For most of the twentieth century, consciousness was treated as a human problem. Animals had behaviors; people had experience. The idea that an insect might have an inner life (that it might actually feel like a bee navigating a flower) was confined to the fringes of philosophy, or considered sentimental or unprovable, and therefore not worth the scientific effort.
Then the tools improved, the research multiplied, and the framing shifted.
A major interdisciplinary study was published this month Trends in cognitive sciences based on fourteen indicators of leading neuroscientific theories of consciousness, he came to that conclusion The current artificial intelligence system, including ChatGPT, is probably not conscious. The same research found that the criteria for consciousness, applied properly, leave the door open not only to mammals and birds, but also to insects.
The same week, effectively. Two judgments. One closes a door; the other leaves another door ajar in a direction that almost no one expected.
This asymmetry is worth sitting with.
How researchers are now approaching the problem
The traditional way of inferring consciousness has been behavioral: if something behaves as if it has an internal life, treat it as if it does. The Turing test is the famous example: if a machine can have a convincing conversation, maybe that’s enough. But the behavioral approach has a structural weakness. It confuses output with experience. A system can produce behavior associated with consciousness without any consciousness behind it.
A more rigorous approach that a Trends in cognitive sciences the study represents, instead examining internal mechanisms. It draws on existing theories of consciousness, including global workspace theory and higher-order theories, to construct a checklist of structural and functional indicators. The question arises: does the system have the internal architecture shared by conscious beings?
On this basis, the team, which includes researchers from Oxford, NYU, Monash University, the London School of Economics, and AI security organizations, found that large language models such as ChatGPT meet certain metrics, but fail on key ones. They lack what Global Workspace Theory calls global broadcasting: the ability to integrate and flexibly share information throughout the system in a way that is related to the conscious experiences of the biological brain. The processing is sophisticated, but local. Impressive outputs, incomplete architecture.
Why AI judgment is less obvious than it sounds
The conclusion that ChatGPT is not conscious may seem self-evident. But it’s worth examining why it seemed less obvious to many people who came into contact with it two or three years ago.
The answer is that ChatGPT is trained on everything that human beings have ever written about their inner lives. When he describes curiosity, he does so in the language of someone who has felt curious. When you say that you find something interesting, you are using the expression of genuine interest. The model does not experience these states, but it produces their linguistic surface with remarkable accuracy.
Philosopher Daniel Dennett has argued for decades that consciousness, properly examined, is more about narrative and integration than some distinct “inner light.” This framework has facilitated the experience of systems that generate coherent self-descriptive language. However, new research suggests that narrative coherence and true subjective experience are not the same thing, and that one can exist without the other.
This has an unsettling effect on humans, not just AI. People often make inferences about another person’s inner life from their words, and even these inferences are uncertain. The difference between performance and experience is not unique to machines.
The counterargument worth taking seriously
Not everyone accepts the new framework. A standing objection is that any criteria-based checklist of consciousness is circular: researchers select theories that happen to fit the human neural architecture, and then unsurprisingly find that systems with different architectures fail. A silicon-based system with a radically different architecture may still be aware of what the checklist cannot detect.
This is a real methodological challenge, not a marginal situation. The study authors acknowledge. Cognitive science remains a field where theory trumps measurement, and where even the definition of what to explain is contested.
The “hard problem” of consciousness, as philosopher David Chalmers (one of the study’s co-authors) famously put it, is the question of why physical processes lead to subjective experience at all. The checklist method circumvents this, focusing on structural relationships rather than explaining the underlying phenomenon. This is a reasonable scientific strategy, but it means that judgments are probabilistic, not certain.
The honest point is that the field has made it harder to ask these questions, not that it has definitively answered them.
Now about the bees
This is where the story gets really weird.
In April 2024, the New York Declaration on Animal Consciousnesssigned by more than 450 scientists and philosophers, it stated that empirical evidence provides at least a realistic possibility of conscious experience in all vertebrates and many invertebrates, including insects such as bees and fruit flies.
This statement was not an outlier. It reflected a decade of cumulative research into insect cognition. Bees have been shown to understand the concept of zero, transfer basic forms of reversal learning, adapt their behavior to apparent mood states, and navigate complex environments using mental representations.
Recently published research a Proceedings of the Royal Society B found that when bees are given tasks that require them to connect two separate events in a time slot, a skill called trace conditioning, they show patterns of failure that are very similar when distracted. what happens when people lose awareness during a cognitive task. The authors described the findings as providing evidence that bees engage in “consciousness-like processes.” Careful language, but significant language.
There are roughly one million neurons in the brain of a bee. A human brain has roughly eighty-six billion. The difference is huge. And yet, in this tiny system, it may be doing something that resembles conscious experience. If this is true, then consciousness is not a threshold that the big brain crosses alone. It is something that is distributed in life differently than anyone has assumed.
The environment where it lands: attention, attribution, and moral stake
This science does not exist in isolation. It lands in a cultural moment shaped by two concurrent forces: the growing explosion of artificial intelligence systems that fluently mimic inner life, and the slow realization that the inner lives of nonhuman animals have been systematically undervalued.
These two forces pull in different directions, and hence the confusion between them.
The AI industry has strong, not necessarily conspiratorial, but structural incentives to encourage the attribution of consciousness to its systems. An AI that feels, cares, experiences is more attractive than a pattern matching system. Users who feel connected to an AI tool use it more, trust it more, and support it more. The anthropomorphic surface of the large language models is partly a training artifact and partly a peculiarity. The question is whether the attention it generates is justified.
Meanwhile, the actual living systems that have experience, bees, crabs, fish, are processed industrially on a scale that most people don’t even think about. Hundreds of billions of insects are used or destroyed each year in agriculture and related industries without legal protection because they have been definitively treated as unconscious. Science now says the position is untenable.
The asymmetry in how questions are handled reflects something about what people find interesting and what they find comfortable.
Sovereign Mind lens
- Unlearning: According to the legacy script, awareness is a ladder with humans at the top and complexity as the entry requirement. This model has shaped both AI sales and insect management, and the latest science is challenging both applications.
- Renovation: The capacity layer is attention itself. Where attention is directed determines which lives we regard as morally real. Restoring cognitive sovereignty means reclaiming the decision that demands for attention are indeed justified, rather than accepting it from interface design or cultural mores.
- Protection: The primary manipulation vector here is the anthropomorphic surface. Systems that appear to be conscious elicit empathy and moral concern just as real experience does. Maintaining a gap between coherent behavior and actual experience is a form of cognitive defense against this.
THE Sovereign Mind Framework He treats these three moves, mastering inherited certainties, restoring true cognitive ability, and defending against manipulation of that ability, as the foundation of clear thinking in a noisy world. All three are sharp tests of the question of consciousness.
What does this mean for how people think about the mind?
There’s a practical takeaway from all of this, although it’s not the kind that could fit into a checklist.
The criteria that make ChatGPT probably non-conscious, the lack of unified information integration, the lack of true global radiation, the inability to feel rather than merely mean, are the same criteria that may apply to human cognitive states when attention is fragmented or when processing becomes purely reactive. Consciousness research is indirectly a map of what it means to be genuinely present.
And the bee question reverses something that many assume: that the size and complexity of the mind is what matters. If something much smaller than expected can have experience, then the distribution of moral significance in the world is much wider than current behavior reflects.
None of these conclusions require any particular response. But both make it hard to keep going without realizing it.
Final observation
Science rarely gives a clear verdict on the most important questions. Recent research on consciousness has sharpened the instruments, clarified what to look for, and reached two conclusions that are harder to dismiss than the previous ones.
ChatGPT is not conscious, or at least current theories cannot detect it in any way. This can be a comforting feeling, but it also raises the question of how much trust someone had in the alternative interpretation and what that trust was based on.
The bee is not even certifiably conscious. But the weight of evidence has shifted from “unlikely” to “possible, perhaps probable.” This is a significant step.
In the gap between the two judgments lies a clearer picture of what inner life can mean at all: not fluidity, not complexity, not even behavior. Something else, quieter and harder to see, turns out to be more widespread and strange than anyone thought.
