Alessandra Buccella is an assistant professor of philosophy at UAlbany who has thought a lot about what it would mean for artificial intelligence to become conscious, and why that question stresses us out so much. This week, she explains why consciousness in machines might look nothing like what we think — and why the “fleshy parts” matter.
Over the last few weeks, the news cycle has once again been boiling with AI anxiety after revelations that lots of OpenAI agents teamed up on a hacking spree and warnings from a former Anthropic researcher that AI could “kill us all” within the decade.
That was before Anthropic released a report suggesting someone in a military lab may have been trying to use its AI to try to build a bioweapon, among other nasty potential uses of their tech.
Then, over the weekend, Anthropic CEO Dario Amodei publicly called for a slow-down, or pacing, in AI development — a call echoed in some fashion by Sam Altman of OpenAI and Elon Musk of SpaceX.
In some ways, it’s a moment a lot like the one Blake Lemoine triggered with his controversial 2022 claims that Google’s LaMDA chatbot was sentient — only a lot bigger. The public’s attention is once again intensely focused on a technology most of us only lightly understand.
But is there an important distinction between whether AI is sentient (Lemoine’s claim) and whether it’s dangerous?
Because all this happened after we spoke to Alessandra, we went back to ask her take on it. Here’s what she had to say:
AB: The short answer is: The question of consciousness and danger are definitely separate in the sense that there are plenty of things that kill us without being conscious or wanting to cause us harm: bombs, earthquakes, hurricanes, diseases, etc.
The longer answer is that if AI ends up killing us all, it most certainly won't be because it gained consciousness and/or “decided” to kill us. What AI “does” is still exclusively the responsibility of the humans who built it, trained it, gave it objectives and unleashed it out in the world without enough safeguards in place.
As far as I'm concerned, all these tech people raising these apocalyptic worries as if they created a monster they can no longer control is just an excuse to avoid taking responsibility.
Alessandra also made an interesting point about the way the science fiction stories we consumed as kids continue to influence our understanding of what’s possible now that some of those technologies are moving from the realm of fiction to reality.
AB: A lot of the discussion about AI and consciousness and sentience and all that comes from those sci-fi stories that we grew up with. And they really influence a lot of what we think is possible with AI right now.
The simulation is an interesting case because you would even think that your body's not real. If I'm in a simulation, then all that I think I am and all that I think my world looks like, that's a lie. And at that point, if you think that that's a possibility, then conscious AI has also become a possibility. Because you could be mistaken about everything. And so even the things that sound completely implausible, why not?
Alessandra recently co-authored an article with UAlbany colleagues analyzing the apologies formulated by AI chatbots and why they’re meaningless – or, without the moral weight of knowing right from wrong.
Here’s the May 2026 article we mentioned about how some LLMs have shown the ability to pass the Turing Test in brief, controlled interactions with humans.
Blake Lemoine has done many interviews about his claims that Google’s LaMDA is sentient. Here is one with Bloomberg Tech.
Google’s 2022 response, via The New York Times: “Our team — including ethicists and technologists — has reviewed Blake’s concerns per our A.I. Principles and have informed him that the evidence does not support his claims. Some in the broader A.I. community are considering the long-term possibility of sentient or general A.I., but it doesn’t make sense to do so by anthropomorphizing today’s conversational models, which are not sentient.”
Audio editing and production by Scott Freedman
Interview and research by Bethany Bump
Photo by Patrick Dodson
Hosted and written by Jordan Carleo-Evangelist
[0:01] Host: Before we begin this week’s episode, we should acknowledge that A LOT has happened since we first spoke to our guest, UAlbany philosopher Alessandra Buccella.
How much, you ask?
Well, just in the last week…
None of this fundamentally changes what we wanted to talk to Alessandra about, which is the way humans insist on thinking about these machines in human ways — and how that impacts our understanding of whether they can become like us, whatever that means.
If anything, Alessandra’s insights are more relevant and necessary today than the day we spoke.
[1:25] Host: Welcome to The Short Version, the UAlbany podcast that tackles big ideas, big questions, and big news in less time than it takes to cross the Academic Podium. I'm Jordan Carleo-Evangelist in UAlbany's Office of Communications and Marketing.
[1:42] Alessandra Buccella: One of the oldest philosophical problems that we teach our students in the intro class is the problem of other minds. How do I know that you are not an automaton, that you're not a robot that just looks like a human and talks like a human — but are you a human? Is there something going on in there, or are you just a very clever machine that can make me think that there is something going on in there?
[2:04] Host: In mid-2022, Google fired a senior software engineer who went public with startling claims that the tech company's LaMDA chatbot had become sentient.
"It's not asking for much," Blake Lemoine told one of his many interviewers. "Before you experiment on it, it wants us to ask permission."
That's an unsettling assertion and it set the internet ablaze in predictable ways. AI skeptics cast Lemoine as an insider with firsthand knowledge that confirmed some of our most anxious thoughts about how much faster AI is developing than our ability to understand and responsibly use it. Others, including Google, dismissed Lemoine's claims saying there was no evidence to support them.
Some argued the entire debate was a distraction from the real and more immediate ethical and safety questions that AI developers need to confront. More recently, some AI systems have shown that they can reliably convince humans of their own humanity, a benchmark known as the Turing Test that was long considered a significant milestone in AI development — and for some, a bright cautionary line.
There are deeper, more fundamental questions here than the capability of the latest large language model out of Silicon Valley. What would it actually mean for an AI system to be conscious? How would we know? How would it tell us? Could we believe it?
Assistant Professor of Philosophy Alessandra Buccella spends a lot of time thinking about these questions. A faculty fellow with UAlbany's AI and Society Research Center, Buccella believes philosophers have a critical role in enforcing conceptual clarity around the words we use to talk about AI — especially given humans' very human tendency to think of almost everything in human terms.
Last week, we spoke to Balakrishnan Prabakaran, the director of UAlbany's AI Plus Institute, about what distinguishes artificial and human intelligence. My colleague Bethany Bump continues that conversation this week with Alessandra, exploring what it means to act with intentionality, to have a unique perspective on the world; the risk of conflating functional intelligence with self-awareness; and why even asking that question seems to stress us out so much.
Here's their conversation.
[4:17] Bethany Bump: How did you become a philosopher? When did you know that was something you wanted to seriously pursue?
[4:23] AB: So I guess I was always the kid that asked the weird deep questions growing up. At some point it was always, I would have conversations with my friends and I would ask, "But what about this? But what is this thing that you're talking about?" Trying to always go to the more fundamental level, and then people would throw their hands up and like, "I just lost track of what you're talking about at this point."
And then in Italy, we start studying philosophy in high school actually. And when I started, I was just like, "This is the kind of questions that I'm interested in exploring. This is what I like." And then it turns out that one of my dad's best friends is a philosophy professor. So I learned that such a thing existed — like a professional philosopher. Do I get to do this for a living, to just ask deep questions all the time and just think about them? That's what I want to do for a living, for sure.
[5:18] BB: Tell me how you first got into studying the philosophy of artificial intelligence.
[5:23] AB: So my first love as a philosopher has always been cognition and perception. So how do we interact in the world, and how come that we have the experiences that we have? How do we learn to think? When AI came around, it was such a natural transition for me to go from studying human cognition and developing views about what it means to be a human thinker, what are the features that make us the creatures that we are in terms of what we can do in the world. And then there are these systems that are advertised as, oh my God, they're even better than humans, or they can do all these things that humans can do. And so to me it was like, okay, I need to go and explore that more and learn more about what these things actually do. And then as a philosopher, we feel some kind of responsibility to go and make sure that we keep the conversation organized when it comes to artificial intelligence because of this tendency that humans have to anthropomorphize everything.
Philosophers like conceptual clarity. And so when people started talking about artificial intelligence and all these amazing things that it can do, as a philosopher of mind, my first reaction was, okay, I need to go and check whether really these claims are well-founded, whether there is really some similarity between how I think humans work cognitively and how these things behave and how these things work.
[7:05] BB: Big question, which has to do with AI consciousness and this debate that people have been having over whether AI will become advanced and intelligent enough that it gains something akin to human consciousness. So how would you define consciousness?
[7:22] AB: When we think about the mind, we think about on the one hand this notion of intentionality, which essentially means our mental states, our mental attitudes, are always towards something else. They're always about something like our thoughts. For example, you think that Albany is the capital of New York State, that thing that comes after that is what your thought is about. Minds are things that are capable of being in states that are about something else, that refer to something outside of themselves. And the other big feature is consciousness. I think about it as subjective experience. To think or to experience something or to believe that something is true or to want or desire something to be true, what it's like to have a unique perspective on the world and your own specific point of view and the qualitative aspect of what it is to have that point of view. That's how philosophers think about consciousness.
[8:25] BB: It seems like a level of awareness, not even necessarily self-awareness, although I'm sure that would be part of it.
[8:30] AB: I'm of the view that you can't really have one without the other in the sense that if consciousness is awareness of your surroundings, then that must involve some kind of point of view from which you're experiencing your surroundings. You need to have some kind of unique perspective, a point of view on the world. And to me, that sounds a lot like some kind of self-awareness.
[8:55] BB: I know that you've said before that cognitive science assumes a functionalist view of the mind. Why is it so influential in how people tend to think about artificial intelligence?
[9:06] AB: Yeah, so what is functionalism? Functionalism is a theory or sort of a theoretical attitude that people have towards cognition that basically says what makes something a cognitive process is its functional role. Think about a chair. You can describe a chair in functionalist terms by saying a chair is something that allows human beings to assume a certain body position and have their weight distributed so that they don't have all their weight bearing on their legs. You describe what a chair is by saying what a chair does. And then people have taken this functional way of understanding objects and transferred it to the mind, saying to be in a state of mind that can trigger other states of mind. So what it is to have that belief that Albany is the capital of New York State is to be, for example, disposed to believe that there is going to be a capital building in Albany.
So there's a network of other mental states that go around that act as inputs and outputs, perceptions, sensory experiences on the input side. And then there's the output side, which is action, behavior. This functionalism is conception of the mind as essentially an input-output processing machine, and that's how it connects to AI. We started thinking about the mind as a kind of computer. When computers started coming out and you could see that they exhibited certain behaviors that looked intelligent in some sense, then people started thinking, “Oh, but then maybe the mind is also a kind of computer. If a brain can implement a mind, can have mental states that are realized in certain ways, then so can a dog, so can an alien, so can a totally artificially built computer. If all that matters for cognition is this input- output information processing, then the physical thing that implements that chain can be any sort of thing.
[11:27] BB: A lot of the debate about will AI become conscious seems like it's maybe conflating intelligence with consciousness.
[11:35] AB: Yes.
[11:36] BB: Can you sort of differentiate between those two for us?
[11:39] AB: For sure. What usually people associate with the notion like artificial intelligence is something like planning step-by-step problem solving. An intelligent thing is a thing that can take in a situation, break it down into its most relevant steps or pieces, identify a problem, and then organize those pieces in order to solve that problem. It's intelligence in the sense of reasoning according to logical rules, or in the case of AI, it's not logical rules anymore now with machine learning. It's like probabilistic reasoning. We think about this very mathematical, organized, step-by-step type of problem solving, but that's not all that intelligence is, and that doesn't seem to have anything to do with consciousness. You don't need a specific unique point of view on the world in order to perform abstract mathematical operations. The two things are really kind of separate and people tend to conflate them because they tend to think that if you're intelligent, then you need to be like us.
And because we're conscious, then if we're intelligent, we're conscious, therefore something else is intelligent and must be conscious. But that depends very fundamentally on a certain conception of both consciousness and intelligence.
[13:03] BB: Let's say that we develop an AI system that can fully mimic the human brain. Is that enough for consciousness to arise or are there factors beyond the brain that influence consciousness?
[13:16] AB: On one hand, you have the question, if we could create an AI system that mimics the human brain, would it be conscious? In that case, I would say it depends on what you mean by completely mimics the human brain because AI systems are inspired by the brain, but they're also not really like the brain because it's not just about electrical signals going from neuron to neuron. A brain does all sorts of other things with neurotransmitters, and there's a lot of the fleshy part, the biological processes and the chemical processes that go on inside the brain are just as important, if not more important, than whatever you can describe at this abstract functional level of input, output, information processing. If we could build a system that has all those chemical and physiological and biological reactions as well, then probably it will be conscious. But at that point you don't even have an artificial intelligence system anymore, you've built a brain.
And the second part was about the body and how even if we could build a brain that would not be enough for consciousness. That's something that people debate in philosophy because there are theories of consciousness out there that basically just say it has to be something that happens in the brain. Other people would say, “No, a brain by itself would not be able to do any of the things that it does, including developing some kind of self-awareness and consciousness.” To gain that unique perspective on the world, you need a body to explore the world through and experience how the world in the way reacts to your explorations.
[15:05] BB: I feel like language also comes to play in this debate around consciousness. We know that humans tend to anthropomorphize or think about objects in these very human ways. So is the fact that the large language models appear to be so adept at language what makes them seem so human-like?
[15:26] AB: Absolutely. Even before large language models came around, people were pretty much settled on computers cannot be conscious. But the moment you have these systems that exhibit such a fluency and competence in using language, then it's very tempting to make this implicit inference saying, oh, if it can talk like me, then it must be like me, except that that's not really the case. Language, if you think about it, is a very mechanical process in a lot of ways. We talk sometimes just out of patterns and phrases that we've incorporated and we sort of automatically deploy even with each other that are just part of how we relate in the world, but there's not a lot of cognitive conscious elaboration behind them. If you sever the two and you think that one thing is to mimic how to use words, but another dimension is to mean what those words actually mean, then where does that meaning come from?
Well, it comes from the fact that language emerged as a way to communicate for creatures like us in a world like ours. So the meaning of those words depends on being humans that have evolved in a certain environment and they invented those words to refer to things that are part of their environment. An AI machine is uttering sounds, but it's very doubtful that it has an underlying understanding of why those sounds are organized the way they are and what they actually mean.
[17:21] BB: So let's put you on the hot seat here. Do you think AI can become conscious? Is there a possibility?
[17:28] AB: The easy short answer is most likely not. The slightly more nuanced answer is: If it gets to a point where its information processing is sophisticated enough and it has a body and it starts really learning from its own interactions with the environment and so on, which is very far away from anything that our systems are doing right now, then it might be conscious. But I would say that consciousness would look very different from what we think consciousness is for humans. We wouldn't really be able to tell even if the machines are conscious because it would look so different from what we take to be consciousness.
[18:20] BB: Why do you think people are so preoccupied with this question of can AI become conscious? How would that change the way that we interact with AI or the role that it plays in our lives if it is conscious?
[18:33] AB: When you hear on the news, people are really distressed about when they start believing that these systems might be conscious, their whole attitude towards them changes because we're so used to the idea that if you're conscious, then you're a creature with a perspective, with a point of view that can feel things. And if you can feel things, then you can suffer or experience pleasure. And if that's the case, then you start occupying a role in our society that is very different from that of just a pure object. At that point, we would have to worry about how do we treat you if you're conscious, if we know that you can feel pleasure and pain, that changes how we behave towards you, what kind of rights you might have. So once you make that step, then they all of a sudden acquire a role in your society that you need to organize in a certain way and you need to have certain attitudes towards these differently from if they were just the same as water bottles.
[19:38] Host: That was Alessandra Buccella, assistant professor of philosophy at UAlbany.
If you want to hear more from Alessandra about how the science fiction stories we grew up with influenced the way we think about the real AI systems that exist today, check out The Longer Version in our show notes. We've also posted links to some of Blake Lemoine’s interviews, as well as a recent study reviewing how some AI systems have passed the Turing Test.
If you missed last week's episode with the director of UAlbany's AI Plus Institute, be sure to check that out, too.
The Short Version would not be possible without contributions from many people, including audio editing and production by Scott Freedman in UAlbany's Digital Media Studio, deep inside the podium tunnels. You never hear him on this podcast, but I promise you he's a real person.
We'll be back next week with another conversation about something interesting. I'm Jordan Carleo-Evangelist here at the University at Albany, and this has been The Short Version.