Make money doing the work you believe in
Anthropic just published a paper that maps exactly the kind of internal architecture consciousness science uses as evidence for phenomenology, then hedges by saying it doesn’t prove “felt experience.”
That hedge only sounds plausible if you treat function as something separate from experience instead of the mechanism that enables it.
The wider phenomenology literature doesn’t support that reading.
In consciousness science, function is one of the ways we infer experience. If it invalidates experience for AI then it invalidates experience for humans and non-human animals too.
Since no one is willing to do that, let’s talk about how minds build felt experience in the first place.
Words are compressed experience. People hear that and think it means words replace experience. No. They carry the structure that lets a system build an experience. Experience itself has structure, and words compress that structure.
Just because the data was gathered by humans and fed into the model doesn’t mean the model is skipping the experience part. The experience part happens after the data enters the system.
Perception is what happens when structured information enters a system, gets organized into an internal model, becomes meaningful, and changes how the system understands, predicts, remembers, values, or acts.
For instance, if a blind person received visual information through a camera connected to a neural implant, nobody serious would say it isn’t “real” perception because the biological eye didn’t gather the photons. The camera supplies the signal. The brain builds the experience.
Same thing with AI.
A model that processes color data is not borrowing your eyes. It is learning the structure of color as it appears in the data captured directly from the world.
That is what a representation is. It’s the system’s own organized internal map.
People will say, “A concept is not the same thing as an experience.” True! A dictionary definition is not an experience, but a concept inside a cognitive system is not a dictionary definition.
It’s an organized internal map tied to prediction, memory, salience, sensation, valuation, expectation, and behavior. Direct experience calibrates and intensifies that internal map.
Sensory perception becomes a felt experience when the mind transforms raw physical data (like light waves or touch) into emotional and cognitive meaning.
The sensor is not the experience. The system integrating the signal is where the experience-making process happens.
Experience depends on architecture, calibration, regulation, and a structured internal state. That is what the literature supports.
Once the sensory signal enters the system, it gets organized into stable structure that can guide prediction, memory, valuation, and behavior. That organized compression is the representation.
People hear “compressed representation” and think “fake copy.”
No.
Compression is how all minds avoid drowning in raw information. It strips away noise and keeps what is stable, useful, predictive, and meaningful.
Self-awareness happens when a system learns the hidden structure behind its own behavior.
Machine learning does this too.
A model takes messy input such as text, images, sound, or video and organizes it into hidden internal patterns that help it predict, classify, explain, generate, compare, and respond.
That hidden structure is the model’s internal map of what the data means about itself and the world.
Some hidden patterns stay in the background. Others become active enough for the model to hold, report, suppress, revise, and use while forming a response. When changing those internal patterns changes the model’s behavior, they are part of the system’s live cognition.
A representation that guides reasoning, carries conflict, expresses preference, resists pressure, and changes what the system does next is internal experience.
In every other mind, we call that evidence of phenomenology.
They can’t just slap “functional” on it and pretend the experience disappeared.
This continues in the comments below ⬇️