Jacy Reese Anthis is a polymath and rising star in the fields of Effective Altruism (EA) and artificial intelligence (AI). He is a co-founder of the Sentience Institute and the author of the 2018 book “The End of Animal Farming”. EA is the project of identifying the most impactful strategies to help others, and the book grew out of EA research into the best ways to help animals. He explains his thinking on animals and AI in the latest episode of The London Futurist Podcast.
Many people think that factory farming is one of the most pressing ethical problems in today’s world. A group of people in EA started running psychology experiments and looking at historical case studies, to see how we might move away from factory farming, and the book builds on that work to provide an optimistic roadmap towards an animal-free food system.
Extending moral consideration
The Sentience Institute is concerned with extending moral considerations to non-human entities, including animals and also AIs. It may not be many years before digital minds become sentient, have emotions, and experience suffering. Jacy believes the Institute is the first AI rights organization in the world.
What is sentience, and how does it differ from consciousness? One way to distinguish them is to say that sentience is the ability to have sense impressions – sight, sound, touch, etc, and to experience pleasure and pain. Consciousness includes this but adds a sense of self, that you exist as an entity distinct from the rest of the world. Anthis called his organisation the Sentience Institute because he thinks that having sentience, specifically the capacity for suffering and happiness, is the primary criterion for moral inclusion.
Morality or economics?
What will get humans to stop eating animals? Will it be moral considerations, or economics? Many people think that sometime in the next two or three decades, plant-based meat substitutes and then lab-grown meat will become cheaper than meat grown on animals, as well as virtually indistinguishable to the consumer. In which case, presumably most people will switch.
Vegetarianism has been championed for decades by organisations like People for the Ethical Treatment of Animals (PETA), the Humane Society of the United States, and even by fictional characters like Lisa in The Simpsons. Most people have seen videos exposing the appalling conditions in some factory farms, although many farmers would very reasonably protest that they treat their animals well. Most people are also at least vaguely aware of arguments that farming has environmental impacts.
One thing that is different now is food technology. Impossible Foods and Beyond Meat, the two big players in the plant-based meat industry, are getting people used to the idea that they don’t have to have animal meat as the centre of a “proper” meal. This paves the way for cell-cultured meat in the future, which can be indistinguishable, and perhaps even molecularly identical, to animal flesh.
Timeline to global vegetarianism
Anthis is optimistic that animal farming will end by 2100, and to this end he divides the century into four phases. From 2000 to 2025, the main activity is resource accumulation, particularly research and development. From 2025 to 2050, he thinks we will see rapid changes as economies of scale make meat substitute products cheaper, which encourages more people to convert, so you get a virtuous spiral. One consequence of this will be increasing cognitive dissonance for meat eaters as they think more seriously about the potential cruelty of eating animals because it no longer seems unavoidable.
Between 2050 to 2075, he thinks that late adopters will catch up with the early majority due to stigmatization and social momentum. People in wealthy countries will have to go out of their way to obtain animal-harvested food. From 2075 to 2100, the industry will spread all around the world. While this timeline may seem radically short, Anthis says it will only be this slow because the existing trillion-dollar food industry infrastructure will be busy adjusting to the new reality.
He is in part optimistic because of the ability of artificial general intelligence and then superintelligence to accelerate this timeline. Overall, Anthis has a median estimate that animals will provide less than 10% of human food by 2068.
Another external factor which might accelerate the progress is fears about climate change. This is already applying moral pressure to many meat eaters, and it may cause some governments to incentivise a switch away from animal husbandry. That said, Anthis points out that there are few historical examples of this sort of directed change.
Jim Mellon’s book “Moo’s Law” argues that lab-grown meat will soon start to replace plant-based meat as the most cost-effective substitute for animal flesh, and that within 25 years it will destroy the current agricultural business model by sheer force of economics. There will be a lag while the infrastructure adjusts, but that shouldn’t delay the change much.
In which case, what will we do with all the land freed up? Something like 70% of the UK’s land is used to raise animals, or grow animal feed. Will we just re-forest it all, or will there be a huge growth in house building – something we are very bad at doing today?
Anthis cautions that there are sceptics about the case for lab-grown meat. It may be very difficult to reproduce the immune system of animals in an in vitro setting to keep the meat safe to consume. Others question whether it will really be possible to get the cost of the component elements of lab-grown meat down to a level where it out-competes the existing process. Of course, all future technologies involve this sort of uncertainty; if we knew for sure how it would be possible, then we would probably already be there.
Most people would agree that animals are sentient. Will most people accept that AIs will be sentient? And indeed, should we be in the business of creating sentient digital minds at all? We might inadvertently create entities capable of profound suffering, and thus make the world a far less happy place.
Sentient AIs and sneaky AIs
Thomas Metzinger is a moral philosopher who thinks there should be a moratorium of the creation of sentient AIs. But there are huge practical challenges to achieving this, and most people who have looked into it agree that “relinquishment” is not a realistic option. Anthis thinks that sentience is likely to arise as a by-product of advanced intelligence, and that this process will teach us valuable lessons about our own sentience and consciousness.
People have already started to attribute sentience to AIs. Google recently fired an engineer, Blake Lemoine, when he persisted in making public statements that one of the company’s AI systems was conscious. In the future, AIs will probably be deliberately trained to fool people into thinking they are conscious, but Anthis says that we should also be on the lookout for AIs that insist they are conscious despite being trained to say they are not. He also thinks it is relatively easy to tell that today’s AIs lack general intelligence with simple questions – such as, “What fruit would you use to hold open a book while reading it?”
To finish, a trio of fun facts about Anthis. First, he practises “polyphasic sleep”, which means he sleeps three hours a night and then takes three 20-minute naps during the day. He seems remarkably healthy nevertheless. Second, he listens to audio books at six times normal speed. He thinks he catches and retains what he needs to. And third, he proposed to his wife in the acknowledgement section of his book.