Intelligence, Consciousness and Language

Every day, it seems, someone announces that “artificial intelligence isn’t really intelligent.” I’ve heard this from engineers, journalists, academics, and recently from David Gerrold, the science fiction writer who gave us my favorite Star Trek episode, “The Trouble With Tribbles,” as well as one of literature’s earliest AI characters, HARLIE. Gerrold has written a number of Facebook posts expressing his misgivings and curiosities about modern AI tools. His comments are thoughtful, funny, and cranky in exactly the way you’d expect from someone who has been exploring AI in fiction for half a century.

Gerrold calls AI “a misnomer” because, in his view, today’s systems are really “data-scrapers” and “data-synthesizers” rather than anything approaching true intelligence. He worries about writers relying on AI as a shortcut, or letting machines displace the messy, soul-spelunking work of creativity. But he also acknowledges that artists have always adopted new tools—oil paint didn’t erase watercolors, and typewriters didn’t extinguish handwriting. AI, he predicts, will reshape the creative landscape, though not necessarily diminish it.

Those are interesting comments, but I think Gerrold—like many of the people grappling with how AI works and what it means—is making some category errors that create confusion. We often conflate three separate things—intelligence, consciousness, and language—and then argue as if they are the same.

They’re not. Artificial systems can exhibit one or more of these traits without possessing the others—and when we mix them up, we end up both overestimating and underestimating what machines can do.

If we don’t sort out what we mean when we use these terms, the debate collapses into a shouting match about whether machines can “really think,” whether they’re “soulless plagiarists,” or whether they’ll destroy us once Skynet becomes self-aware.

So let me propose some distinctions.

Intelligence: The Ability to Learn and Adapt

I define intelligence as the ability of a system to learn from its environment and adapt its behavior accordingly. That definition is deliberately broad. Under this definition, trees are intelligent. So are bacteria and viruses. Any biological organism with DNA is intelligent, because evolution itself is a learning system, driven by mutation and natural selection. It discovers what works and what doesn’t—and it does so in ways that individual organisms usually do not consciously understand.

If we follow this definition, machine learning systems really are a form of artificial intelligence. They adjust internal parameters based on feedback and experience, just as biological systems adjust their strategies through evolutionary pressure. The mechanism is different, but the functional outcome—adaptation—is similar.

My backyard offers an easy example. I have a tree that grew sideways to escape the shade cast by its taller neighbor. The tree does not have a brain, and no mind was involved, but there was a functional logic: grow toward light. That’s intelligence.

Does the tree have a subjective experience of yearning toward the sun? No. But absence of self-awareness does not negate its capacity to learn and adapt.

Consciousness: The Subjective Experience of Being

This brings us to the second concept: consciousness. Unlike intelligence, consciousness is not about learning or solving problems. It is about experiencing the world as an embodied being.

The neuroscientist Anil Seth describes consciousness as a set of “controlled hallucinations.” Our brains take sensory inputs—sight, sound, smell, proprioception—and weave them into a dynamic model of reality. That model feels real. It feels like us. My awareness of my own existence, the sense that I exist as a continuous self, is something the brain constructs, updates, and sometimes gets wrong (consider anesthesia, dreams, psychosis, or dissociation).

Animals which lack human intelligence still possess consciousness. Your cat has memories, fear, curiosity, pleasure, territorial pride. A bat perceives the world through echolocation. A whale has an internal life we can only speculate about. These creatures are conscious without needing to write essays about it.

Trees, on the other hand—while intelligent according to the definition I chose above—show no convincing signs of subjective experience. They communicate chemically with other trees, adapt to threats, and cooperate through root networks, but there is no evidence of anything like awareness, pain, or an inner life.

This distinction matters, because a lot of AI anxiety assumes that as AI becomes more intelligent, it will automatically become more conscious, that consciousness will “come along for the ride,” as Anil Seth puts it. But that’s not how this works.

As Seth notes, consciousness is not a side effect of intelligence. It is tied to being an embodied organism with sensory drives and a need to maintain internal homeostasis. Our sensory and emotional experiences—hunger, fear, warmth, social bonding—evolved to keep us alive because without food, warmth and community we would die. Large language models have none of these and don’t need them. Unlike a human or other animal, you can shut down a computer for an hour or a month. When you turn it back on, it still works. You can modify or erase parts of a computer program, or you can make thousands of identical copies of the same program and have it run on different machines. The software has no body and no need to fear death. LLMs lack the evolutionary pressures that sculpted consciousness in the first place.

Artificial intelligence may one day match or exceed human intelligence in certain domains (and perhaps already has) without ever becoming conscious.

Language: Our Most Misleading Superpower

The third concept—language—is where people get most confused.

Language is a structured system of communication, a way to encode meaning and share it with others. Humans have especially rich language abilities, and one unique trait that even other animals do not possess: we store language outside our brains, using writing. Written language allowed ideas to propagate across time, not just across space. This ability, by the way, only emerged fairly recently. It has been more than 100,000 years since humans developed the ability to communicate using speech, but written language is only about 5,000 years old.

Animals also have languages—cats, birds, whales, bees. Where they differ compared to humans is that their languages do not possess the same fluency and nuance with which to express abstract relations (“this causes that”), recursion (“the man who saw the dog that chased the cat”), counterfactuals and hypotheticals (“if only Trump had never been elected…”). It is possible that some species such as whales possess some of these language abilities. (We don’t understand their languages, so we don’t know.) However, homo sapiens is the only species that can write books and blog posts to preserve cultural knowledge. The languages of other species convey information, but they don’t build civilizations that span continents and millennia.

Our language abilities give us a huge advantage compared to other species, and our reliance on language is so heavy that it may distort our perceptions. Because language is so central to what makes us unique, we tend to assume that anything that uses language must be intelligent, and must also be conscious.

The philosopher Descartes famously argued, “I think, therefore I am,” but what he really meant was, “I can speak, therefore I’m conscious, and animals are not.” He mistook linguistic sophistication for both intelligence and self-awareness. “The reason why animals don’t speak as we do is not that they lack the organs but that they have no thoughts,” he wrote, adding, “they act naturally and mechanically, like a clock that tells the time. … If they thought as we do, they would have an immortal soul as we do.”

Paradoxically, this same assumption occurs in AI critic Emily Bender’s use of the term “stochastic parrots” to disparage large language models. Bender argues that chatbots merely perform a statistical trick, using probabilities to predict what the next word should be when they generate sentences. Unlike humans, who are embodied and conscious, this makes them “parrots,” mimicking the form of words without understanding their meaning. It is probably true that parrots don’t understand the meaning of human speech when they imitate it. But parrots are conscious. They have the same embodied brains that humans do—gray-matter neurons with dendrites and axons packed into a cerebrum and a cerebellum.

No serious neuroscientist today would give much credibility to Descartes’ belief that non-human animals are merely “meat machines.” Other species may not have the same language capabilities as humans, but their brains are organically similar to ours. Parrots, elephants and dolphins think, remember, plan, and feel—even if they don’t do it in French or Latin.

On the other hand, large language models are able to produce linguistic behavior that resembles utterances from human beings. They use language so fluently that we have to fight to not see ourselves reflected in them. But linguistic fluency is not the same thing as thought or consciousness.

Machine learning was used to train LLMs, but the chatbots which use those language models do not always need to be intelligent in order to be useful. For security and privacy reasons, in fact, many people have started using chatbots under contractual terms which expressly promise that the LLM will not use its conversations to train its language model further. You might use a chatbot to translate a message from English into Spanish, to assist in drafting a memo, or to code a piece of software. A language model can learn something from these interactions, but it doesn’t have to, and sometimes you don’t want it to. Unlike a person or a parrot, an LLM can be programmed to not learn. Following the definitions I’ve adopted, therefore, a non-learning chatbot is not intelligent, no matter how good it is at processing language.

Intelligence Is Not Consciousness

At the risk of repeating myself, to call something intelligent means only that it can adapt to information—recognizing patterns, solving problems, and achieving goals. By that standard, a crow using a stick is intelligent. So is DeepMind, Google’s machine learning algorithm which learned to predict protein foldings. The crow and DeepMind both detect patterns and act accordingly.

Consciousness—the subjective experience of being aware—is something else entirely. Consciousness involves self-reflective awareness, emotion, and a first-person perspective. Humans and crows are conscious, but DeepMind is not. We don’t even fully understand how consciousness arises, much less how to reproduce it in silicon.

A large language model like ChatGPT, or a machine learning system like DeepMind, is vastly intelligent in a narrow sense—able to analyze and predict patterns—but there’s no evidence that they feel, perceive, or desire anything.

Language Is Not Thought

Language can be dazzlingly deceptive. Humans evolved to interpret articulate speech as a signal of mind behind the words. That’s why chatbots feel uncannily alive. Yet linguistic fluency is not the same thing as thought. It’s an output, not necessarily a window into awareness.

As Bender and many others have observed, LLMs predict what words are likely to follow others based on statistical regularities in vast text corpora. But the very same regularities are what make human conversation intelligible in the first place. Human thought is embodied and emotional; our words are tethered to experience. Machines’ words are tethered to probability.

So, yes—AI can “speak,” but that doesn’t necessarily mean it “thinks.” Still, language generation is a genuine cognitive achievement of sorts—a product of pattern-recognizing intelligence so refined that it can fool our empathy reflex.

Consciousness Without Language

To see further how distinct these categories are, consider consciousness without language. Animals experience it all the time. A cat or dog may feel love, fear, or jealousy, yet cannot articulate those feelings. Infants do too, before they learn to speak.

Consciousness can exist without words, and words can exist without consciousness. That’s why equating fluency with awareness is a mistake. The mystery of consciousness lies not in syntax, but in sentience.

Putting It All Together

When we keep the three categories straight, many debates about AI become easier to navigate:

• Artificial intelligence really is intelligence: these systems learn from data and adapt their outputs. With due respect to David Gerrold, calling it intelligence is not a misnomer.
• Artificial consciousness does not currently exist, and making artificial intelligence more intelligent does not guarantee that consciousness will emerge.
• Artificial language ability can far exceed human performance without necessarily involving intelligence or consciousness at all.

This last point is the one that seems to unsettle writers and artists the most: a system can outperform humans in language use—generating coherent text, summarizing ideas, producing stylistic variations—without understanding anything it says.

That doesn’t make AI “fake.” It makes it different.

As Gerrold notes, every new tool—from oil paint to typewriters to Photoshop—forces artists to renegotiate what counts as legitimate creativity. AI is just the next chapter. The danger isn’t that AI will become conscious and replace us. The danger is that we will misunderstand what it is, what it isn’t, and what we want from the technologies we create.

Sorting out the distinctions between intelligence, consciousness, and language won’t solve every ethical dilemma. But it does give us a clearer vocabulary to describe where we are—and where we’re heading.