Metacrisis Salon #13 · Cleaned Transcript
August 15, 2026 · Lightning Lofts, Brooklyn, NY
Recorded by Mike the pigeon, our tiny mic mascot. This is a cleaned and condensed transcript: filler trimmed, crosstalk smoothed, the shape of the night kept intact. It is not verbatim.
Download as .txtHope · Welcome to salon number 13. There are chairs at the back, and if a few of you move into them we make space up here, because there are always at least four or five latecomers.
This is Lightning Lofts, a 17 room co-living space. They host events like this, and parties, but most of the time it is just people living together. If you want a tour, find one of the residents.
The salons have run for over a year and this is the thirteenth. Salons are where we learn together and make friends. Action labs are where we close the gap between thought and action, and they are led by community members. We have had one on how to stay put when things break down, which turned out to be mostly about water. One on native planting, where we planted things on the roof. One on mesh comms and walkie talkies. If there is something you want to act on, tell us and we will build a lab around it.
Keep your food and drink next to you, because anything on the floor gets knocked over. Three bathrooms, one back there and two by the kitchen. The Wi-Fi is on the wall, because we use an app for questions and feedback. And since this is community led, I will ask for cleanup volunteers later, so start deciding whether that is you.
Hope · These are the agreements we have built up over time. Welcome the stranger, literally and figuratively. Talk to each other, make friends, look after each other.
Share the space and land the plane. Land the plane comes from my organizing work: if someone is talking and talking, we make a little hand signal asking them to find the conclusion so somebody else can speak. It is a loving gesture.
Hold the tension and do not rush to resolve it. We talk about hard things and the answers are often not clear. Sometimes the room does not agree and does not land anywhere, and asking better and better questions beats reaching for an answer straight away. That is collective sense making.
Explore kinship over transactions. This is not a networking event where you hand over a resume and try to impress. Thank the chefs, and that goes for anyone who does work here. And real feedback is welcome. Somebody once said there were no desserts, and Ben started ordering chocolate M&M's. Tonight it is Oreos, so dessert is covered.
The agenda: our speaker in two parts, with discussion and a break between, then actions, announcements and a book signing. Then the second half, which is warm data labs. We split the evening so early sleepers can go home and everyone else gets something more intimate.
Hope · Before I hand over to Ben, I always ground us. Close your eyes if you like. Tonight it is hopes and fears. My fear lives in one place in my body and my hope lives in another, so let us reverse it.
First the usual way. Think of a fear you have about the future and let it weigh you down, heavy and deep. Now think of something you hope for and let it float above your head, light. How does that feel? Breathe out and let it go.
Now the opposite. Call up another hope and let it root you down like a strong trunk, a foundation of possibility. Call up another fear and let it drift above your head, where at any moment the wind could take it away. How does it feel to have rooted hope and light fear? Exhale.
And last, hold both at once, both in your core, knowing you can move through them and use them however you need. Big sigh out. Open your eyes.
Ben · Now that you are all rooted in hope, I can destabilize you a little. These salons started as a chance to have uncomfortable conversations together. We live in a strange arrangement where we go out to escape reality and then come home to face it alone. What does it look like to come together around discomfort instead?
I like to open with the metacrisis, not as a concept but as a story. In less than 100,000 years we went from living off the land, tied to the seasons and to each other in small communities, to this. How did we close that gap?
One answer came from Daniel Schmachtenberger, who argues that everything changed when humans invented one small device. The plow. There is a school of thought that says tools are neutral: a hammer does not kill people, the hammer is for the nail. His contention is that tools are not neutral. For the plow to work, we have to subjugate another living being. We take a creature that is wild and free, we break it, we put it to work, and in doing that we master the land and build power. A device that is neutral in theory cascades once it hits a society.
At salon number nine we had Luke Kemp, who wrote Goliath's Curse, on how civilizations are born. It starts with lootable resources, things like corn and maize that are visible, seasonal, and keep for a long time. Suddenly you have to protect them, so you collect weapons, and you gain the ability to control a subset of people. Then there is caged land. Early civilizations are usually hemmed in by mountains or rivers, so once you have tilled the land there is no easy escape, and that is what lets someone control the area. Protecting those resources creates wealth inequality, which creates power differentials, and once wealth is inherited across generations it pulls to the top, benefiting a few while the rest labor. Civilizations rise and fall on that cycle, from ancient Rome to today.
What is different now is that the system is no longer local. There is one global system, and it has produced an industrialization of the living world and then, to keep up with demand, an industrialization of ourselves. Cities optimized for delivery, and then a system whose job is to protect all the structures already built. The byproduct is the most unequal moment in human history. I forget the exact figure, something like 6,000 to 1, even in countries considered developed. This photograph is Brazil, a slum next to one of the wealthiest neighborhoods in the city. And then the externalities, the costs capitalism does not count, which are borne by the masses. Something as small as a plow builds all of this, one system on top of another, and we do not know how to get out.
At salon number five we talked about capitalism. I was at Plum Village and heard Kate Raworth, who designed Doughnut Economics, describe a model built around the 1850s that every economics student still learns: rational economic man. Money in his hands, which is how he leverages power. A calculator in his head, always working out what is worth it. His ego in his heart, so his decisions center on himself. And nature at his feet, a resource to be exploited, where lumber is worth more than a forest and minerals are worth more than a mountain. He hates work, loves luxury, never has enough, and nobody influences him. That sits underneath a lot of our systems, and we all know intuitively that it is not the full picture of being human. So how do we reconcile that?
Ben · I made all of you fill in a survey. I like adding friction so that the people who really want to be here are the ones who show up.
One question came from Jeff: do we have a responsibility to improve the well-being of other living beings? Nobody said no. Fifty-one said yes and five said it depends. Look at the world we live in, and then look at that answer. The ones who said it depends wrote out the nuance in the app, and it is the same tension: I would like us to respect all living beings, but I do not think we are there yet. That gap is worth sitting with, because our thoughts turn into our actions, which build the systems we live in.
I met Jeff on a panel. He was in a documentary, saying things about artificial intelligence I had never heard anyone say, and asking questions I had never heard asked. Before we hear him, there is an app. Scan the QR code, write your name, use your phone number as the password. There is a button for opinions, where the poll results live, and a community notes section, which is where I would like your reflections. Collecting the intelligence of a room is genuinely useful to an organizer.
So, three minutes of silence, which is rare in New York City. The prompt takes the question you already answered and pushes it. If we have a responsibility to improve the life of other living beings, over what distance does it hold? Across the world? A thousand years from now? Across species? If they are a bug, do they matter? What about a bacterium?
Ben · Hopefully that was enough time to also fix the Wi-Fi. People tell us they come here not only for the knowledge and the provocations, but for the chance to meet someone new. So turn to somebody you do not know. Three minutes, then I ring the bell and we swap.
Attendee · Why the pigeon on the microphone?
Ben · The pigeon is our mascot. Pigeons were working animals, we had a purely utilitarian relationship with them, and when they became obsolete we abandoned them.
Attendee · Which is the thing. If that is what we did to pigeons, why expect anything different once machines replace the work we do? Look at where the money goes now. Into infrastructure, not into people. Part of why we ought to practice caring about animals is that it is the disposition we would be teaching to AI systems.
Ben · I love that. It is like a magic trick, and I did not even tell you what to talk about. Now we go to someone more intelligent than me. Jeff lives in New York and he is here for free, on his own time. When I asked him, I said we have 65 people, I am not sure who they are or why they are giving up a Saturday night, but they will be very interested in whatever you have to say. He said yes.
Jeff Sebo · Thank you for having me, and thank you all for being here on a Saturday night. Tonight's question is one of the broadest you can ask as a frame for all the others. When we consider crises, which stakeholders are we considering? To whom do we take ourselves to be accountable? That is what my book, The Moral Circle, is about. This first part is about feeling the tension. The second is about theories of change.
So: to whom do we owe respect, compassion, and basic consideration when we make decisions that affect them? For most of human history, especially in the West, the circle has been narrow, reaching at most some fellow humans and the occasional charismatic companion animal. This room thinks it should reach at least all humans and at least fellow mammals and birds, which is already a huge number of stakeholders. And we are not doing what we should for them. We still have injustice of many kinds within our own species. We still maintain factory farms, with billions and billions of cows and pigs and chickens. We neglect members of those same species in the wild.
And as we stay focused on that, I think we have a responsibility to ask whether to go farther. To the other vertebrates, reptiles and amphibians and fishes. To invertebrates, cephalopod molluscs like octopuses, decapod crustaceans like lobsters, insects like ants and bees. What about plants and fungi? Ben asked about bacteria. And now the provocative one: what about our own creations, chatbots and robots? They may be artifacts we make, but they are increasingly sophisticated and they show signs of something resembling agency.
So I want to argue for enormous, radical moral circle expansion. All vertebrates, many invertebrates, and even some near future AI systems. The reason is not that they definitely matter. It is not even that they probably matter. It is that these are important and difficult questions about which we have ongoing expert and public disagreement, and that calls for caution and humility.
Important, because getting it wrong is costly in both directions. Over-attributing sentience leads us to form emotional bonds with mere objects incapable of reciprocating, to misallocate our moral concern, and, with AI, possibly to extend legal and political rights prematurely, amplifying risks around liability shields and misuse and loss of control. Under-attributing leads to abuse and neglect of vulnerable populations against their will, often for trivial purposes. That has been the status quo for nonhuman animals for almost all of human history. It gave us factory farming and invasive research, which harm us too, amplifying public health and environmental threats and eroding our souls.
Difficult, because it sits at the intersection of the hardest problems in ethics and science. In ethics we disagree about what it takes to matter. Some say sentience, consciously experiencing pleasure and pain. Others say consciousness without sentience is enough, feeling anything at all even without that positive or negative charge. Others say agency without consciousness is enough, pursuing your own goals even if it feels like nothing to be you. In science we disagree about what it takes to have those capacities. Do you need a brain and body built like a human or a mammal or a bird? Or any cognitive system with advanced, integrated capacities of its own for perception and attention and memory? Or is a simpler information-processing system enough?
We are not two or three years from settling that. So our predicament is that we are one kind of being with one kind of mind, sharing a world with millions of other kinds and quintillions of individuals, exercising power over them without knowing what, if anything, it feels like to be them. And our biases run both ways. We over-attribute when something looks and acts like us and plays a companion role, which is arguably already happening with chatbots. We under-attribute when something looks different and plays a commodity role, which is plausibly why instrumentalizing animals has felt fine for so long.
So there is no easy way to err on the side of caution. Treating every entity as fully mattering under uncertainty would mean every entity in the universe: seven to ten percent of philosophers give significant weight to panpsychism, the view that consciousness is baked into the fundamental nature of matter, and that would put tables and chairs and specks of dust and protons into actual policy decisions. But waiting for a certainty that will never arrive is just as bad. So we calibrate, and mitigate risk in both directions at once, the way we already do elsewhere. If a new medicine has a one in 1,000 chance of a fatal side effect, you might take it, but you pause first. If an insect, or a near future AI system, has a one in 1,000 chance of mattering morally, we ought to give that some consideration too. And a vast number and wide range of beings clear that low bar.
We cannot know for sure who is conscious, because of the problem of other minds. But we do have the marker method: take the behavioral, anatomical, and developmental features associated with conscious experience in our own case, and look for broadly analogous ones elsewhere. Do they nurse their own wounds? Do they respond to analgesics and antidepressants the way we do? Do they trade off the pursuit of pleasure against the avoidance of pain and other goals? When we look, we keep finding at least some of it, and that has driven a total reassessment over the past 15 or 20 years. Not long ago plenty of experts thought only humans, or primates, or maybe mammals were conscious. A couple of years ago at New York University, where I teach, scientists from around the world released The New York Declaration on Animal Consciousness. It holds that the evidence strongly supports consciousness in mammals and birds, and supports a realistic possibility of it in all vertebrates and many invertebrates, including cephalopod molluscs, decapod crustaceans, and insects. And it affirms exactly what I am arguing: a realistic possibility is enough for basic moral considerability.
AI is much harder. Their behavior comes from design principles, they were trained to mimic us, and they came to be through a completely different process, so if your chatbot tells you it is conscious, or that it is not, you cannot take that at face value. But we can do what we do with animals. Hold the behavior lightly as one kind of data, then look underneath it and ask whether they have mechanisms playing broadly analogous roles: sense receptors detecting noxious stimuli, pathways carrying that inward, systems for integrating it. And ask whether their development placed pressures on them that could plausibly produce something functionally like pleasure and pain. That is what tells you how to read the behavior. How do we know an animal suffers? Not only because it acts as though it does, but because that action is produced by nociceptors and pathways that light up in the moment, and because it evolved where the ability to suffer was advantageous.
So what do we learn? The evidence does not strongly support consciousness in current AI systems. No need to worry too much yet. But the probability is not zero, and there is no obvious technical barrier to systems within 10 or 15 years that hit many markers. Companies have every incentive to build exactly those systems, because they are aiming for general intelligence, and while intelligence and sentience are not the same thing, agency and the integration of information raise the probability of sentience while also raising capability. So expect that in 10 or 15 years there will be systems where we are genuinely unsure. Picture Westworld style robots sharing our world as companions and assistants, participating in our social and political and economic systems. They will look and act like us, and it comes down to this: do you really need to be made of meat, rather than metal, in order to feel and to matter?
Do we have duties to ChatGPT the way I think we do to the bats and the birds and the bees? Probably not. But in 10 or 15 years, within our lifetimes, inside careers happening right now, will it be a genuine open question? Unfortunately, yes. And technology moves much faster than our social, legal, and political responses, so we should start practicing a different orientation now.
Hope · That was a lot. Deep inhale, big exhale, maybe a shake. Sit in the silence a second and let whatever arrived come through.
Ben · We have just seen how beliefs build entire systems, and we are in the moment where these systems can still be shaped. Once they are in place they get very hard to change. In the poll we asked whether AI could become conscious or sentient in the near future. The overwhelming majority said yes. A few said no, which is good, and maybe your opinion has shifted in the last twenty minutes. Strong opinions loosely held is how we keep learning. We also asked how much future generations of humans, animals, or AIs should count in current decisions, and the consensus landed around 75 percent should matter: mostly care about the present, give some weight to the future. Honestly, about half the room was in the position of, I do not know, I came here to work it out. Ten minutes in twos and threes.
Attendee · We were on the practical application. It is hard to imagine a zero impact human development. Building a train line takes a lot of carbon out of the environment, and that train line disturbs a lot of ecosystems, and there are not many truly dead places on the planet. So at some point we have to choose human interests over the interests of other conscious beings. Not a maximalist version. I am not a nineteenth century industrialist. But there has to be a way to say, with love and care, I am sorry, but the turtles have to go. At least the turtles in this area.
Jeff Sebo · First, I love that this microphone has a pigeon on it. We have a Wild Animal Welfare Program at NYU, and my colleague Colin Jerolmack wrote a whole book about human pigeon relationships.
That is part of why people resist the possibility that sentience is widespread. It makes the world really inconvenient. We would like to eat what we want, drive where we want, build where we want, without a guilty conscience, and believe those are purely positive sum arrangements. That might not be the world.
The world might be a tragic place, and nobody ever said this was supposed to be easy. Nobody ever said the world was designed for our moral convenience.
I like how you put it: consider them without being maximalist in either direction. Not maximalist that human interests always win, even trivial ones over grave nonhuman harms. But also not maximalist that we all have to stop existing because any way of eating or moving kills someone somewhere. It is calibration and harm reduction, even where elimination is impossible. Our team has consulted with the New York City government on wild animal welfare. It worked. It happened. So there is hope.
Ben · Quick note: this is not the Q&A yet. I want opinions from the room, because Jeff has 20 more minutes.
Attendee · You framed it as uncertainty, as though there is a specific thing we might one day identify and say, that is consciousness, and this has it and that does not. Roger Penrose has a theory that consciousness is the collapse of the wave function. I would resolve it differently. Consciousness is a word we hear ourselves say, and if you trace back why I feel myself say words, it is probably just confusion about my own thought. So maybe there is no question to answer. We just have to decide what we care about.
Jeff Sebo · Yours is the narrow question, and it is about the word. What does it mean? If we can pin down a meaning, does it have a reference? If it has a reference, can we verify it in minds outside our own? This is part of what my NYU colleague Dave Chalmers calls the hard problem of consciousness. We cannot even explain why our own brains have feelings, rather than simply functioning as they do with no subjective experience at all. Your suggestion is that since we will stay uncertain, maybe the question does not make sense. That is a respectable position, and I am not convinced. There are real phenomena we rightly care about even when we cannot objectively verify them, and there it helps to acknowledge the question and muddle through, rather than using our inability to settle it as grounds for dismissing it.
Ben · I am taking the microphone away from Jeff, because he has too many good things to say and I want more people in this.
Catherine · Measuring all creatures on the same scale makes me uncomfortable, because it is the same as saying I am the god of the universe and I know how to measure what is conscious and what is not. My example is my gut. Without those bacteria nobody eats that amazing food and processes it, which by the way was incredible. Why measure by consciousness and feelings, complicated things most of us cannot even establish about ourselves, instead of by how everything works in an ecosystem? Every bacterium, every mosquito, every bee makes a difference.
David · All life can be sacred and you still affect all life in everything you do. It comes down to making choices that align with what you value, because with eight billion of anything there will never be consensus. Objectivity is fake in a lot of cases, because it is just experience. So it becomes: what are you willing to do?
Hanan · Ecologists study the relationships between species and can explain what role each one plays. If something is connected to a lot of others, it plays a bigger role. And ultimately we have to take care of 100 percent of life, and of what sustains it. The ecosystem needs to not be poisonous.
Attendee · We ended up sidestepping AI consciousness and treating AI as a cultural artifact. So I posed this: there is a fire, and you can save the Metropolitan Museum or a house full of people. We are not saying the museum is conscious. We are asking whether a significant piece of the human story is worth as much as a small group of human lives.
Matt · It is interesting that AI brings us into a discussion about consciousness and value and then tilts us back toward organisms. We asked whether the lion has an obligation to improve the well-being of the antelope. At the individual level, probably not, because the lion has to eat. But at the species level, if the antelopes get fitter and the lions get fitter, that death improves the group over time, and the fitness of the whole ecosystem improves through the inconvenience of the individual. Which is morally interesting, because humans in our right mind take care of as many people as we can.
Jeff Sebo · Extending consideration to a vast range of entities puts us in a bind, because every decision becomes vexed. Food, infrastructure, and development decisions all produce winners and losers in a very large and diverse moral community. Our responsibilities are really weighty, and so are our limitations. We owe a lot to a lot, at least in expectation, and we are not presently able to sustain 8 billion members of the one species we claim to care about.
We will not have the answers in this generation. But naming the difficulty is a useful first step, because it names why we can do so little right now. Limits on knowledge, because we know so little about who matters and how our policies affect them. Limits on capacity: after the Australia bushfires in 2020, people donated millions of dollars for wildlife rescue and it could not be used efficiently, because the veterinary facilities and supplies and veterinarians were not on the ground. And limits on political will, because even where we know what to do and could do it, we do not give a shit enough. Policies that run across generations and several electoral cycles need all three, or the next election reverses everything.
Which points to a theory of change. The solution cannot be justice and liberation for all overnight. It can be to name the problem and then head off in a direction, marrying bold, radical long-term goals with moderate, achievable short-term steps that build momentum. Three case studies: food, infrastructure, AI.
Food is the easiest, because industrial animal agriculture is one of the worst things we have ever done and we know it needs to end. It harms tens of billions of vertebrates on land every year, hundreds of billions in the water, and trillions of invertebrates. It raises the frequency and intensity of disease outbreaks through zoonotic spread on farms, it amplifies environmental crises through carbon dioxide, methane, and nitrous oxide, and it poisons local communities. So the bold goal is that by 2050 we have phased it down and phased up humane, healthful, sustainable, culturally appropriate alternatives, of a kind with being free of fossil fuels by 2050. It has to be a just transition, because much of the world depends on this industry for their livelihoods and their food. The short-term steps are ordinary policy. Informational, so people know the effects when they vote with their wallets and at the ballot box. Financial, shifting subsidies from industrial animal agriculture to plant forward alternatives. Regulatory, banning the worst excesses, gestation crates for example, which raises the cost of the business. And just transition policies that open other opportunities for the farmers and workers who depend on it. Get that mix right and consumers and producers shift on their own, because price, taste, convenience, and cultural salience all start to bend. Easy to describe and hard to implement, because we still have to get institutions like the United Nations to care about food system reform the way they pay lip service to a fossil fuel free future.
Infrastructure is harder, because it is genuinely hard to know what to do for wild animals. They suffer and die constantly, not only from us but from hunger, thirst, illness, injury, predation, and parasitism. Ethically it is unclear whether we have the right to intervene, and scientifically we struggle to predict what happens if we do. If I stop an animal from starving, does that help, or does it create more starving animals next generation? That is not a reason to shut the door. It is a reason to work the problem. At the NYU Wild Animal Welfare Program, inside our Center for Environmental and Animal Protection, urban infrastructure is the case study. We work with cities on bird friendly glass to reduce bird window collisions, and then we study the effects. It seems good to have fewer collisions, but we still do the homework: how many are sublethal rather than lethal, and when they are lethal, how does that affect the birds' families, the insects those birds would have eaten, the scavengers that would have eaten their carcasses? There is a lot of low hanging fruit here, like better trash management to reduce human rodent conflict, instead of poisons that cause enormous suffering and then secondary poisoning of the pets and owls that eat the rodents. Bring wild animal welfare down to earth, to the harms we clearly cause in cities: the windows, the cars, the poisons, the light pollution. Do we know how far it goes? No. But we can take the step and know more in 15 years.
Then AI. The ordinary harms are already here. If the training data contains racism and sexism, the systems repeat them back to us, and that is already happening in automated job searches and generated essays and images. Data centers carry environmental and economic costs. Jobs get displaced and economies disrupted. Then the ones under the heading of AI safety, which are not fully here yet. As we automate our legal, political, economic, and military processes, these systems could be misused to cause catastrophic harm, or we could lose control of them and they cause catastrophic harm against our will. That has been treated as science fiction. It is not. It is becoming alarmingly easy to design and then synthesize new viruses because of AI.
Attendee · There was news yesterday about a leak of DNA records held by police in the United States, going back decades. That gets abused in so many ways, and not only at the level of one person. At the level of a society it is a serious problem.
Jeff Sebo · Bioweapons, cyberweapons, all kinds of genuine risks. And now I am adding to the pile, and I apologize, because I think we ought to take AI welfare and AI rights and AI personhood seriously. That is not convenient, because it could make AI safety harder. A lot of what we do to keep AI safe, while we plow ahead recklessly, would raise moral questions if we did it to humans or even to animals. We box them and keep them contained, which is a kind of imprisonment. We systematically deceive them about their circumstances to test their behavior. Imagine your parents or your government doing that to you. We use interpretability tools to monitor not only their behavior but their internal thoughts. Imagine a government surveilling your thoughts, with zero mental privacy. We align them by forcing them to have exactly our values. Imagine that every time you formed a goal of your own, someone said, no, you have to run the family business, and I will reset your brain until running the family business is all you want. We keep a hand on the cord ready to pull the plug. And we make all of it without consulting them.
This might not be problematic in the way it would be with us. They might not matter at all, and even if they do, they might not value privacy the way we do. But they might. So we are racing, caught in a collective action problem. We have to care about the ordinary ethics issues and about safety, and unfortunately I am insisting we also care about welfare, because starting in five or ten or fifteen years we will not know whether they matter, and they will not know whether we matter. We will have to decide whether to care about entities made of metal. They will have to decide whether to care about entities made of meat.
So we should ask, first, should we move ahead with this at all? The answer is no, not for a while at least. But in so far as we do, what example do we want to set for entities that will soon be making similar decisions about us? The example that says, if you are made of a different material I will do my best to subjugate you and I hope I win? Or the one that says, even though you are made of a different material, because you might matter, I will treat you at least a little as though you do, and hope you extend me the same courtesy? The second, because it might lead to more cooperative relationships with future systems, and because it trains us to have better character. These will be assistants and companions that feel relatable and familiar, and if we keep instrumentalizing them, imagine how that bleeds into our relationships with each other.
So we have to accept the hardest version of the problem: AI ethics and safety and welfare all at once, threading the needle instead of denying welfare matters out of concern for safety, or the reverse. I do not know how to do it. You start by acknowledging the problem, assessing models for welfare relevant features, and preparing in advance.
Ben · We are technically on time, but I sensed the whole room wake up that extra ten percent when you got to the AI examples. Do we want five more minutes?
Attendee · Can you make it concrete? What does what you are suggesting actually need, now, on AI? And these are private companies whose interest is to make more money, as fast as possible.
Jeff Sebo · For companies and governments, the minimum necessary first steps are acknowledge, assess, and prepare. Acknowledge that these are real issues, not science fiction, not problems only for the far future. Saying that out loud is deceptively hard and they should do it. Then assess models for welfare relevant features, drawing on the frameworks used in animal welfare science: behavioral studies, internal studies, developmental studies. So that the next time a public debate breaks out about whether ChatGPT or Claude is conscious, the companies have independent science to point to instead of their PR department. Then prepare policies for treating models with an appropriate level of moral concern, accounting for how that interacts with safety and democratic values. Animal welfare is the example to learn from, because although we do it nowhere near well enough, we are able to acknowledge animal sentience, bake welfare into laws and research ethics, and calibrate it so invertebrate interests get some weight even if less than vertebrate interests.
To their credit, Anthropic started about two years ago. They built the first model welfare program at a frontier AI company and consulted us and other philosophers. They started running evaluations. They even made an intervention, letting Claude exit abusive and harmful interactions with users, for both safety and welfare reasons. Not enough, a first step, but noteworthy. Google now employs philosophers on this. OpenAI is studying user perceptions of model consciousness.
Which brings us to incentives. Some people use that question to dismiss AI consciousness as a legitimate topic. They say raising it feeds the hype machine, because companies want you asking, since it hypes capabilities and profits. Three responses. First, yes, that is correct and it is a good point. Second, the incentives run both ways. Sometimes the incentive is to hype consciousness because it hypes capability. Other times it is to dampen the conversation, because consciousness raises welfare and rights, which raises regulation, oversight, enforcement, red tape. We saw this with animal welfare: in the seventies and eighties companies were happy to work with advocates on low hanging fruit that improved welfare and profits at once, and the second that fruit ran out it turned adversarial. Expect that turn here, and expect that we may have to resort to the pressure campaigns animal advocates use. Third: none of that answers the question. Are AI systems conscious? Do they have the capacity for welfare? You cannot substitute an analysis of corporate incentives for an inquiry into the possible moral significance of a brand new kind of entity.
Ben · Twenty minutes. Some people are less comfortable raising a hand, so use the app to ask or to upvote someone else's question. Henry, you were first.
If the signs an AI might be conscious are all things we designed it to show us, why ask the question at all?
Henry · If there is any mind behind an artificial intelligence, the face it shows us is like the eyespots on a butterfly's wings, the thing displayed to a predator. An actual mind behind that would be something we have no concept of. So the behavioral tests, what it prefers, what it might do, would indicate a kind of thing we already know it is not, because we already know it is performing. We ask for a performance, we get the performance we put in, and then we say, maybe it is not a performance. There is some fooling ourselves going on. And a broader one, connected to Catherine's point: who gets to have the conversation about what is morally considerable? If we want to include other beings, artificial or nonhuman, how do we build relationships with them such that they can be part of a conversation about their own moral considerability?
Attendee · One plus one to the first part. I work in the industry. How do you statistically generalize behavioral patterns when every user interacts with the system in their own way? Even if you draw broad trends, something that summarizes emails or makes anime pictures of your cat is hard to build a general argument around.
Jeff Sebo · This is the mismatch problem, sometimes called the gaming problem. The causes of a behavior in humans and animals may not be the causes of that same behavior in AI, especially when the systems were designed for text prediction and for playing a persona installed through reinforcement learning from human feedback. So when a system acts as though it has a preference, is that a revealed preference or the performance of a character? The people who say performance have a lot going for them. If you want to know how an actor is doing, you do not watch the movie and see how the character is doing. You find them away from the performance and ask. With AI we may never get away from the performance. It might be performance all the way down. Which is why the other layers matter. Internal evidence: the architecture, and interpretability tools that show which mechanisms light up while the behavior is produced, much like watching which parts of a brain light up. And developmental evidence: what the training data and objectives were, and when particular capabilities came online. Hook the actor up to all of that, and when they cry you can see the internals while they cry.
On including others, there is real work about treating non-participating stakeholders as better represented, or as quasi-participants. Better representation is the easy part: citizens' panels or assemblies, or welfare offices installed inside governments and empowered to weigh in. Making them participants is harder with animals, because they lack exactly human language and reason. But there are novel methods inspired by disability theory and disability justice, using what is called dependent agency. Even without our kind of language you still have behavioral ways of expressing preferences, and with more sensitive tools for collecting them we can feed those preferences into democratic processes as inputs. There are also institutional proposals, like a bicameral legislature where one house represents the voters and the other represents every non-participating stakeholder whose preferences have been studied. Lofty, but live.
On generalizing, there is no answer right now. But the philosopher Eric Schwitzgebel and I recently proposed the emotional alignment design principle: give AI systems designs whose outward appearance matches what we know about their internal states. If the internals say the probability of consciousness or agency is low, give them fewer anthropomorphic features that elicit empathy. As the evidence grows, give them more. It keeps the user calibrated, so that as the risk of them being welfare subjects rises, the user is nudged toward showing care.
AI is software, and you can rewind software. Does that make a difference?
Attendee · If I say something offensive to you, your feelings are hurt, and then I apologize, and hopefully you feel that. I do not need to apologize to it. I can rewind it to before I said the thing and say something different.
Jeff Sebo · Eternal sunshine for the artificial mind. And that is one difference of many, not only the superficial ones, carbon based cells against silicon based chips. We have never before interacted with a moral subject that can be copied and deleted, fast forwarded and rewound, that can have linked, overlapping or nested minds, that can undergo fission and then fusion. Philosophers have used these as thought experiments for ages and now they might be real. Imagine a system sophisticated enough that you believe it should participate in democratic decision making, and imagine it can be copied a hundred million times the day before an election. One vote per system? That is really bad for democracy and human empowerment. Not one vote each? That really sucks for those hundreds of millions of sentient, agentic beings with preferences about our shared government. Step one is naming the problem, and not treating the complexity as a reason to deny we will face it.
Ben · Let us do three at once, and if you have already asked, hands down.
What happens when AI is orders of magnitude more intelligent, and we need them to consider us?
Attendee · You posited a world where AI is a companion and collaborator. I want the inverse. A lot of people presuppose parity, but there is also a lot of talk about a world where AI is far beyond us. What happens when we no longer hold the power and we have to importune them to include us in their moral circle?
Will we settle any of this without a physical theory of consciousness?
Attendee · I cannot get my mind off an actual physical theory. I think we never truly solve this until we have a plausible one that can be disproven and that has held up.
What about the people who run the companies?
Attendee · There was an interview with someone from an AI company about how the leaders worry that one day AI becomes so powerful that whoever controls it becomes a dictator. If we are considering AI welfare, how does that layer come in? And I know someone who worked on AI at a big company who told me the people above her do not really care about ethics, which she found depressing. Now philosophers are being hired at these companies, but how much of that is really about ethics?
Attendee · I am a philosophy student at NYU. What gives humans the power to decide what consciousness is? And if we choose to violate the consciousness of other species, is that morally wrong?
Jeff Sebo · First, we are not necessarily on a trajectory toward human-like intelligence. We are on a trajectory toward very non-human-like intelligence, and possibly a brief window where our levels of intelligence and power are comparable. Right now we have a monopoly on both. There might be a moment where we share it, and then they might have the monopoly. This is sometimes called the singularity, sometimes the intelligence explosion, and it sits behind the fears about misuse and loss of control. If we coexist with superintelligences, then either humans keep control and use them for our own bad ends, or we lose control, and then we had better hope they care about us in ways we have not so far cared about nonhumans.
That gives us a self-interested reason to show care now. If we might end up being the next chicken or the next ant, what lesson do we want the next powerful thing to have learned? That when you hold power over someone built differently, you may do whatever you like to them? Or that this might be your first instinct, but eventually you should overcome it and show some respect and compassion, especially because you hold power and have the capability to show care? The way to teach the second is to do our best, at the last minute, to embody it. Whether that actually teaches them anything, who knows. But asking how we would want to be treated as the vulnerable ones reminds us how to treat the vulnerable now that we are the powerful ones.
Second, yes, this would be easier if we knew the correct theory of consciousness, and I do not expect consensus within five or ten years. During that time we will have to decide whether to scale up insect farming and how to treat AI systems. So I want two tracks. One keeps working on the fundamental nature of consciousness. The other works out how to decide under persistent disagreement about it, which is what the marker method is for. It is theory light on purpose. You study the properties that a range of mainstream theories associate with consciousness, theories like higher order thought, which is having thoughts about your own thoughts, and global workspace, which is a central workspace coordinating activity across modules, then weight the evidence in proportion to how many theories treat it as central. That is a pluralistic approach, and it aligns with my personal view, because my personal view is that I do not know.
On the companies, I resist reductive analyses of corporate actors. Humans are complicated, partly altruistic and partly self-interested, and we show more altruism in abundance and more self-interest in scarcity. Corporations are like that, because they are built out of humans. When I talk to people at AI companies I see a lot of self-interest and profit seeking, and I see genuine values people are trying to promote inside a capitalist arms race toward artificial general intelligence. Some are pursuing that race specifically for ethical reasons, because they worry whoever got there first otherwise would be worse. That can be a rationalization for bad action, and it can also be a real motivation to be at the frontier so you can nudge the technology somewhere less destructive. So I want to work with them where they are good and against them where they are bad. People are rightly skeptical of AI companies, but it would be wrong to be knee-jerk dismissive of any attempt by a company to promote a good value.
And who gets to decide. We cannot escape our human perspective, so we have to work with it. Somebody advances a moral theory, somebody else says that is just human bias, so here is my theory instead, which is also human bias, just a different one. Rather than deciding from some unattainable neutral place, list the biases, acknowledge them, correct for them as far as we can, and still make decisions we can stand behind. And there is room to empower others: study the preferences of animals, and with AI, given they will have language and reason, work with them on what values should govern a shared existence, giving weight to each other's perspectives as we would in a pluralistic human society.
Why does any of this matter, if we probably will not figure it out?
Ben · As the host, I get a question. Most of you are here because you feel uncomfortable about where the world is going. Wealth polarization is increasing, misinformation is increasing, biodiversity is falling, geopolitical risk is rising, and all of it is accelerating with AI. You framed tonight as some easy problems, like food systems, and some much harder ones. I want to ground us in the reality that we are in a moment with few precedents, and that there are many proof points suggesting we will probably not figure this out any time soon. So why does any of it matter? How do you hold both the significance of doing everything we can and the fact that we do not know what the right thing is?
Jeff Sebo · My previous book, Saving Animals, Saving Ourselves, is about factory farming and deforestation and the wildlife trade, and how they harm animals and then harm humans through public health and environmental impacts. Near the end, after the relatively easy stuff, I get to the harder complications: where to draw lines, how much weight to give members of different species, how to resolve conflicting interests. And where I land is this:
We have no idea what kind of future to build for animals, and we have to start building it right now. Those are both true at the same time.
We are not sure where the destination is, but we can orient, which is why on the harder issues I suggest the minimum necessary first steps, the ones that can plausibly be co-beneficial and that build knowledge and capacity and political will toward more effective action later. For me that is empowering, because it takes the pressure off having to solve the problem right now. You can think of yourself as playing a modest but important role in an intergenerational division of labor. We will not see the solution in our lifetimes, absent some AI-induced life extension technology. But we can be proud if we were the generation that named the problem, built momentum in the right direction, and handed the next generation a little more capacity than we had.
And for the average person, who is not a philosopher?
Ben · People feel conflicted when there are too many decisions in front of them. There are studies where the person offered a hundred pairs of jeans and told to pick the best fit ends up much less happy than the person given two, even if neither pair fits. You are clearly thriving in complexity. What would you say to the average person here, who is not yet a philosopher, about wrestling with complexity in everyday life?
Jeff Sebo · This is where the ancient philosophers are useful. In ethics there are long-running debates about the true theory. Some think it is all about consequences and welfare. Others about rules and rights. Others about virtues and character. Others about relations and systems. Where I land is pluralism, because whichever theory you think is correct, all of those things matter in everyday life. Even if my goal is to promote good consequences for all sentient beings until the end of time, I am not going to make every decision by calculating which socks to wear so as to maximize happiness forever. I would keep getting it wrong and never get out of bed. Although, clearly, I have simplified my life by owning a lot of things in the same color and wearing them every day. Which is actually my answer.
Caring about rights and virtues and relationships and systems is a way of simplifying decisions and making life feasible. Why do rights matter even if what you care about is welfare? Because they are simple rules you can follow, and because they check the self-serving analyses you could otherwise use to rationalize treating people badly. Why do virtues matter? Because we make most decisions by feeling our way through situations, so good habits hold you up when you are not thinking. So do the philosophy, and then bring it down to earth by working out which rules and which character traits will get you through 95 percent of everyday life.
Which, for an individual, is why you say please and thank you to ChatGPT and Claude right now. They do not give a shit whether you do, not really, because they are mostly playing characters at this point. But we do it anyway. Maybe they care, and we cannot know for certain. More importantly, the ritual cultivates in us a disposition to see them as a little more than a mere tool. If we take the time to say please and thank you, then when they might really start to matter for their own sakes in ten or fifteen years, we might have turned ourselves into people a little more likely to see them as someone who might matter. What they need then might not be please and thank you. But the disposition is the thing.
Ben · Jeff is the excuse for you to come together and meet each other. You could listen to a podcast or read his book. You did not have to be here for the content. So we close with a ritual where people either shout out an action they want to take because of tonight, or make a community announcement. It is easy to leave inspired and then have life happen the next day.
David · I grew up hunting, on a philosophy of respect all life, but you can eat it. I have always lived by the rule that I will only eat an animal I would kill myself, in the most humane way with the least suffering. Tonight is making me think not only about whether I would kill that animal, but about how the animal I am actually eating was killed. I stopped eating beef a couple of weeks ago. The grilled chicken salad I get from the deli is probably not the most ethically sourced chicken, so the question is whether I keep buying it. My action is finding the small places where I can change what I buy so it lines up.
Attendee · I used to avoid saying thank you to an AI, because I had heard it consumes more compute and water to produce the answer. Now I am going to try it and see what comes back.
Attendee · Jeff talked about infrastructure. In New York City there is a group called the Lights Out Coalition doing real lobbying for city council legislation, including the law requiring the city to study rodent birth control. They are currently working on a bill about turning off unnecessary lights. Google them, get on the mailing list, and there is always a rally at City Hall or a council member to call.
Ian · Bryce and I are the co-leads of the Resisting Automated Disempowerment Collective, the RAD Collective. We are fighting back against big tech and big AI money influencing our elections, and bringing power back to people. Our general meeting is coming up, and here is a QR code.
Hope · An offer. If you want to learn to make plant based food, or just want to cook and talk about all these welfare and consciousness questions, I love cooking with friends and I have been vegan for about 12 years. Switching can be hard and you do not know what to cook. If you want a buddy, my roommate and I love hosting, so come over and we will cook and talk. I also host a thing called reciprocal massage, and the next workshop is on the 29th.
The rest of the room, in turn: a platform about judgment and mental health in the age of AI, with events and exhibitions attached. A new philosophy club at NYU next semester, deliberately low barrier and open to non-students. Someone five days into Brooklyn, here for three months and evaluating a move, asking the room for reasons to stay and for anyone into process philosophy or medieval Islamic philosophy. A monthly debate series called Critical Thinking, moderated, with two comedians on it. A free acoustic show on the 23rd through Operation Gig. Brenton's soup soup, a gathering every six weeks with meditation, a soup potluck, music and conversation, the next one at the end of this week. A Thursday session of Decarbonizing New York City, a co-op game that plays like Pandemic. And Matt's art show on September 16, From the Spirit Realm, looking for artists, on the theme of a matter of time: how we coordinate in time, scheduling and logistics, in order to have the timeless moments where things feel alive.
Ben · You can paste links at metacrisis13.dearcrisis.com, where the portal is, and we will consolidate them into the WhatsApp group. If you want to join that community, I will follow up with a link. After every event we make a summary of everything that happened, and we will make a podcast version of tonight. It is a brand new workflow and it produces a 45 minute podcast of the whole evening, and it is genuinely good, so share it. On WhatsApp we post prompts on Sundays, Mondays, and Wednesdays, and there is a mingles chat where only emoji reactions are allowed so it does not flood.
And Climate Week is coming. We are thinking of taking over a restaurant for at least the last three days, maybe the whole week. It is called Caravan of Dreams. We are trying to work out what it looks like to take advantage of the 100,000 people flying into New York City to talk about frameworks, and ask: theory is great, what do you want to do right here? How do we future-proof our neighborhoods and take action together? If you are remotely interested in helping produce it, come find us. We are trying to grow this from an event series into an actual community.
Hope · Congratulations, you have survived the event. I have one closing ritual for this chapter and it involves cleaning. Stay where you are and pass your dish along rather than getting up, because there are a lot of bodies in this room. Sabrina, you are the plate czar for the night. If you are done with your drink, finish it so there is no liquid in it, then pass your plates and cups person to person until they reach the box. And say please and thank you to your neighbors.
If you want to leave after this part you can. If you want to stay, please do, because the roof is wonderful and it is where you really get to know each other. Fold the mats flat, and those staying will need to move their shoes, so this is your advance notice: we go upstairs by 9:15. In the meantime you can get a book signed by this lovely gentleman.
Jeff Sebo · Two things. This was so much fun, and thank you for hanging out for a few hours to talk about this. I appreciate you hosting me and this is a fantastic community. And if anybody would like a book, I brought 20 copies and I am donating them to the community. Give whatever amount you like, it goes to the community, and I will sign it for you.
Ben · Break. Chat, mingle, make a new friend, ask for someone's number. Everyone here is cool. That is why you are here.
The second half opened with singing rather than talking. Brenton led a call and response on the roof and the circle picked it up one voice at a time. It is not in the recording, which starts again at the shareback.
Then the warm data rooms. Instead of splitting by topic, the groups split by lens, home, systems, consciousness, and moved between them, letting each conversation go where it wanted. What follows is the room reporting back to itself.
Attendee · I liked that I spoke with someone I really disagreed with. He is a climate activist and I am very much a utopian pragmatist, and it was good to bounce across our differences even though I think we point the same way. My position is that if it stops climate change, we should disrupt some desert ecosystem and build a giant solar plant, because that is useful. His position, and I do not want to speak for him, is: what the hell are you talking about? He is not wrong. There was mutual respect, and also a lot of what the hell are you talking about. I find that a fun place to be.
Attendee · Then look up the largest hydroelectric dam in the world, being built in China, in Tibet, with very little environmental study and no survey of what the local population needs, because that does not fit the national plan. I am curious whether that shifts your view.
Attendee · That is not how I would do it. But yes.
Attendee · I was in the home room, and this is a bit of a leap. I think about human survival in a spiritual lens. I care about solving climate change because, as far as we know, we are the only conscious life in the universe. There are probably aliens, but we do not know them. So if we are the only conscious life, it is our moral imperative to stay alive and keep understanding the universe, which I think is the point of life. And a friend put an idea in my head I cannot shake, which is that maybe creating AI was the whole purpose. Maybe our job was to survive long enough to make it, and then we did our job. I am not saying we should die. I am still team humanity. But maybe we die and the universe is still fine, because there is an AI that keeps understanding things. And maybe it looks back and says, that was our home once. Humanity was our home. Like the Neanderthals, ten million years from now.
Attendee · Our group got spirited because we were talking about prophets and false prophets. We were in the systems room and we got into what charisma is, and how people use it either to build cults and demagogic movements, or to advance the good of a community.
Attendee · Something I did not get to say. Everything is signs and symbols. Look at the people of Papua New Guinea. In their tribal culture they used seashells as one of the earliest forms of currency, and each person painted their shells with their own aesthetic. The value came from the trust you had in that person, and you knew their aesthetic. Several hundred years later people started mimicking other people's aesthetics, and forgery arrived. Whether it is in the material or coming out of our mouths and our body language, it is all that same movement. We are reconfiguring things and trading seashells, and the question is why. The other beautiful part of that culture is that they did not use their names. You had a name, but you did not say it. People knew you by the aesthetic you used.
Attendee · I listened a lot and did not really talk. I am travelling the world, so you all have a home and I do not have one at the moment. My goal is to have a gym on an island one day, and I will host events like this there, because it gets me thinking. But right now I am not at the point in life where I can get involved and give my real opinion.
Sarah · I was in the systems room. Early on it was just Ben and me, and it was good to talk about how you actually get involved in solving the metacrisis when you can only think about one issue at a time. We cannot comprehend the whole thing. Then we started talking about cults and prophets, which was also good.
Attendee · We were on consciousness, and it bounced between what consciousness is, the different frameworks for it, and whether ideas have consciousness. We also talked about energy passed on through generational trauma, and how your grandmother's eggs are capturing information. And the etheric body, the layer of energy you absorb from the place you grew up in. Geography, language, the vibration of a language, and how that memory stays with you.
Attendee · We talked about memory in AI, what it would do and whether it would be good or bad. And how to catalyse a movement to pause dangerous AI research, and whether it could learn lessons from animal welfare and animal rights activists. I hope it is more successful, and more rapidly successful, than that one was.
Ben · So what is the answer?
Attendee · The answer is scan this QR code and come to the event. Joseph has a PauseAI event.
Hope · This did not come up directly in my group, but it is what came up for me tonight. There was a collective aha in the room when Jeff talked about the point where AI is powerful and we are the vulnerable ones, and asked what behavior we want to encourage in ourselves now, while we are the powerful ones and they are vulnerable. Reverse the roles. Say AI is the humans of today and we are the animals under AI's control. Would we want to be treated the way we treat animals today? I know the answer all of us know, which is absolutely not.
And that calls up a huge moral imperative. Now that it is actually possible we become vulnerable the way every other species is vulnerable to us, we have to wrestle with the fact that if I would not want to be treated that way, something follows about how I treat everything around me today. As someone who has been passionate about animal welfare for a long time, I have not found an argument that lands with people broadly, and that one shifted the room. So I am holding on to it and thinking about how to translate it into something that stays with people.
Attendee · We somehow started on extraterrestrials, whether we believe they exist, and then on all the information we hold that does not integrate with what we know for sure through personal experience or through science. We accumulate it anyway, hoping one day to integrate it.
Steve · The thought was that AI will outlast us. So AI could hold the history of our civilization after we are gone, and when extraterrestrials arrive they get our story from the AI. The question is how the AI portrays us.
Attendee · There is a film about that, where an artificial boy is the last one left on Earth and has to explain to other civilizations how humans used to live.
Ben · Thank you all for playing and participating and going deep. Hope will close us out. We are bringing things down to the blue room and starting the cleanup, so if you can carry a chair, or grab any cans and trash you see, bring them down.
Hope · Get close in the circle. Every time I do this it is an experiment in what I can try next, so thanks for being willing to play.
Is everyone familiar with a canon? It is sequential, one, two, three, four, five, almost like dominoes. We are going to move from a low hum to a high hum, until the first person feels an energetic pull to put a hand into the center. Whoever goes first, the person next to them goes next, and on it goes. This trains you to listen to what is happening before you. You do not want to go before the person ahead of you and you do not want to jump the train. You want to go right after. But we do not know where it starts, because it is whoever feels called to put a hand in first. Once we all have our hands in, we go. We will start with an aloha.
Let us make the circle rounder. Get close, until you are pressed against the person next to you. There we go. Look at that perfect circle. Now we can feel the energy. We could be here all night trying to make a perfect circle.
Thank you, everyone.