should always prioritize output/​impact and treat (a)symmetries as a framework and as proxies with nuance.

  4. ^

    The “moral” truth angle may not necessarily help us much beyond the “consumer alignment” angle—unless we assume AI agents to find and act on moral truths.

  5. ^

    Perhaps, one could explore what the offense/​defense balance can tell us about finding asymmetries in the AI-animal advocacy space or what conventional industry strength (such as political entrenchment) could be turned into a weakness.

  6. ^

    Some of these motivations may be more charitable interpretations than others.

Kevin Xia 🔸16 May 2025 6:16 UTC
69 points
6 comments5 min readEA link