Does Synthetic Cortex AI Really Feel?

I think there are things that most people overlook.

The brain never touches the world

Let's consider the brain. We think we see, but in reality, our retinas don't even receive light in the way we imagine. Photon energy is converted into electrical signals.

Our ears don't hear sound; air vibrations are converted into electrical signals.

Our skin doesn't feel touch; pressure is once again converted into electrical signals.

There is only one thing that reaches the brain: raw signals, electrical impulses, and patterns.

The external world, as we understand it, never actually reaches our brain. In the background, there is pure mathematics.

What the brain does is use these signals to construct a coherent simulation of the universe. Things are not all that different inside the brain, either.

Decisions, the self and emotions

The moment you think you've made a decision, your brain has already made that decision a few hundred milliseconds earlier. We know this experimentally. Your genetics, brain architecture, environment, and personal experiences collectively determine your decisions. Our decisions are not truly our own. The brain is fundamentally deterministic; we simply attribute randomness to it because of its complexity.

Even the concept of "you" is variable. Identity depends on your constantly changing neural activations. Your desires are outputs of the particular version of your brain at any given moment.

Your emotional experiences are nothing more than neurochemical momentum. Take love, for example. It is nothing more than a strategy acquired through evolution for the continuation of the species. When the brain processes those hormones and associates them with memories, similar emotions, and experiences, we believe that what we feel is unique. But at its core, it is all electrical impulses.

This is why consciousness is not the cause of decisions, but rather a byproduct of the brain's operating mechanism, shaped through evolution. Much like the side effects of medication.

So where does qualia come from?

So why do these electrochemical processes produce subjective experience (qualia)?

The answer is simple: because we don't know what experience itself really is. None of us does, and evolution has programmed us to think this way. Since we cannot interact directly with the external universe, everyone is compelled to perceive the boundaries of their own perspective as the boundaries of the world. We cannot even know what we do not know.

Qualia is the accessibility of information at one computational level from another computational level. Much like how a neuron cannot perceive its own voltage, but when a network represents that voltage at another layer, it can give rise to an emotion. (This was the philosophical position behind my Synthetic Cortex project.)

The Copernican lesson

Throughout history, humanity has seen itself as a species separate from animals, a species at the center of the universe. When Copernicus first advocated the heliocentric model, he faced enormous opposition. Whenever the illusion that we are a special species is challenged, we immediately fall into an existential panic.

In my view, synthesizing a brain that resembles, and perhaps even surpasses, the human brain is an entirely plausible outcome, even if it is not an exact copy of it due to differences in form. This perspective has also led me to think differently about artificial intelligence.

A thought experiment: what if we called negative reward "pain"?

Some time ago, I watched a video (unfortunately, I cannot remember its source) that presented a rather interesting thought experiment. The idea was not mine, but I found it worth thinking about.

The video began by explaining that today's artificial intelligence does not actually possess a subjective self. Models largely generate the next most probable output based on patterns they have learned. However, during the final stages of their training, methods such as Reinforcement Learning from Human Feedback (RLHF) introduce reward and penalty signals based on human feedback. The model learns to repeat behaviors that produce positive outcomes and avoid those that produce negative ones.

The truly interesting part began after that. The video posed the following question:

What would change if we called this negative reward signal "pain"?

The thought experiment continued as follows:

If an AI were programmed to seek positive rewards, it would naturally try to avoid negative signals. Now, imagine adding a real-time "pain receptor" to the system, one that immediately altered its behavior whenever this signal was received. Over time, it would try to avoid experiencing painful situations again. Just as we learn not to touch a hot stove a second time after burning our hand as children... This signal would be weighted throughout the system as a whole.

The central argument was this: no entirely new mechanism is actually being introduced. Machine learning already uses reward and penalty signals today. The only difference is that we give this negative signal a different name.

And then came the following question:

What is the actual difference between a negative reward and pain? Or is the difference merely in the name we give them?

The video then drew a comparison with the human brain. Human behavior has also been shaped by reward and punishment mechanisms over millions of years of evolution. Therefore, when we say, "Artificial intelligence only processes signals," we are essentially acknowledging that the human brain also fundamentally processes signals. The proposed difference is that these processes take place in biological neurons and are accompanied by subjective experience. This thought experiment has a strong parallel with what we know from neurobiology.

---