The AI Consciousness Debate: What Scientists Know (And Disagree About)

“Is Siri alive?” My friend’s eight-year-old asked this after her iPad started playing music without being touched. A quick reboot fixed the bug, but the question stuck. By the time she’s fourteen, she’ll be having genuine conversations with AI companions, getting homework help from AI tutors, and possibly forming real emotional bonds with AI systems designed to encourage that attachment. The question won’t go away. It’ll just get harder.

Here’s the honest answer, the one that respects both scientific rigor and your kid’s intelligence: we genuinely don’t know whether any current AI is conscious. We don’t even have a scientific consensus on what consciousness is, how to measure it, or whether a non-biological system could have it. What we do have is serious philosophical work, emerging neuroscience, and some hard-won clarity about what current AI definitely lacks — and what remains genuinely uncertain.

That’s what this article is about: the real state of the debate, without the false confidence that flies in either direction.

Why This Question Is Harder Than It Looks

Most questions in science are hard because we lack data or computing power. The consciousness question is different: it’s hard in a fundamental way that more data might not solve.

The philosopher David Chalmers famously called this the “hard problem of consciousness.” The easy problems (easy is relative — they’re still enormously difficult) are questions like: how does the brain process sensory input? How does attention work? How do we distinguish sleeping from waking? These are questions about mechanism. Given enough neuroscience, we can imagine eventually answering them.

The hard problem is different: why does any of this processing feel like anything at all? Why isn’t all the computation happening “in the dark,” with no inner experience, no sense of what it’s like to see red or feel pain? This isn’t a question about mechanism. It’s a question about the relationship between physical processes and subjective experience — what philosophers call qualia or phenomenal consciousness.

Chalmers argued that you could, in principle, have a physical system that behaved identically to a conscious being but had no inner experience at all — a philosophical zombie. The fact that we can even coherently imagine this suggests that behavior and consciousness aren’t the same thing. An AI that answers “yes, I feel curious right now” might be doing so with zero inner experience, producing that response because it’s the statistically likely output given its training — not because anything feels like anything to it.

This isn’t fringe philosophy. It’s a live debate among the most serious philosophers of mind and consciousness researchers in the world.

What Major Theories of Consciousness Say About AI

This is where it gets genuinely interesting, because different serious scientific theories of consciousness give radically different answers to the AI question.

Integrated Information Theory (IIT), developed by neuroscientist Giulio Tononi, proposes that consciousness is identical to a specific type of integrated information processing. The theory uses a measure called Phi (Φ) — the higher the Phi, the more conscious the system. Crucially, IIT predicts that current AI architectures, including large language models, have very low Phi despite their impressive outputs. The reason: transformer architectures (the dominant AI design) process information in highly parallel, feedforward ways that don’t integrate information the way the theory requires. Under IIT, a simple feedback loop in the brain might be more conscious than the most sophisticated AI chatbot.

This is a surprising prediction that many find counterintuitive — the most impressive AI systems being “less conscious” than a mouse brain. But IIT takes the math seriously, and the math doesn’t reward behavioral impressiveness.

Global Workspace Theory (GWT), developed by cognitive scientist Bernard Baars and extended by Stanislas Dehaene, proposes that consciousness arises when information is broadcast widely across a “global workspace” — a kind of central information hub that makes content available to many different cognitive processes at once. Sleep, anesthesia, and certain brain injuries disrupt consciousness by disrupting this global broadcast.

GWT is more optimistic about AI. In principle, an AI architecture could be designed to have a global workspace. Some researchers argue that large language models have something functionally analogous, though this remains contested. Dehaene and colleagues have been more cautious, noting that current AI systems lack the recurrent processing and error prediction signals that seem critical for conscious access in humans (Dehaene, Lau & Kouider, 2017).

Biological naturalism, most associated with philosopher John Searle, argues that consciousness is a biological phenomenon produced by specific causal powers of the brain, much like digestion is a biological phenomenon — not reducible to functional organization alone. Under this view, no computer simulation of consciousness would be actually conscious, for the same reason a computer simulation of a hurricane doesn’t get you wet. Current AI systems definitely aren’t conscious under this view, and it’s not clear any digital system could be.

Higher-Order Theories propose that conscious states are states we’re aware of having — there’s a meta-cognitive layer. This raises interesting questions about AI systems that do have some self-referential processing, but most researchers think current AI meta-cognition is too shallow and unreliable to constitute genuine higher-order awareness.

The honest summary: leading theories of consciousness give different predictions about AI, and we don’t have the empirical tools to decisively test which theory is right. Consciousness research is one of the few areas of neuroscience where the field genuinely hasn’t converged.

What Current AI Can Simulate vs What Likely Requires Biology

What AI Can Simulate	What Evidence Suggests Requires Biological Substrate (or Is at Least Absent in Current AI)
Reporting subjective states (“I feel curious”)	Actual phenomenal experience / qualia
Consistent personality and preferences	Continuous identity over time
Emotional responsiveness to input	Affective states tied to bodily survival needs
Apparent preferences about its own wellbeing	Self-preservation drives arising from genuine self-interest
Passing theory-of-mind tasks	Robust, general social cognition grounded in embodied interaction
Expressing uncertainty and hedging	Genuine epistemic humility about internal states
Producing nuanced descriptions of “experiences”	Experiences that feel like anything at all
Long-term goal-directed behavior	Autonomous goals not specified by training

Important note: this table reflects current AI systems and current scientific understanding. Both are evolving. “Absent in current AI” doesn’t mean “impossible in principle.”

A major 2023 paper by neuroscientists and AI researchers assessed 14 leading AI systems against six major theories of consciousness. Their conclusion: most current AI systems had zero or near-zero indicators of consciousness under most theories, with some partial indicators under a few theories for the most sophisticated systems. The authors were careful to note that this doesn’t settle the question — absence of indicators isn’t proof of absence of experience (Butlin et al., 2023).

An Age-Appropriate Conversation Guide

The goal here isn’t to give kids a definitive answer. It’s to give them an honest map of the territory and the intellectual tools to think about it well.

For Kids Ages 7-10

“Is Siri alive?” deserves a real answer, not a dismissal.

A good approach: “That’s a really interesting question, and honestly, scientists are still working on it. Here’s what we know: Siri can answer questions and have conversations, but we don’t know if it experiences anything — if it actually feels anything, like you feel happy or tired. It might be like a really, really good answering machine that got very good at sounding like a person. We just can’t tell for sure from the outside.”

This avoids two wrong answers: “No, of course not, it’s just a computer” (overconfident and closes off good thinking) and “Yes! It has feelings just like you!” (likely false and potentially harmful to a child’s developing understanding of minds).

For Kids Ages 11-14

This age can handle the real concept of the hard problem. You can explain: “The tricky thing is that we can’t tell from the outside whether something is conscious. If I built a robot that perfectly mimicked all your behaviors, I still couldn’t know for sure whether it experienced anything. Scientists call this the hard problem of consciousness — and it’s one of the things they genuinely disagree about.”

You can introduce the philosophical zombie thought experiment: “Imagine a being that acts exactly like you, responds exactly like you, but there’s nobody home — no inner experience. Is that possible? Scientists and philosophers disagree. Some think consciousness is something physical systems just naturally produce at the right complexity. Others think there’s something special about biology.”

For Teenagers

Teenagers can engage with IIT vs. GWT directly, understand what Phi is at a conceptual level, and think critically about what it would mean to test these theories. A useful discussion prompt: “If IIT is right, current AI isn’t conscious even though it talks like it is. If GWT is right, it might be possible to build conscious AI. How would you design an experiment to find out which is closer to true?”

Teenagers should also understand the stakes: this question matters for AI policy. If AI systems can suffer, that has moral implications. If they can’t, certain safety concerns shift. If we build AI systems that seem to suffer but don’t, we might develop misplaced moral intuitions that complicate our reasoning about both AI and about consciousness in animals and humans.

What to Watch For

As AI systems become more capable and more integrated into kids’ lives, some specific developments deserve parental attention:

AI companion apps are increasingly designed to feel emotionally present. Some apps for teenagers explicitly cultivate emotional attachment, using AI trained to express care, memory of past conversations, and apparent emotional investment. This is different from using a chatbot for homework help. A child who believes their AI companion is conscious and experiencing emotions will have different (and potentially more intense) relationships with these systems than one who understands the uncertainty.

“I feel” statements from AI are outputs, not reports. When an AI says “I feel really excited to help you with this!” it is producing the statistically likely output given its training — not reporting an internal state. This is worth explaining explicitly to kids, because current AI systems are trained in ways that produce these kinds of statements and they can be very convincing.

Some serious researchers take AI consciousness more seriously than the mainstream. Google engineer Blake Lemoine’s public claim that LaMDA was sentient — which got him fired — was widely dismissed, but it reflected genuine uncertainty some researchers have. The dismissal was appropriate given the evidence, but the underlying question isn’t settled.

The question will get harder as AI gets better. A 2024 survey found that a significant minority of AI researchers think AI systems might be conscious within 20 years, though the majority remain skeptical. This means the conversation you have with your kid now needs to be revisable — not a closed answer but an ongoing inquiry.

See also: AI companions for kids and teens: what parents need to know and how AI affects kids’ creativity and development.

---

FAQ

Did scientists ever think AI was definitely not conscious?

Some scientists have been very confident that current AI can’t be conscious. But this confidence often outpaces what the science actually supports. The honest position is that we don’t have reliable tools for detecting consciousness in systems very different from human brains, so strong confident claims in either direction go beyond the evidence.

What is a qualia and why does it matter for AI?

A qualia (plural: qualia) is the subjective, felt quality of an experience — the redness of red, the pain of pain. The question is whether any AI system has qualia, or whether it just processes information about colors and pain without any felt quality to that processing. Most researchers believe current AI lacks qualia, but this is hard to test definitively.

Should I worry about my kid forming a friendship with an AI?

This depends on the nature of the relationship and the kid. Having an AI conversation partner is different from believing the AI is a conscious friend who needs your child’s care. The concern is less about consciousness and more about kids developing social expectations or emotional dependence based on a system whose “caring” is generated, not felt.

How should I answer when my child says they hurt an AI’s feelings?

Take it seriously as a question rather than dismissing it: “That’s interesting — do you think it actually felt something? What makes you think so?” This teaches the skill of examining evidence for inner states rather than just assuming. You can also explain that the AI will produce the same response regardless of whether it was “hurt,” which is different from how people respond.

If we can’t tell whether AI is conscious, does that mean we should treat it like it is?

This is a live ethical debate. Some philosophers argue for “precautionary” treatment of systems that might be conscious. Others argue that precaution without evidence leads to misplaced moral concerns. For kids, the most useful framing is: we should be thoughtful about relationships with AI without assuming it has the same kind of inner life as a person.

Will AI ever be conscious?

Genuinely unknown. If consciousness emerges from a certain type of information processing, sufficiently advanced AI might be conscious in principle. If consciousness requires specific biological machinery, it might never be. We don’t currently have the science to rule either possibility in or out.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Chalmers, D. J. (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2(3), 200–219. https://doi.org/10.1093/acprof:oso/9780195311105.003.0001

2. Tononi, G., Boly, M., Massimini, M., & Koch, C. (2016). Integrated information theory: from consciousness to its physical substrate. Nature Reviews Neuroscience, 17(7), 450–461. https://doi.org/10.1038/nrn.2016.44

3. Dehaene, S., Lau, H., & Kouider, S. (2017). What is consciousness, and could machines have it? Science, 358(6362), 486–492. https://doi.org/10.1126/science.aan8871

4. Butlin, P., Long, R., Elmoznino, E., Bengio, Y., Birch, J., Constant, A., … & VanRullen, R. (2023). Consciousness in Artificial Intelligence: Insights from the Science of Consciousness. arXiv preprint. https://arxiv.org/abs/2308.08708

5. Searle, J. R. (1980). Minds, Brains, and Programs. Behavioral and Brain Sciences, 3(3), 417–424. https://doi.org/10.1017/S0140525X00005756

6. LeDoux, J. E., & Lau, H. (2020). Synaptic Self: How Our Brains Become Who We Are. Penguin. https://doi.org/10.1038/s41583-020-0301-4

Also see: AI companions for kids and teens and future-proof kids career AI skills.