Thanks for this thoughtful and well-structured response.

I agree that the arrival of Precision Livestock Farming (PLF) in industrial animal agriculture is likely, if not inevitable, and that complete disengagement would be a strategic error. The key question, as you frame it, is how we should respond given that inevitability i.e. how we can influence trajectories in ways that mitigate harm and preserve our longer-term strategic goals.

Where I’d propose a refinement to your view is in the specific structure and framing of the engagement model. I support a “multiple bets” strategy in theory, but I would argue that with PLF in particular, the nature of each bet matters greatly, especially in terms of long-term systemic lock-in.

1. Engagement Does Not Necessarily Require Cooperation

PLF presents a unique case in that it is not simply a welfare-relevant technology, it represents a structural optimization tool that increases the scalability, opacity, and economic viability of intensive animal farming systems. That is, even in its “best-case” deployments, it likely reinforces the underlying system we aim to replace.

Engagement, therefore, need not mean co-development or partnership with PLF actors. There are other legitimate and impactful modes of engagement, including:

• Regulatory friction (e.g., mandating transparency, requiring human oversight)
• Strategic litigation
• Public campaigning aimed at investor and consumer skepticism
• Cross-sector coalition-building with labor, privacy, or anti-monopoly actors

These forms of non-cooperative engagement allow us to remain actively involved in shaping outcomes without legitimizing or entrenching PLF’s role in the future of food production.

2. The Good Cop / Bad Cop Model Should Be Deliberately Asymmetric

Your suggestion that some actors could play a “good cop” role while others adopt a more oppositional stance makes sense in principle. However, I would propose an important modification to the typical good cop / bad cop framework: the good cop should be engaging not with industry, but with governments and regulators, and their function should be restrictive, not constructive.

Specifically:

• “Bad cop” groups should argue for outright bans on fully automated factory farms.
• “Good cop” actors can propose moderate but strictly framed regulatory constraints (e.g., minimum welfare baselines, human presence requirements, or legally binding data transparency standards).

This model creates a strategic gradient: opposition at the moral and public level, paired with narrowly scoped institutional engagement that increases friction, compliance costs, and reputational risk.

3. Global Arbitrage Opportunities Can Be Weakened

You rightly point out that if countries like China adopt PLF aggressively, efforts to restrict it in Europe or North America may result in the offshoring of suffering. This is a valid concern. However, I would caution against using global arbitrage as a justification for preemptively conceding ground on domestic policy.

Instead, the strategic response could include:

• Advocating for import standards and trade restrictions based on animal welfare criteria
• Building alternative protein innovation ecosystems that reduce domestic reliance on high-intensity imports

Summary

I agree that:

• PLF presents significant and urgent risks.
• Advocacy must engage with it rather than ignore it.
• A pluralistic strategy is warranted.

Where I’d recommend sharpening the approach is in how we distribute and constrain roles:

• We should avoid positioning any actor as a constructive collaborator with PLF developers or agribusiness coalitions.
• Instead, advocacy should be bifurcated into (a) principled opposition focused on public narrative, and (b) pragmatic regulation focused on limiting harms through oversight mechanisms.
• The narrative must remain clear: PLF is a threat to animal welfare, food system justice, and democratic accountability. Our engagement exists not to optimize it, but to slow, restrict, and, where possible, prevent its expansion.

Thanks Aidan, really thoughtful response, I appreciate you taking the time to engage with this.

Let me clarify a few points that I think get to the heart of our disagreement.

1. On “net positive” factory farms

I think the idea that factory farms could someday produce “net positive” lives misunderstands the nature of factory farming, both structurally and ethically.

Structurally, factory farming is defined by its goal: to maximize output while minimizing cost. That necessarily imposes constraints that are incompatible with what we’d consider high welfare, things like space, autonomy, species-typical behavior, social bonding, and individualized care. If a system provided those things, it wouldn’t be a factory farm in any meaningful sense of the term.

So I don’t think this is just a question of philosophical pessimism or optimism about wellbeing, it’s an economic and definitional constraint. Essentially, factory farms can’t become welfare-positive, because doing so would mean ceasing to be factory farms.

On top of that, there’s a deeper ethical issue that’s often sidelined in welfare debates: killing is itself a harm. Even if a farm animal experiences moments of pleasure, they are ultimately bred into existence for premature death. That death (especially when non-consensual and systematically commodified) represents a major negative welfare event.

We rightly acknowledge this in human ethics: even lives with some suffering are considered worth preserving, and death is treated as a profound harm, not a neutral endpoint. There’s no reason to apply a different standard to nonhuman animals, especially when their strong behavioral avoidance of death indicates it matters to them too.

If we accept that the premature killing of sentient beings is a major harm (and that welfare is not just about net pleasure during life, but includes the deprivation of future wellbeing) then the idea of a “net positive” life that ends in commodified slaughter starts to fall apart.

To illustrate this more clearly: imagine we had an industry that bred and killed human infants for consumption, ending their lives when they reached the equivalent cognitive development of a pig or a chicken. Even if these infants experienced "net-positive welfare" during their short lives, we would not see that as a justification for killing them. We should apply the same ethical reasoning to animals, whose interests in continued life are no less real.

2. On reform vs abolition

I absolutely agree that reform can be a pathway to abolition, and I’ve written elsewhere about the importance of non-reformist reform as a strategic tool. This distinction matters: some reforms genuinely weaken the economic and cultural foundations of exploitative systems, while others inadvertently reinforce them.

As a rough heuristic: does the reform increase or decrease industry profitability and resilience? Cage-free transitions, for example, often raise costs, reduce scalability, and create a visible narrative of harm. These can support broader movement goals and act as stepping stones towards the abolition of factory farming. Others, like PLF, do the opposite.

So I’m not against reform. I’m against reforms that entrench harm.

PLF is a clear case of the latter: a technological intervention that reduces costs, scales the system, increases opacity, and hardens industry infrastructure, making factory farming more difficult to regulate, challenge, or replace.

I’ve intentionally chosen not to name individuals or organizations here, not because I haven’t seen support for PLF (I've seen it from across the movement: academics, NGOs, individual advocates, funders), but because I think the conversation is more productive when it stays focused on ideas rather than people. 

Within the animal advocacy movement, I have a lot of respect for those exploring or supporting PLF and I believe that in most cases, their motivations come from a sincere desire to reduce suffering and make tangible progress for animals.

By keeping the focus on strategy and system-level impacts, I’m hoping to avoid unnecessary division and instead invite reflection on how different interventions interact, align, or conflict with broader movement goals. I think we do our best work when we interrogate ideas rigorously, assume good intentions and stay anchored in our shared values. 

That’s why I’ve chosen not to single out individuals for engaging with this area of work. I also believe in movement ecology, the idea that strong movements benefit from diverse approaches. Disagreements about strategy are both inevitable and healthy, as long as they’re grounded in mutual respect and shared goals.

For anyone reading this who’s involved in PLF-related work, I want to emphasize that I’m always open to dialogue, collaboration and shared learning. Differing views on how to achieve change should never be a barrier to working together. We can disagree on strategy whilst still recognising that we are on the same side.

Thanks for your reply and for clarifying your perspective. I do agree that the most harmful applications of PLF technology we’re currently seeing are driven by machine learning and deep learning, rather than generative AI. When I refer to AI in factory farming, I’m using the term in its broader sense to include these technologies as well—beyond just large language models specifically.

On the main point, I think campaigns for restrictions or bans on AI in factory farming can actively strengthen the push for transparency, rather than being at odds with it.

Broadly speaking, transparency campaigns without accompanying pressure tend to fail across cause areas. Companies are unlikely to willingly share data unless there’s significant public scrutiny or regulatory threat. Calls for a ban increase that scrutiny by raising public awareness about the risks AI poses to animals, highlighting the need for accountability and uniting broad coalitions that increase political power.

The risk, if the movement focuses solely on promoting “positive” uses of PLF, is that we create an environment where welfare washing and complacency thrive. Companies will only adopt welfare improvements where they align with profitability, and even then, these measures are often incidental rather than intentional. In many cases, welfare "improvements" serve to entrench factory farming further, creating the illusion of progress whilst masking systemic harm. For example, technologies that reduce disease outbreaks may allow producers to justify increasing stocking densities, leading to even greater overall suffering, despite the initial appearance of progress.

To meaningfully challenge these systems, we need radical counterpressure—calls for bans or restrictions. Without this counterbalance, we increase the probability that AI will cement factory farming's dominance rather than dismantle it. History shows us that meaningful action—particularly changes that hurt industry interests—rarely happens without radical demands to push the boundaries of what’s politically acceptable.

Campaigns for bans aren't in opposition with calls for transparency, they're a strategic neccessity in achieving them. They apply the pressure needed to drive reforms, expose harmful practices, and keep the ultimate goal—fighting factory farming—at the center of the conversation. Without this pressure, transparency risks becoming toothless, co-opted as a tool for welfare-washing or superficial improvements that merely serve industry interests. Coupling bold demands for bans with transparency-focused efforts ensures that any improvements are not only genuine and accountable, but also prevent the illusion of progress from entrenching the very systems we aim to dismantle.

In this way, the two strategies can complement each other: bold calls for bans provide the pressure and visibility needed to make transparency campaigns more effective.

Hi @Wladimir J. Alonso and @saulius,

First of all, I want to emphasise that I see value in both approaches—advocating for the abolition of AI in factory farms and pushing for welfare-oriented reforms. These strategies are not contradictory, rather, they can complement each other to achieve broader progress. Radical proposals shift the Overton window, making moderate reforms appear more reasonable, while moderate approaches secure practical wins and build momentum for more ambitious goals.

That said, I remain skeptical that AI in factory farms will have a net-positive impact on animals. The gap between technological development in academia and its real-world implementation in industry has historically favoured profit maximisation at the expense of welfare. For example, CRISPR gene editing was initially intended to address genetic defects but has instead enabled selective breeding that exacerbates welfare issues—like chickens bred to grow so quickly their bodies cannot support their weight.

The argument that factory farming cannot get worse through further optimisation strikes me as overly optimistic. AI is already contributing to worsening conditions:

• The rise of 26-story AI-powered pig farms in China, where extreme intensification creates severe welfare risks.
• The replacement of soy-based feed with black soldier flies housed in AI-automated containers for chickens, further entrenching exploitative systems.

Historically, unforeseen consequences of new technologies—like antibiotics enabling extreme overcrowding—have harmed animals, and it’s unrealistic to assume that future AI breakthroughs won’t follow similar patterns. This research paper outlining 12 harms caused by precision livestock farming provides a useful starting point for thinking through some of these concerns.

On the question of bans, I agree that such campaigns are unlikely to succeed in the near term in the U.S., but the political dynamics in other regions—like Europe, Australia, and New Zealand—are different. In these contexts, smaller political parties (e.g., animal justice and green parties) hold influence and could plausibly campaign for bans or significant restrictions on AI in factory farms.

Importantly, campaigns for a ban have value beyond their immediate outcomes. They can:

• Build broader coalitions across political divides, uniting meat industry labor unions (concerned about job losses) with environmental and animal welfare advocates.
• Drive moderate reforms by setting ambitious demands, which create space for compromise and incremental progress.
• Establish global precedents that legitimise concerns about AI and animal welfare, shifting the political landscape toward stronger regulations and, eventually, abolition.

In short, while a ban may not be immediately tractable, campaigning for one could yield significant benefits: building coalitions, achieving meaningful reforms, and paving the way for larger victories down the line. Ultimately, both approaches—welfare reforms and abolition—can reinforce one another, driving the systemic change we all aim for.

This is a notoriously hard problem to measure overall (there's lots of variation in actual consumption vs reported diets, social desirability bias etc.), but there are several easier sub-sections of the problem that we can more easily measure and they tend to show exponential growth.

We see this exponential pattern in the growth of vegan restaurants in Europe, the percentage of the UK population identifying as vegan and the number of products labelled as vegan worldwide, just as a few examples.

As un-scientific as it is, I also think the anecdotal evidence from long-term vegans is worth considering, Most people who have been vegan 10+ years (myself included) will acknowledge that the rate of growth over the last 5 years has been significantly faster than the 5 years before that across virtually every metric, from the number of vegans you meet in everyday life to the number of restaurants and products available to the overall attitude that the public has towards veganism etc.

My point was that the 10-70% range reflects different outcomes depending on the actions we take, not just optimism as a feeling or a belief. Optimism can certainly motivate us to act, but without those actions, it has little to no impact on the actual probabilities.

It seems our main disagreement lies in how we view these probabilities. I see them as dynamic and heavily influenced by the actions we take as a movement, while you seem to view them as more static and inherent to the situation, essentially outside of our control.

I think that's an extremely important distinction because it fundamentally shifts how we approach this challenge. If we believe the odds are fixed, we become passive observers, resigned to whatever fate has in store. But if we recognise our power to influence those odds through strategic action, technological innovation, and effective execution, we become active participants in creating the future we want. This empowers us to take responsibility, to strive for optimal solutions, and to push beyond the limitations of the status quo.

To better understand your perspective, could you provide your estimated probabilities for ending factory farming by 2060, considering these scenarios:

• Scenario 1: Complacency. We maintain the status quo, with minimal changes to our current approach. We fail to identify and effectively target the key pressure points within the system, and we don't create the necessary feedback loops to amplify our impact and hinder the growth of industrial animal agriculture.
• Scenario 2: Moderate Improvement. We make incremental progress in our strategies and adoption of new technologies, but we don't fully capitalize on opportunities for exponential growth. We achieve some success in identifying and influencing key pressure points, but our efforts are not comprehensive or optimally coordinated.
• Scenario 3: Optimal Execution. We proactively identify and exploit every opportunity for exponential growth within the movement. Simultaneously, we precisely target the most influential pressure points within the system and create powerful feedback loops that accelerate our progress while hindering the expansion of industrial animal agriculture. We achieve maximum coordination and efficiency, fully leveraging every available resource and opportunity with optimal strategic foresight.

I'm really interested in seeing how your probabilities compare across these scenarios, especially for scenarios 2 and 3. While 1-5% might seem understandable (albeit quite pessimistic) for scenario 1, where we assume minimal change, it seems considerably less likely that those odds wouldn't drastically increase with the improved strategy and execution described in scenarios 2 and 3.

To put it into perspective, imagine a basketball team with a 1-5% chance of winning a game. If they then acquire a star player, develop a brilliant new strategy, and execute it flawlessly, wouldn't their chances of winning increase substantially?  Similarly, in the fight against factory farming, if we effectively leverage exponential technologies, coordinate our efforts optimally, and execute our strategies with precision, it seems almost impossible that our probability of success wouldn't see a major boost.

1. It's important to emphasise how much our actions as a movement influence those odds. We're not just bystanders, our strategies, dedication, and execution all play a role. That's why I've given a range of probabilities based on the actions we take as a movement. Each individual member has the power to decide their level of contribution—0% or 100%—so that probability is something everyone can decide for themselves.
2. You're right about the balance needed with optimism. Over-optimism about one solution, like cultivated meat, can hinder exploration of other options. But optimism about our overall goal is a different story. It leads to self-fulfilling prophecies: if we don't believe ending factory farming is possible, we automatically decrease the probability. But if we believe it's possible and that our actions determine the outcome, we massively increase our chances of success.
3. Yes, I use an LLM for almost everything I write. I usually draft my ideas and then refine them through a conversation with the LLM, making them clearer and easier to understand. This saves me time and improves the quality of my writing. I'm also autistic and sometimes find it challenging to get the right tone across in my writing, and LLMs helps me with that too.

By "ending factory farming," I mean a 95% reduction in animals raised in intensive industrial farming operations globally by 2060. Predicting the likelihood of this is complex, but I'd estimate it as:

10-30% if we continue with current strategies and resource allocation.

• Pros: Growing veganism, plant-based options, investment in alternatives, and public awareness are all positive signs.
• Cons: Cultural habits around meat consumption, powerful industry lobbies, and potential increased meat consumption in developing countries pose serious challenges.

35-50% if we effectively leverage exponential technologies and focus on strategic leverage points.

• Pros: AI-powered advocacy shows promise, and targeting key global hubs can create ripple effects. The movement itself appears to be entering a phase of rapid growth, and history suggests society tends to expand its moral circle over time.
• Cons: AI advocacy may need a strong global movement to be effective across diverse cultural contexts. The identified leverage points may not be as influential as predicted, and preventing animal agriculture from leveraging those same technologies to compete is crucial.

55-70% if we achieve exceptional movement coordination and execute optimally on key interventions.

• Pros: Combining exponential social movement growth with technological advancement creates immense potential for rapid change. AI advocacy is proving effective, and a systems-level approach generates powerful feedback loops.
• Cons: Achieving and sustaining global coordination is incredibly difficult. Unforeseen consequences, potential loss of momentum, and adaptation by the industry are all risks.

These are rough estimates, and many unknowns could influence the outcome. It's easy to get caught up in predictions, but the future of factory farming rests in our hands. Our strategies, dedication, and ability to overcome challenges will ultimately determine success or failure.

Instead of fixating on a fixed probability, we should adopt a mindset of radical responsibility. 

Every animal advocate can shift the odds. Imagine two extremes:

• Scenario 1: Complacency. We lose focus, funding dries up, infighting weakens the movement, and opportunities are squandered. The probability of ending factory farming plummets towards 0%.
• Scenario 2: Optimal Execution. We embrace technology, forge alliances, and execute flawlessly. We inspire millions, drive innovation, and hold industry accountable. The probability of success skyrockets towards 100%.

The reality will fall somewhere in between. But the key takeaway is this: we are not passive observers, we are active participants in shaping the future for animals.

One major obstacle I see is the slow rate of adoption of AI by animal advocates. Currently, about 50% of animal advocates rarely or never use AI in their work: https://www.openpaws.ai/research-and-reports/report-on-the-use-of-ai-in-animal-advocacy

Funding is another major obstacle, we clearly don't have the resources to compete with animal agriculture on computing power. That's why I think our best bet is open sourcing models and data (which animal agriculture won't do because they give them a competitive advantage) and leveraging the power of a passionate community to improve our models, rather than "throwing money at the problem". 

Whilst it's not really an issue of exponential growth not applying to animal advocates, one other major concern is that exponential growth can also apply to the animal agriculture industry, as @GoodHorse413🔸 pointed out. I think that's a threat we should take very seriously as a movement and something we should aim to disrupt through a combination of lobbying for legislative changes and engaging in corporate campaigns to restrict or ban various uses of AI in factory farms and slaughterhouses.