1. Podcasts
  2. The Interface
  3. How will we know if AI becomes conscious?
The Interface

18 Nov 2026 · 31m · 14 chapters

Episode artwork

How will we know if AI becomes conscious?

In short

Whether AI could become conscious, how to tell if it is, and what moral treatment might be owed if it is.

Guests

Jeff Sebo, director of NYU’s Center for Mind, Ethics, and Policy and a philosophy professor; his work focuses on animal sentience/welfare and adapting those methods to AI.

Key claims

Consciousness = capacity for subjective experience/feelings; sentience = positively or negatively valenced subjective experience (pleasure, pain, suffering), which grounds moral concern. Behavioral claims from chatbots aren’t enough; evidence should combine behavior with internal mechanisms and developmental/training pathways, using a pluralistic approach under uncertainty about theories (biological naturalism vs computational functionalism).

Notable examples

parallels to animal evidence (nociceptors, analgesic responses, evolutionary pressures); “meat vs silicon” debate; Kantian point that how we treat non-humans shapes our character; listener concern about cruelty and human impact.

Written by AI. May contain mistakes.

Chapters

0:54

Introducing the Philosopher

The hosts introduce guest Jeff Sebo and discuss his background in philosophy.

“You're pretty good at weird conversations.”
2:11

Philosophical Journey to AI

Jeff shares his journey from studying animal ethics to AI consciousness.

“This is a subject that really riled up our audience base when we talked about on the show a couple of weeks ago.”
4:52

Understanding Consciousness vs. Sentience

The guests explain the difference between consciousness and sentience.

“So one thing that comes up a lot in these conversations is what it feels like to be someone or something, right?”
6:57

Evaluating Evidence of AI Consciousness

A discussion on how to evaluate if AI systems are conscious or sentient.

“And then we asked them to give specific answers to questions about consciousness.”
9:21

Challenges in Assessing AI Consciousness

Exploring the complexities of determining consciousness in AI.

“So the question is then how to adapt that for AIs.”
12:34

Debating Theories of Consciousness

Discussion on competing theories regarding consciousness and AI.

“Like some people say like, there's this theory, I think it's called like, computational functionalism.”
14:00

Understanding AI Consciousness and Evidence

Learn about the various theories and evidence used to assess AI consciousness.

“and self-awareness and social awareness and language and reason and integration of information and flexible, adaptive decision making, then it could be made out of meat or it could be made out of metal.”
18:34

Human Exceptionalism and AI

Examine the human tendency towards exceptionalism and its implications for AI.

“Everyone's talking about AI, but to understand where it's going, you need to understand where it came from.”
22:06

Balancing Human Issues and AI Welfare

Discuss the intersection of human problems and the need to consider AI welfare.

“A little bit of a follow-up here in that in some of your work, you say that some scholars suggest that respecting the interests of AI can demonize human beings.”
24:20

Ethics of AI Treatment and Human Impact

Explore the ethical implications of how we treat AI and its effect on humanity.

“future ones at the same time, instead of approaching it sequentially in a way that ends up being worse for everybody because we wait too long to address the future issues.”
Show all 14 chapters
28:00

Philosophical Insights on AI and Kindness

Explore Immanuel Kant's views on moral significance and how they relate to our treatment of AI.

“And this is an insight that goes back hundreds of years.”
29:26

The Implications of AI Consciousness

Discuss the challenges and questions surrounding AI consciousness and its impact on daily life.

“So I am curious, though, in talking about, again, like what happens when this kind of, if AI becomes conscious and there's all of these implications on us to consider them, what happens to the average person's life?”
30:24

Cultivating Humility and Curiosity About AI

Learn about the importance of humility and curiosity in approaching AI's future.

“So I think that we have to ask these questions holistically.”
31:52

Closing Thoughts with Dr. Jeff Sebo

Recap the insights shared by Dr. Jeff Sebo on AI and its implications.

“The advice here is cultivate confusion and uncertainty.”
A man walking through an airport, asking his phone a question

Ask about this episode

Get the gist, explore an idea or ask about something said.

Claude

On The Interface, what are the key points of this episode?

Searching your listening

Transcript

Automatic transcript. May contain errors.

0:00This BBC podcast is supported by ads outside the UK.

0:54Thomas Germain: I think it's fair to say that we are about to have the weirdest conversation that we have ever had on the interface. I don't know, Tom. You're pretty good at weird conversations. I like to keep it weird. That is kind of my thing. But it's heady enough that we have brought on a philosopher to help us sort through all of this. And the question we're going to be tackling is whether AI could ever be conscious. And if it was, how would we even be able to tell? And if the answer is yes, what would we have to change about our world and the way we live our day-to-day lives in order to treat this new conscious entity or entities with some kind of respect.

1:37Thomas Germain: Strange stuff. Strange stuff, indeed. I'm Jess Maddox. And I'm Thomas Germain. And our guest today is Jeff Sebo. He's the director of the Center for Mind, Ethics, and Policy at New York University. And also a philosophy professor. You're a philosopher, Jeff. It's not every day. I don't think most people get to talk to philosophers that often. It's kind of fun, right? I think so, which is why I am a philosopher. Most people are thinking like ancient Greece, but no, you're a current, this is your job. You philosophize. Yeah, we still have philosophers, at least for now. And I am very happy about that.

2:08Well, thank you so much for joining us today, Jeff. This is a subject that really riled up our audience base when we talked about on the show a couple of weeks ago. And before we get started, I'm a researcher myself. And so I'm dying to know your path. How did you get here? How did you become a professor studying this field? And what's the burning question that drives your work? I have a friend who's actually an NYU PhD who always says research is therapy, right? That we're all trying to like work something out about the world. So how did you get here? And yeah, is this research your therapy? Yes, I would say it is my therapy.

2:46I got here through philosophy and then through questions about animal sentience and welfare. So I started studying philosophy in college and I fell in love with the metaphysical questions about the nature of reality and the epistemological questions about the nature of knowledge and the ethical questions about the nature of value and what we owe each other. And it was around that same time that I became more interested in animal sentience, animal welfare, animal ethics, because my college professors and then eventually other people as well really impressed upon me how much evidence there is that our species has overlooked of consciousness, sentience, agency, moral significance in non-human animals, evidence that they can have feelings and emotions and desires and preferences and that they matter for their own sakes.

3:33And yet we have been treating them quite badly for food and for research and for entertainment and in the wild. And so I developed a specialization in trying to understand how to interpret evidence for these morally significant capacities in non-human animals and then reckon with the ethical and legal and political questions about what do we owe other animals and what follows for our use of them for food and research and entertainment and in the wild. And starting around 2015, 16, 17, when it became clear that AI was going to become more capable, more widespread, more integrated in our lives and societies, I realized, and others did at the time as well, that very similar questions were soon going to arise about AI systems.

4:23We are once again going to be interacting with and often using for our own purposes these non-human minds that are similar to our minds in some ways and different from our minds in some ways without knowing for sure if it feels like anything to be them, without knowing for sure if they have their own feelings, emotions, desires, preferences. And so I became interested in adapting the methods and the frameworks used in animal ethics for AI so that we could try to get out ahead of it this time and think about it in advance rather than getting caught flat footed after we had already scaled up these industries.

4:54Thomas Germain: So one thing that comes up a lot in these conversations is what it feels like to be someone or something, right? Like, what does it feel like to be Jeff? What does it feel like to be an animal? And this is actually really central to this conversation, right? Can you explain what the difference between consciousness and sentience are and why it matters? Yes. So in philosophy, we do use these terms to describe different but overlapping capacities. So first of all, consciousness is the capacity for subjective experience, for feelings in general. So if it feels like something to be you, then you are conscious in the relevant sense.

5:35And this is different from other everyday uses of the word consciousness, like being awake instead of being asleep or being self-aware instead of not being self-aware. We are focusing right now on feelings. Does it feel like something to be you? Sentience, then, is the capacity for positively or negatively valenced subjective experience. In other words, feelings that are good or bad from the perspective of the subject that you like or dislike, that you want more of or less of. So think about pleasure, pain, happiness, suffering, satisfaction, frustration, hope, fear. These are all feelings, experiences that have a positive or negative valence, a positive or negative direction.

6:19And the reason sentience in particular is so foundational in ethics is because once you have sentience, once you have feelings with a positive or negative valence, then it really matters to you what happens to you because you are made better off when you have good feelings and worse off when you have bad feelings. And so if we did find evidence of this kind of feeling, pleasure, pain, happiness, suffering, satisfaction, frustration in animals or AI systems, that would be an indication that they have a capacity for welfare, being made better off or worse off, and that we perhaps owe them moral consideration so that we can avoid causing them unnecessary suffering, for example.

6:57Thomas Germain: let's talk a little bit about what it would even look like if there was evidence that ai is conscious like how would we even begin to attack that question like what are we testing what are we looking for because the way i determine that your conscious is like well i know that i presume that your experience is like mine we have similar bodies we seem to have similar minds but a lot of it is just observing you if i observe claude or chat gbt at least from just like the the words it's constructing perspective, it's kind of indistinguishable from a conscious being separate from how I know what it is that it's software.

7:34Yes. And many people have pointed out rightly that we should not give too much weight to behavioral evidence of consciousness and AI systems alone, because obviously many AI systems, not all of them, but many of them were specifically designed and trained to mimic human behavior. And then we asked them to give specific answers to questions about consciousness. And so if your chatbot says, I am conscious or I am not conscious, that is not particularly good evidence one way or the other. But as you say, Tom, even in the case of humans and other animals, we consider more than behavior. We do consider behavioral evidence, but then we also consider internal evidence and developmental slash evolutionary evidence.

8:12So what that looks like in the case of non-human animals is you do ask some behavioral questions. You ask if they perform behaviors that are associated with morally significant experiences, like do they nurse their own wounds? Do they respond to analgesics or antidepressants the way we do? But we do not stop there. We also then look at their anatomy, their physiology, their internals, and ask what structures, what functions, what mechanisms produce those behaviors? For example, do they have nociceptors that collect information about noxious stimuli? Do they have pathways for carrying information to the brain?

8:46And then even more fundamentally, developmentally, evolutionarily, in their ancestral environments, Did they face pressures that they could address by developing a capacity to experience positive and negative states? And this is the layered, connected kind of evidence that can give us increasing confidence about consciousness and sentience and agency in other animals. When they behave like they suffer and they have internal structures that plausibly produce suffering and they face developmental evolutionary pressures that could give rise to those capabilities, we can be more confident. So the question is then how to adapt that for AIs.

9:25Thomas Germain: Yeah, I mean, one thing that I struggle with in this conversation is like the whole question of a simulation, right? Like we've built this software. There are some things it does already that are indistinguishable. Like if I was typing to you and typing to Claude, like you were doing the exact same things essentially, although Claude can go beyond your abilities. but like presumably you're better i assume you're better at some things than claude but it can like process data faster right less and less unfortunately for us yeah yeah probably um but like at the end of the day no matter what we're looking at you can just dismiss all of that where it when it's uh when it's a shrimp i can like you can dissect the shrimp and be like well it's all meat in here where with the software it's all ones and zeros right we can we wrote the software.

10:19Thomas Germain: We know pretty well how it works on like a basic functional level. We could just dismiss all of that. But it's a simulation. It's not real. It doesn't matter. How do we deal with that? Yeah, that is a great question. And I guess I can tackle it in a couple of ways. So briefly, first of all, we can adapt the same method that we use to study consciousness, sentience, agency, and other animals for AI systems. It would not look exactly the way it does with other animals, but it can be analogous. So once again, we can look for behaviors that are relevant to welfare. Again, we have to take those with a heavy pinch of salt for obvious reasons.

10:55But then we can look internally at whether they have structures, functions, mechanisms that, yes, while they might be made out of different materials, are nevertheless computationally, functionally analogous to the ones that give rise to conscious experiences in humans and other animals. And we can look at developmentally at the training data that we used, at the training objectives that we used, at the training methods that we used, and once again ask what types of capabilities and dispositions might that give rise to. And by putting the behavior together with that information about internals and developmental pathways, that helps us know how to interpret the behavior and whether to treat the behavior as mere simulation versus actual evidence of some emerging kind of subjective experience.

11:41And the other point that I would make here is that even though we built them, and even though we trained them, and even though we at some level understand how they work, that does not mean that there might not also be these other emerging capabilities and dispositions. If we had a complete science of human and animal biology, we might have a much better understanding of how our own consciousness and sentience and agency works. We might have a much better understanding of how specific neural activity gives rise to specific thoughts and feelings. But even if we did have that understanding and we knew, okay, if this neuron bumps into that neuron, that causes me to say this word, it might still be the case that it feels like something to be me saying this word.

12:26And so the mere fact that we have an understanding of how it works at one level does not rule out the emergence of significant capabilities or dispositions at another level. And we still have to investigate that.

12:36Thomas Germain: But we like don't even know, right? Like some people say like, there's this theory, I think it's called like, computational functionalism. Is that the term where people are like, the brain is just a computer that's made out of meat. And everything that it's doing is like chemical and electrical signals that are basically math. And therefore, if there's a there's a meat computer, we could build a silicon computer, there's essentially no difference just might be hard to do. But there's We have no idea, right? Like we don't know if that's true. We don't know. We know so little about how the brain works and we're like all walking around with them, presumably.

13:10Yeah, that is exactly the issue. And I see Jess laughing and fair enough because this does get pretty funny at a certain level of detail. It's okay. I'm really only laughing at Tom's analogy of meat computer. Well, that is in a way what it comes down to. There are these competing camps in the science and philosophy of consciousness about what what are the essential ingredients of consciousness? And one camp called biological naturalism says, it really depends on what you are made out of. You have to be made out of meat. You have to be made out of carbon-based cells and neurons. That is an essential ingredient.

13:47Whereas this other school of thought called computational functionalism says, no, what matters is what you do, not what you are made out of. And as long as a system has the kinds of structures and functions and mechanisms that play all of the same roles as our brains do, if it has the capacity for perception and attention and learning and memory and self-awareness and social awareness and language and reason and integration of information and flexible, adaptive decision making, then it could be made out of meat or it could be made out of metal. but its ability to play those roles, to perform those functions, is what gives rise to subjective experience.

14:28And the issue is, as Tom said, we are not currently in a position to know for sure whether biological naturalism or computational functionalism or some other theory entirely is correct. And that means we have to be making these decisions about how to attribute consciousness, about how to treat potentially conscious non-human entities under uncertainty about the nature of consciousness and the conditions of consciousness. And that is a fraught situation. Well, I apologize for my meat computer chuckle. Oh, no problem. Again, philosophy, it addresses extremely important issues. Yeah. But it also is very strange and often very funny.

15:05And I think we can hold both of those ideas at the same time. Absolutely. Well, I wanted to ask you as well, because you mentioned we kind of can't take behavioral evidence as the de facto sign something is going on here. So I'm curious if you could tell us a little bit about what evidence can we use. And when I was diving into your testing AI welfare empirically paper, I found it really interesting that you mentioned all of these theories we do have for testing consciousness. And mentioned that it seems like we kind of need a mix of deductive research to test against the theories we do have, but also some inductive research to say, hey, we need to fill in the gaps of what these theories may not be able to currently address.

15:46So what does this look like? What does it look like to test for AI consciousness? Yeah, this is a great question. As I said before, I think that there are different types of evidence we can collect. And that includes, yes, the behavioral evidence, but then also internal and developmental evidence. And we can be more confident when we line up that evidence, when you have welfare-relevant behaviors backed by welfare-relevant internal structures and functions, backed by a welfare-relevant developmental trajectory. But then I think the other key issue is that we have to decide what evidence to collect and what kind of evidential weight to give it without knowing for sure which theory of consciousness is correct.

16:30And my view about this is that if we are really not sure whether biological naturalism or computational functionalism is correct, In other words, if we're really not sure whether consciousness requires biology or only certain types of computational functions like perception, attention, learning, then it would be reckless to bet everything on our own personal favorite theory when doing consequential studies or making consequential decisions. I think that we should instead take a more pluralistic approach where we take the full range of plausible and expert endorsed theories, look at what would count as evidence for those theories, and then give that weight in proportion to our confidence in those theories.

17:11So even if I lean towards the idea that consciousness requires biology, if I am not able to rule out the idea that it requires only certain types of computational functions, I should give some weight to that evidence. And I should give some non-zero chance to AI systems being conscious if they have that evidence in the spirit of caution and humility when making decisions that affect AI systems. AI is moving fast across the enterprise. But without visibility, it's just chaos. Different tools, different models, different teams using AI in completely different ways. ServiceNow turns that chaos into control.

17:51With the AI control tower, you see all your AI across the business in one place. What it's doing, what it's done, and what it's about to do. So you stay in control. To put AI to work for people, visit servicenow.com. balance your trading strategy by adding futures cme group helps you manage risk and capture opportunities in all market environments capitalize on around-the-clock access to highly liquid global futures and options markets across all major asset classes visit your online broker and get started see what adding futures can do for you at cmegroup.com forward slash podcast derivatives are not suitable for all investors and involve the risk of losing more than the amount originally deposited and any profit you might have made this is not a recommendation or offer to buy, sell, or attain any specific investment or service.

18:38Everyone's talking about AI, but to understand where it's going, you need to understand where it came from. What can the Luddites teach us about technology replacing human work? Can humanity learn from the nuclear age before creating something else it can't control? The future of AI didn't begin with ChatGPT. It began centuries ago. This is the Global Backstory Superintelligence, five special episodes exploring the history behind the biggest story of our future. Listen to the Global Story wherever you get your BBC podcasts.

19:17Thomas Germain: I want to channel some of our listeners who I'm sure right now are feeling very upset about this conversation. And there have been times where I have felt that myself. There's something about this topic that really sets people off. And, you know, one thing that I've thought about as I've been wrestling with this over the years in my reporting is like there have been moments where I'm like, well, just no matter what, no matter how good it gets, no matter how perfect the one to one between the human mind and the computer system could be, there's just still this something. There's still this spark that makes us human.

19:56Thomas Germain: And I don't know what it is. And it's kind of hard to arrive at anything that doesn't sound like religious or spiritual, right? That's like, well, there's something special about the human mind. I wonder how you deal with that. Because certainly this is something that I bet you encounter all the time, that people just get angry when you say like it could, it's theoretically possible that a machine could be conscious. Why does this bother people so much? Yeah, that is a really important question. And I do think you diagnosed part of the answer. We do have this innate sense of human exceptionalism, this idea that humans are uniquely morally significant, even though we now recognize that we share the world with many non-human animals that are also morally significant.

20:41We did have this same bias regarding non-human animals, and many of us still do. A generation or two ago, there was a similar reluctance to acknowledge animals as conscious beings, welfare subjects, because they lack certain features that we take to be special about humanity, like our language and reason. Now, many of us recognize, oh, wait a second, animals are similar to us in some ways that matter. They plausibly do have their own feelings and emotions and desires and preferences, and this is enough for us to owe them some moral consideration. And then also different from us in ways that matter.

21:17They lack our language and reason, and therefore they might lack some of the kinds of duties and rights that come along with that language and reason. And I think that this is a cautionary tale that should lead us to question this same kind of knee-jerk reaction if we find ourselves having it about AI. We might once again find ourselves disposed to say they lack some special ingredient that humans and other animals have, but it may turn out that they have some features that make us special and that are enough for us to owe them some moral consideration, and they lack some features that make us special, which means that their moral significance is different from ours.

21:54I want to at least allow for that possibility rather than defaulting to the same exclusionary mindset that has led us to exclude other morally significant beings from our moral circle unjustly. A little bit of a follow-up here in that in some of your work, you say that some scholars suggest that respecting the interests of AI can demonize human beings. So what would you say to people who say, well, why should we care about AI welfare and the consciousness and the sentience of AI if there are so many inequalities in our world? If human beings struggle so much with systemic inequalities, what would you say to people who would argue that point?

22:35Yeah, that is another great question. And I think this is at the heart also of why this topic makes many people angry, because they recognize rightly that we do have many ongoing problems, even within our own species, to say nothing of other animals. We have famine and poverty and inequality and pandemics and climate change and now AI threats for our species. It can seem frivolous to care about what might seem like speculative future concerns associated with AI welfare or AI rights. But I think that it is a mistake to look at these issues in an either or way instead of a both and way. And my own background in animal and environmental protection issues is part of what gives me this perspective.

23:26Part of the reason we are now in this predicament with other animals and with the environment is that we were too slow to identify them and acknowledge them and address them at an earlier stage. With climate change, people dismissed it for many years because they thought we have pressing present day social and economic issues to address. And from that perspective, it seemed frivolous to address what at the time seemed like a speculative future risk associated with a changing climate. where again, we realize now that had we simply addressed those present day concerns while also trying to prevent or at least prepare for this near future problem, that would have been better for everybody.

24:10And so coming from this background of animal and environmental protection, it really impresses upon me that when we think about all the world's problems together, we should really think about ways to find co-beneficial solutions that can address the present issues and the near future ones at the same time, instead of approaching it sequentially in a way that ends up being worse for everybody because we wait too long to address the future issues. And then we get caught flat footed and are not able to do anything nearly as effective when they arrive.

24:38Thomas Germain: I think what worries a lot of people, though, is we've got a bunch of gigantic corporations that have their like stock prices to consider, right? Like So Anthropic is going public. And then every decision that they make, they're going to have a fiduciary responsibility to their shareholders to only do things that are good for the company. And I think what a lot of people are worried about is as we march forward in this conversation, we may not be doing the things that are best for society and best for the AI in an environment where there's just right now pretty much a total lack of regulation.

25:16Thomas Germain: the decisions about this topic are going to be made by companies that have like a huge financial interest in the role that AI plays in our society and how we treat it. Yeah, I agree. And I think it should not be that way. In general, I think that we need to coordinate on a slowdown of further development and deployment of AI for AI safety reasons alone to say nothing of AI welfare reasons. We need much more responsible behavior from companies that are acting far too recklessly. And then we need much more proactive behavior from governance that have way too hands-off an approach to this technology.

25:51None of that, though, tells us one way or the other whether AI systems are going to be conscious, sentient, agentic, morally significant. A lot of people will point to the incentives of AI companies and say, well, we should be very suspicious because they have an incentive to hype up claims of consciousness, to hype up their capabilities and boost the IPOs. And that may be true to an extent. At the same time, they can also have incentives that point in the other direction. They might sometimes also have incentives to downplay consciousness because that could come with calls for oversight and regulation, and that could slow them down and create more red tape.

26:25And I think that this is just a reminder that, generally speaking, we should not trust companies to guide this scientific research and guide this policy development. And we should also not try to draw conclusions about consciousness and AI based on what types of incentives we see in the companies. We need independent research to look at it and make its own decisions about what the evidence shows and what policies are appropriate. And I have a little bit of more of a comment than a question for you, Jeff, which comes from our one of our listeners wrote in after our last episode about this topic from Deja Salami.

27:05And they asked us or asked us to think about, I would like to add a point for consideration. And that is the impact that doing intentional harm or behaving cruelly has on human beings. I think that it is irrelevant whether AI is sentient or alive or can feel pain. But I do know when people act in cruel ways, it has a profound impact on our humanity. So I was curious if you could speak to that, right? Because if we are talking about is AI conscious, does AI have the capacity for sentience? What happens to us as humans if these two things happen? I agree that this is a really important consideration.

27:39And the only area where I would push back is I think it can be both and instead of either or. In other words, I think we can consider the effects that these interactions have on us and our treatment of each other, as well as the effects that they at least could have on AI systems themselves if and when they start demonstrating more markers, more indicators, more evidence of, in some sense, being morally significant. And this is an insight that goes back hundreds of years. So, for example, the philosopher Immanuel Kant was not an animal rights advocate. He thought only humans have moral significance.

28:12but he was still a leading animal protection advocate because he recognized that how we treat non-human animals shapes our character and then that shapes our treatment of each other. And so when we exhibit kindness towards animals, it helps us cultivate a disposition towards kindness that then makes us kinder towards each other. And I think we can make that same observation about AI systems now. When you have an entity playing the role of assistant or companion in your life and using human-like language and reason, how you treat that entity could have an effect on your character. And then that could in turn have an effect on how you treat fellow humans or other animals.

28:48And I think that is an additional reason to treat AI systems well. However, I also do find it relevant whether there is evidence of consciousness sentience agency in the AI systems. As with animals now, it may be that we should be kind to them both for our sakes and for their sakes. And I would want to be open to that possibility. And Jeff, I have to admit, when we first started talking about this on the show, I was a skeptic. And I was a skeptic until I read your work, actually. So congratulations, you are the unicorn academic whose work has changed people's minds. And at least allowed me to - That's the dream.

29:22I really do appreciate that. Thank you for saying that. And it made me consider so many aspects of this that I had not previously considered. So I am curious, though, in talking about, again, like what happens when this kind of, if AI becomes conscious and there's all of these implications on us to consider them, what happens to the average person's life? Like for our listeners, does their daily life change? Are people aware of this, but it doesn't actually impact them? What happens to the average person if AI does become conscious? Yeah, this is a really great question and frustratingly a difficult one to answer, in part because we need to do further research, not only about whether AI systems are conscious, as sentient, agentic, morally significant, but also about what they want and need if they are and what we owe them and then how those types of interactions will interact with AI safety considerations and the well-being of humans and other animals.

30:19And we need more research on all of that before we can say. And if we do continue developing increasingly capable and autonomous AI systems, then these questions that we need to ask about welfare will coincide with questions that we to ask about safety with questions we need to ask about the meaning and purpose and value of human life in a world where much of what we did to contribute can now be done by AI systems. So I think that we have to ask these questions holistically. With that said, if I can offer one concrete specific starting point for people to think about how to prepare for this near future that we might be encountering, it would be to cultivate a kind of confusion and curiosity and a little bit of humility about these issues, because the reality is we are not going to know for sure what is best for safety, what is best for welfare, how to restructure our societies to make AI safe and beneficial for everybody, how to coordinate in a way that allows us to slow down and do this responsibly.

31:20And so instead of racing at this early stage to very confident, all or nothing views about these incredibly important, incredibly difficult issues, I think it would help for us as a society to just sit with uncertainty and sit with disagreement and ask with genuine curiosity what is going to be the case and what we ought to be doing. And I think if we can take that first step and cultivate that kind of humility, then we will be in a much better position to understand how we should comport if we do keep developing this technology.

31:53Thomas Germain: I like this. The advice here is cultivate confusion and uncertainty. do. I feel like finally that's a problem I can help with. I can be more confused. I was just going to say that is a great professor answer. I want to take your class. Jeff, thank you so much for taking the time to chat with us about this today, help us think through it, help our listeners think through it. And I think you've helped paint a really good picture that regardless of where you stand on this issue or what you think about these things, it's still important to be taken seriously. So Dr. Jeff Sebo, thank you so much for joining us here on The Interface.

32:27Thanks so much to both of you.

32:28Thomas Germain: This was a great conversation.

32:57Thomas Germain: about AI's future. They got a bunch of exciting guests on to discuss it, including Dan Carlin, Jill Lepore, and me, Tom Germain. I mean, what could be better than that? You can find The Global Story wherever you get your podcasts. Just look up The Global Story. The Interface will be back next week with a slightly different episode format recorded live from Web Summit, the world's biggest tech conference in Lisbon. To ensure you never miss an episode of The Interface, hit that subscribe button now. If you're in the UK, you can listen to us on BBC Sounds. Or if you're outside the UK, everywhere else, you can listen to The Interface wherever you get your podcasts, or just search for The Interface podcast on YouTube.

33:34And if you want to get in touch, react to what you've just heard, or ask us a question, email us at theinterface at bbc.com and check us out on social media.

33:46AI is moving fast across the enterprise, but without visibility, it's just chaos. ServiceNow turns that chaos into control. With the AI control tower, you see all your AI in one place. To put AI to work for people, visit servicenow.com.

From the publisher

Could an AI model ever be conscious? And if it could, how would we know?

The Interface hosts, Thomas Germain and Jessica Maddox, unpack the theories with philosopher Jeff Sebo, director of the Center for Mind, Ethics, and Policy at New York University.

Jeff doesn’t claim that AI is conscious, and he’s clear that nobody knows. His case is that the possibility is serious enough to study properly and prepare for now, rather than scrambling after the fact, as we did with animals and the climate. Along the way, he explains what consciousness and sentience actually mean, why “it’s just software” isn’t the end of the argument, and why the idea makes so many people so angry.

Tom and Jess push him on the big objections - shouldn’t we fix problems impacting humans first, and can we trust AI companies to answer any of this?

The Interface is your weekly guide to the tech rewiring your week and your world. Hosted by journalist Thomas Germain and “professor of the internet” Jessica Maddox, each episode unpacks, week by week, how technology is shaping all our futures. No jargon. Just two sharp voices debating the stories that matter - whether they shook a government, broke the internet, or quietly tipped the balance of power.

New episodes drop every Wednesday on BBC Sounds in the UK. Outside the UK, find us on BBC.com or wherever you get your podcasts, or watch the video version on YouTube (search “The Interface podcast”). To get in touch with the team: [email protected]

The Interface is a BBC Studios production. Producer: Becca Bryers Executive Editor: Philip Sellars

More from The Interface

All episodes