Why DeepMind's AGI advances do not answer the key question of machine consciousness

The central debate about AGI is not whether machines will become smarter, but whether they will develop an inner experience. The author argues: digital systems can increase their intelligence, but this does not bring them closer to consciousness.

Protein and Life

The author starts from a grounded premise: rather than speculate about hypothetical forms of intelligence, we should look at what science already knows. All observed forms of life are associated with proteins and cellular organization. This leads back to the classical definition of life formulated by Friedrich Engels as early as 1883.

What matters is not only the protein base itself, but also the constant metabolic exchange without which a system cannot sustain itself as living. Even the viral paradox, usually cited against such an approach, actually supports the argument here. A virion contains protein, but outside a cell shows no signs of life and only begins to behave as living when inside the metabolic system of the host.

From this the author draws a cautious but firm conclusion: protein is likely necessary, though insufficient by itself. We can imagine life on a different substrate, but for now that remains imagination, not confirmed fact. He connects this thesis to the recent AI agenda.

When on October 9, 2024, Demis Hassabis and John Jumper received the Nobel Prize in Chemistry for AlphaFold2, it was not simply a reward for a powerful algorithm. In the author's logic, this is a reminder that the most notable AI breakthrough turned out to be connected not with creating a conscious machine, but with studying proteins — the material foundation of the life we know.

Intelligence Without Experience

The next step is to separate intelligence and consciousness, which in public discussions of AGI are constantly conflated. The author draws on David Chalmers' distinction between the "easy" and "hard" problems of consciousness. Easy problems concern information processing: how a system recognizes patterns, responds to stimuli, plans actions.

The hard problem is different: why subjective experience arises at all, that very qualia without which there is no inner world. Here, according to the author, engineering ends and philosophy begins. Here too we recall a simpler empirical fact: all known bearers of consciousness are living creatures.

We have not a single confirmed example of a computer, stone, or any other inanimate system possessing subjective experience. This is not a strict prohibition or final proof, but a very persistent correlation. Therefore, even if AGI learns to solve more problems, computational power alone does not answer whether the system experiences pain, color, or time.

The author particularly disputes the techno-optimistic reading of DeepMind's statements. When in February 2026 Demis Hassabis spoke of continuous learning, long-term planning, and consistency as properties of future AGI, he was talking about intelligence, not consciousness. The author draws the same boundary through mathematics: a machine can prove, calculate, and iterate through options, but so far shows no intuition, no ability to "see" a beautiful idea before formal proof.

In this sense AI remains a very powerful problem-solver, not a bearer of inner experience.

Why an Analog is Needed

If we nonetheless admit that consciousness could theoretically arise not in a biological brain, then architecture becomes decisive. The author believes digital computers are too far removed from the brain in their basic principles of operation. The brain uses continuous signals, ionic currents, and dense coupling of memory with computation. A classical digital machine works differently. Therefore, he considers direct comparison of the brain with an ordinary computer too crude a simplification:

• discrete states 0 and 1 instead of continuous transitions
• rigid clock synchronization instead of asynchronous biological dynamics
• separation of memory and processor in the von Neumann architecture tradition
• constant data transfer instead of computation where information is stored

From this emerges the paper's key engineering conclusion: if non-protein consciousness is possible at all, it would more likely require analog or neuromorphic systems than ordinary GPUs and TPUs. As candidates, the author lists memristors, photonic chips, and architectures like SpiNNaker2. They share an attempt to approach how the brain actually works. It is no accident that researchers of such systems increasingly speak of ionic rather than purely electronic logic in computation.

"Ions are a better medium for embodying brain principles than electrons."

Against this backdrop, the forecast of imminent AGI ceases to look like an automatic route to a conscious machine. Yes, digital models will get better at writing text, winning games, and assisting in science. But the author's point is different: intelligence can be scaled on digital systems, but consciousness, if it ever arises outside biology, will require a completely different substrate and completely different physics of computation. Power alone, in his view, is not sufficient.

What This Means

For the AI industry, this is an important clarification: even very powerful AGI does not equal a conscious being. The discussion is gradually shifting from the question "how many more GPUs do we need" to "what architecture could ever generate subjective experience," and the market has no convincing answer yet. That is precisely why the debate over AGI increasingly comes down not to software, but to the material foundation of computation.