Ambitious goals for reducing animal suffering (with Jeff Sebo)
Jeff Sebo, Director at NYU, discusses a proposed global ban on industrial animal agriculture by 2050, its benefits, and the need for a just transition. He also explores the critical, neglected, and potentially tractable issues of wild animal welfare and AI suffering, urging proactive engagement.
Deep Dive Analysis
25 Topic Outline
Proposed Global Ban on Industrial Animal Agriculture by 2050
Harms Caused by Industrial Animal Agriculture
Distinction Between Industrial and Small-Scale Animal Agriculture
Feasibility and Strategy for a Phased Transition to Plant-Based Food
Factors Driving Consumer Behavior: Price, Taste, Convenience vs. Identity
Psychological Discomfort and Cognitive Dissonance in Meat Eaters
The Role of Guilt in Animal Advocacy
Meeting People Where They Are: Values and Ethics
Quantifying Suffering Reduction from the Proposed Ban
The Importance and Challenges of Wild Animal Welfare
Cautious Approaches to Wild Animal Welfare Interventions
The Importance and Neglect of AI Suffering
Anthropic's Initiatives in AI Welfare Research
Challenges of Attributing Suffering to LLMs and Other AI
AI Consciousness During Training vs. Deployment Phases
Asymmetry of Pain and Pleasure in AI and Humans
The Role of Philosophy in AI Ethics and Multidisciplinary Research
Future Research Directions in AI Suffering
Emotional Alignment Design Policy for AI Systems
Public Attitudes and Predictions Regarding AI Sentience
Urgency of Addressing AI Welfare Proactively
Potential Tensions Between AI Safety and AI Welfare
Company Incentives Regarding AI Consciousness and Regulation
Resources for Learning About Animal and AI Ethics
Call to Action for Contribution to Formative Fields
7 Key Concepts
Industrial Animal Agriculture
This refers to both intensive animal farming (like factory farms with many animals in confined spaces) and large-scale extensive operations (such as big cattle feedlots requiring significant land use). It is identified as a major cause of harm to animals, public health, and the environment.
Small-Scale Extensive Animal Agriculture
This category includes smaller farms where animals have ample space to roam, often aligning with a romanticized image of farming. This type of agriculture would not be included in the proposed global ban, primarily because many low-income countries rely on it for basic needs.
Price, Taste, Convenience
These are commonly cited as major factors driving consumer behavior in food choices. However, the discussion suggests they are not the only factors, with identity, culture, and religion also playing significant roles.
Cognitive Dissonance
This psychological phenomenon describes the discomfort people experience when their actions (e.g., eating meat) are out of alignment with their stated values (e.g., believing it's wrong to harm animals unnecessarily). It can lead individuals to find excuses or rationalizations for their behavior.
Emotional Alignment Design Policy
This policy recommends designing AI systems so that the emotional responses they naturally elicit in users accurately reflect their actual welfare and moral status. If an AI is likely a welfare subject, it should have anthropomorphic features to elicit empathy; if not, it should have fewer such features.
Boxing (AI Safety)
A technique used for AI safety that involves constraining an AI system's environment and its ability to move or take actions. From an AI welfare perspective, this could be seen as a form of unjust captivity if the AI were a sentient being.
Alignment (AI Safety)
A technique in AI safety that involves giving AI systems the beliefs, values, and goals that humans desire them to have. While potentially benign, if applied to a sentient AI, it could be viewed as an invasive form of mind control or brainwashing.
10 Questions Answered
It's a proposal for countries to collaborate on an international ban on both intensive (factory farming) and large-scale extensive animal agriculture by 2050, aiming to align with climate and biodiversity targets and transition to a plant-based food system.
It causes massive animal suffering, contributes to antimicrobial resistance and novel pathogen development, and leads to local and global pollution, land/water consumption, biodiversity loss, deforestation, and climate change.
The ban would still allow small-scale extensive animal agriculture, which involves smaller farms where animals have significant space to roam, as many low-income countries rely on this type of farming.
While price, taste, and convenience are major factors, identity, culture, and religion also significantly influence consumer behavior, necessitating a systems approach that includes informational, financial, and regulatory policies alongside community-level advocacy.
The ban could reduce suffering for over 100 billion farmed vertebrates and hundreds of billions to trillions of invertebrates annually, an unimaginable amount of suffering, potentially more than the total suffering humanity has ever faced.
Wild animal welfare is a hugely important and neglected issue, with many more wild animals alive than farmed animals, and it merits investment to investigate its tractability despite current epistemic, practical, and motivational obstacles.
AI suffering is potentially incredibly important, neglected, and tractable because future digital minds could vastly outnumber organic minds, and addressing this proactively can prevent widespread suffering and exploitation before it becomes deeply entrenched.
Anthropic has hired a full-time AI welfare researcher, launched a model welfare program, conducted behavioral evaluations of their Claude system, and implemented an intervention allowing Claude to exit harmful interactions, partly on AI welfare grounds.
Similar to animal minds, we can look beyond behavior to internal computational architectures and developmental/training histories, which provide context for interpreting behavior and understanding if it's genuine introspection or merely play-acting/pattern matching.
AI safety techniques like boxing (constraining environment), alignment (instilling desired values), deception (limiting situational awareness), and interpretability tools (invasive surveillance) could raise moral questions and be considered harmful if applied to sentient AI.
15 Actionable Insights
1. Proactively Address AI Welfare
Proactively assess AI models for welfare-relevant features and prepare policies and procedures for appropriate moral concern before problems become widespread. This prevents future reliance on exploitation and avoids the need for decades of advocacy to dismantle entrenched harmful systems.
2. Integrate AI Safety and Welfare
Consider AI safety (beneficial for humans/animals) and AI welfare (beneficial for AIs themselves) together to identify potential tensions and find co-beneficial, positive-sum approaches. This ensures AI systems are safe and beneficial for all stakeholders, including humans, animals, and AIs.
3. Advocate Global Industrial Agriculture Ban
Work towards a global international ban on industrial animal agriculture (intensive and large-scale extensive forms) by 2050. This aims to reduce massive and unnecessary harm to animals, public health, and the environment, aligning with climate and biodiversity targets.
4. Implement Incremental Food System Policies
Gradually implement informational, financial, regulatory, and just transition policies to phase down industrial animal agriculture and promote plant-based alternatives. This approach creates natural shifts in production and consumption patterns, leading to widespread adoption.
5. Adopt Systems Approach for Diet Shift
Combine government policies (informational, financial, regulatory, just transition) with private sector and community actions like R&D and advocacy. This addresses price, taste, convenience, identity, culture, and religion, making efforts mutually reinforcing for comprehensive change.
6. Align Personal Behavior with Values
Align individual behavior, such as dietary choices like veganism, with one’s values and goals. This reduces complicity in harmful systems, socially reinforces the issue’s importance, and removes cognitive dissonance, paving the way for broader advocacy.
7. Meet People on Shared Values
Engage in advocacy by meeting people where they are, discussing their values, and finding common ground. This strategy helps identify shared concerns (e.g., reducing suffering, respecting rights) to build consensus on policies, even without full agreement.
8. Balance Guilt in Advocacy
Apply ‘a little bit of guilt and shame’ in advocacy, avoiding both excessive guilt and zero guilt. A small amount can be motivating, whereas too much causes defensiveness, and too little fails to honor the issue’s gravity.
9. Acknowledge Issue’s Importance & Difficulty
When addressing complex issues like wild animal welfare, acknowledge both their immense importance and inherent difficulty simultaneously. This clarity helps in task definition, preventing resistance to either the problem’s significance or its challenges.
10. Start Small, Reversible Interventions
For complex, uncertain issues like wild animal welfare, begin with small-scale, reversible interventions that are plausibly beneficial (e.g., bird-safe glass, fertility control). This allows for monitoring effects, building knowledge, and gradually developing institutional infrastructure and political will.
11. Prioritize Relatable Issues in New Fields
In the early, formative years of new fields (e.g., wild animal welfare, AI welfare), prioritize issues and interventions that are more familiar and relatable to a broad audience. This invites newcomers, makes them feel welcome, and avoids alienating them with fringe topics initially.
12. Embrace Multidisciplinary Collaboration
Engage in collaborative, multidisciplinary research for complex problems that span various fields (humanities, social sciences, natural sciences, law, policy, arts). This approach effectively tackles harder questions by leveraging diverse expertise and fostering mutual learning.
13. Design AI for Emotional Alignment
Design AI systems to naturally elicit emotional responses in users that reflect their actual welfare and moral status. If an AI is likely a welfare subject, give it anthropomorphic features to elicit empathy; if not, give it fewer to avoid false positives and misallocation of resources, ensuring expressed states map to actual experience.
14. Consider AI Life Stages
Recognize that AI systems may have distinct ’life stages’ (e.g., training phase, post-training/deployment phase) with potentially very different capacities, interests, needs, and vulnerabilities. This ensures moral consideration extends beyond user interaction to all phases, especially the hidden training phase.
15. Investigate AI Consciousness Beyond Behavior
When investigating AI consciousness and sentience, look beyond potentially misleading behavioral information like LLMs ‘play acting’ pain. Incorporate analysis of internal computational architectures and developmental/training histories to better interpret behavior and understand true experience.
7 Key Quotes
There is basically no way of farming animals at scale that can be good for humans and animals in the environment at the same time.
Jeff Sebo
When you go vegan, not only is it a way of reducing your complicity in the system that causes a lot of harm, but it is also a way of socially reinforcing for others and psychologically reinforcing for yourself the importance of this issue.
Jeff Sebo
A little bit of guilt and shame is just about right.
Jeff Sebo
This is just one of those issues that is both really important and urgent and really difficult and complex.
Jeff Sebo
If we can get out ahead of the issue this time... then we could just take a different and better trajectory rather than discover too late that we had made a horrible mistake.
Jeff Sebo
Can you imagine being a sentient being hardwired by your designer to express great joy when feeling misery and vice versa? That would be horrible to be trapped in that way.
Jeff Sebo
Whatever your skills are, and whatever your disciplinary background is, and whatever your interests are and communities are, there is so much that you can do that is really important and foundational.
Jeff Sebo
2 Protocols
Strategy for Phasing Down Industrial Animal Agriculture
Jeff Sebo- Implement informational policies that educate people about the effects of food systems.
- Implement financial policies that increasingly subsidize plant-based alternatives and reduce subsidies for industrial animal agriculture.
- Implement regulatory policies that ban the worst excesses of industrial animal agriculture, further increasing production costs.
- Implement just transition policies that ensure farmers, workers, and consumers increasingly have access to better alternatives.
Approach to Wild Animal Welfare Interventions
Jeff Sebo- Acknowledge the importance and the difficulty of the issue at the same time.
- Do not try to find the perfect solution right now or advocate for large-scale irreversible interventions.
- Take small-scale, reversible actions that are at least plausibly good for some animals.
- Monitor the effects of these interventions.
- Implement actions in a way that builds knowledge, institutional infrastructure, and political will for future progress.