Quick Summary
Current AI systems display severe speciesist biases and the industries that harm animals are already decades ahead of the animal rights movement, both in utilising existing AI tools and in developing their own.
Our newly launched nonprofit Open Paws is dedicated to building open source AI specifically for the animal rights movement, providing free technical support to nonprofits and working with major AI labs to ensure that the future of AI benefits all sentient beings.
We are currently seeking funding and estimate that we would save 20-70 animal lives per dollar raised, making Open Paws approximately 5-17 times more cost-effective than the average Animal Charity Evaluator recommended charity.
Further details can be found in our Pitch Video and Funding Proposal.
Research shows AI is speciesist
Current AI systems like ChatGPT have rapidly gained popularity in recent times, but they exhibit a concerning bias against farmed animals.
"The more an animal species is classified as a farmed animal (in a western sense), the more GPT-3 tends to produce outputs that are related to violence against the respective animals."
Speciesism, or discrimination based on species, can harm efforts to protect animals and drive positive dietary changes, and correlates with other prejudices like racism and sexism.
Sources:
- Speciesist language and nonhuman animal bias in English Masked Language Models
- Speciesist bias in AI: how AI applications perpetuate discrimination and unfair outcomes against animals
- The Moral Standing of Animals: Towards a Psychology of Speciesism
AI is Driving Animal Exploitation
The industries that harm animals not only benefit more from these speciesist systems than our movement does, they are also heavily invested in developing their own custom AI that allows them to exploit more animals more efficiently.
McDonald's has their own AI lab that automates their digital menu boards to increase sales, JBS (a large meat processing company) has their own proprietary AI for sorting carcasses in slaughterhouses and factory farms have specialised AI to reduce their operating costs.
In contrast, animal rights organizations are massively lagging behind in both adopting and developing contemporary AI tools.
Soon, AI will be smarter than humans
"If science continues undisrupted, the chance of unaided machines outperforming humans in every possible task [is] estimated [by a survey of thousands of AI experts] at 10% by 2027, and 50% by 2047."
If these superintelligent AIs of the future retain the same speciesist biases they do today, animal exploitation could become entrenched forever in history, making animal liberation impossible for humans to achieve.
Sources:
- Thousands of AI authors on the future of AI
- Shared Path - Service Design and Artificial Intelligence in Designing Human-Centred Digital Services
Our Initial Traction
Here's a small sample of some our initial achievements since launching in January 2024
Won the Jury award during pitch night for ProVeg's Kickstarting for Good nonprofit incubator.
Over 300 individuals submitted 170+ unique nonprofit ideas to Kickstarting for Good, with 9 selected for the program and 6 reaching pitch night; our nonprofit emerged number one, as judged by animal rights leaders based on potential animal impact, cost-effectiveness, team strength, and the neglectedness of our issue.
Attracted 75 volunteers within one week of launch.
40% ML & AI developers & 60% non-technical animal advocates. The quick and varied response to Open Paws reveals the animal rights movement's readiness to embrace AI, reflecting its competitive advantage in AI through a wealth of skilled talent and data resources, equipping it to effectively contend with more resource-rich adversaries.
Imagine an AI activist that never sleeps
This AI could be trained on the entire collective knowledge of the animal rights movement.
An AI trained on the collective knowledge of the animal rights movement would be extremely effective in crafting highly personalised and persuasive messages on animal issues.
An AI designed for animal advocacy could also be used to generate unique activist email and petition templates, improve donor communications, provide mentorship for vegan challenges through chatbots, automate replies to comments and emails, write blog articles, craft social media captions, predict the success of digital advertisements, assist political activists draft plans to shit funding from harmful industries to animal friendly ones and more.
This open-source AI could significantly increase the impact and cost effectiveness of the entire animal rights movement.
Potential Impact for Animals
Animal Charity Evaluators estimates 4,056 animals are saved per $1,000 donated to one of their recommended charities and 7 animals are saved per $1,000 donated to an animal shelter.
Farmed Animal Funders estimated $200 million was donated to farmed animal causes in 2021, whilst the Animal Agriculture Alliance estimated it was $800 million in 2022.
If we use the average between both sets of upper and lower bounds, that would be 2,031.5 animals saved per $1,000 donated and $500 million donated to animal causes yearly.
That means even if our AI only helped 10% of animal charities become 10% more effective, it would save an additional 10 million animals per year.
Based on these calculations, we would save 20-70 animal lives per dollar spent, making Open Paws approximately 5-17 times more cost-effective than the average ACE recommended charity
Sources:
This Estimate Is Conservative
Generic AI tools already improve employee productivity by 66% so 10% is a conservative estimate for how much more effective our AI would likely make animal rights organisations.
Likewise, 10% adoption by animal rights organisations is also a conservative estimate, considering the tool's free access, ease of adoption due to provided training and support, and the increasing familiarity with similar technologies.
This estimate focuses solely on the initial development and release of our open-source AI.
However, the broader impact will be more significant, encompassing the creation of additional tools based on our AI, the influence on major AI labs to adopt animal alignment, and enhanced awareness of speciesism in AI amongst the wider public.
Source:
Frequently Asked Questions
Shouldn't organisations first learn to use regular AI tools effectively?
Specialised AI tools for animal advocacy are crucial, as generic AI can lead to suboptimal or speciesist content, especially in automated and external applications like chatbots. Whilst training organisations to use existing AI tools for internal use cases has many short-term benefits (such as improving the efficiency of organisations) customised AI specifically designed for animal advocacy will have the greatest medium to long term impact.
How can a smaller nonprofit like Open Paws keep up with giants like OpenAI in AI development?
Focusing on specific areas like animal advocacy allows for the development of powerful AI models with smaller budgets and datasets, offering tailored solutions without needing vast resources. Currently and historically, specialised AI systems have tended to outperform their generic counterparts in specific tasks. However, our approach doesn't rely solely on creating superior AI tools for animal advocacy. We will also focus on influencing the development of AI more broadly and training non-technical animal advocates on how to use AI, both of which have enormous potential for impact, regardless of whether specialised or generic systems perform best in the future.
Why would animal advocates have any more success in lobbying AI labs than any other group?
Animal advocacy groups have several advantages in advocating for change in AI, like common connections within the Effective Altruism movement and AI industry. Focusing on reducing speciesism and harmfulness in AI models also aligns with standard AI safety and ethics concerns, especially as speciesist biases might also be risky for humans if AI applies the same logic towards us.
Can't we just use better prompts to get less speciesist responses from AI?
While prompt engineering can mitigate biases for internal use, the challenge intensifies in external or automated environments like public-facing chatbots or automated responses. Most users, including many animal advocates, lack experience in prompt engineering. An open-source, animal-aligned AI is the only way to systematically address these biases at their source, whilst prompt engineering is akin to placing a bandaid over a much deeper wound.
Further Information
Further details can be found in our Pitch Video and Funding Proposal.