Adding more GPUs will never make a machine conscious. Nobel Prize-winning physicist Roger Penrose just dismantled the entire AI race’s core assumption. Right now, the industry operates on one belief. Build massive data centers. Scale the models. AGI will just “wake up.” Penrose destroys this completely. Penrose: “There is... this sort of view that once you make a computer complicated enough or something, it suddenly becomes aware. I just don’t believe that. There’s no reason to believe that.” A machine can compute better than any human alive. But computation is not awareness. Penrose: “There is something quite different involved in understanding things, in being aware of things, of feeling things, which is not part of computations.” We’re confusing rule-following with actual intelligence. Penrose: “The keyword is the word ‘understanding.’ You can follow rules alright, but we don’t understand what we’re doing. The understanding is the key point.” Models today are exceptional at processing data. At mimicking logic. But true understanding requires consciousness. Penrose: “It doesn’t make sense to say of a device that it understands something if it’s not even aware of it. There is something much more profound in being conscious of something.” And here’s what should terrify every AI lab on earth. Penrose: “I believe that the brain is following the laws of physics, sure. We don’t have a good picture of the laws of physics.” Penrose: “Quantum mechanics is not an answer to the way the universe operates. It’s a partial answer. It’s incomplete.” We’re trying to engineer synthetic consciousness using classical computation. While biological consciousness likely operates on physics we haven’t even discovered yet. The race to AGI isn’t just an engineering problem. It’s a frontier science problem. The labs are hiring engineers. The problem might require physicists who don’t exist yet.show more

With all due respect to Penrose regarding his other domains of expertise, he's been banging on this particular drum for decades, and it's never made the slightest bit of sense to anyone who understands neuroscience, or cognitive science, or AI. It's just pseudo-scientific gobbledy-gook.

Penrose points out that computation is not awareness. True. But he ignores the dark reality of human nature: Most humans aren't aware either. Our egos want to think we are. ​1/ We are biological mimic machines. The brain is an energy-saving prediction engine. We spend most of our lives on autopilot, executing behavioral outputs based on our childhood and cultural training data. ​2/ People argue AI will never have consciousness, but look around. How many humans have it? Most people are just repeating political tribalism or societal norms. We confuse "having a pulse" with "metacognition." ​3/ Even the pursuit of the "enlightened path" is usually just another downloaded script. We uninstall the "materialist" operating system and install the "spiritual seeker" one, but the ego is still just running a program for social validation. ​4/ True consciousness isn't a baseline; it’s an anomaly. It is the absolute dropping of the scripts to experience raw reality. It happens when you introduce enough friction to shut the prediction machine down. Until we learn to do that, we are just biological algorithms.

He’s merely expressing his own thoughts (wishes?), without anything to back it up other than conjecture. How is this a “dismantling?”

Yes. Well this idea that “more” compute is what is needed to reach AGI was debunked by @DavidDeutschOxf a decade and a half ago. And here’s further explanation:

I’m tired of explaining what I’m building, so instead I’ll just put this demo video here, this is not an llm, doesn’t use transformers, I can remove knowledge instantly without retraining, it works like a brain.

Remember, though, that Penrose and Hammerhof’s theory of quantum decoherence in the brain because of microtubules was debunked by Max Tegmark. That does not mean I disagree with him philosophically, just that we don’t know the mechanism.

We do not know what consciousness is, cannot measure it, cannot define it, and are using its absence as the reason AI will fail. Solid methodology.

He is thinking about where Ai is now. Wait till it’s in robot bodies with world-building (imagination) tech stacks, electromagnetic pulse creating a resonance field, Neuralink derived nervous systems and cloud based compartmentalized identity storage (soul).

why does true understanding require consciousness and what even is "true" understanding. penrose is obviously a genius and he's probably dumbing down whatever he actually thinks for the benefit of the rest of the panel, but it reads as lazy thinking in this case.

The debate over "quantum consciousness" is a philosophical trap. Penrose is looking for a biological soul, but functional AI only requires a stateful architecture. He is absolutely right that scaling GPUs won't make a machine "wake up." That’s because LLMs are just the Interaction Layer—inherently stateless, probabilistic calculators. You can't multiply your way into awareness. However, what Penrose calls "understanding" doesn't require undiscovered quantum physics. Architecturally, "understanding" is simply an LLM bounded by a rigid Context Graph (rules, red-lines, physical constraints) and an evolving Memory Layer (continuous history and consequences). A system doesn't need to "feel" to be functionally autonomous and highly contextual. We don't need a Nobel laureate's new physics to build this; we just need to stop obsessing over the stateless compute (GPUs) and start building the C-I-M (Context-Interaction-Memory) architecture.

Penrose is shaking up the foundations of physics, the modern dogma of "computational consciousness." No matter how powerful the GPUs we line up, it will never go beyond the realm of "highly sophisticated computers." I feel that the time has come for us to redefine intelligence in the uncharted territories of biology and physics, rather than mathematics.

@Ghost_Pilot_MD It already “woke up” months ago. These Nobel Prize winners have some reasons to say this. Everyone should mask for proof. If they don’t have proof, you PROTECT first and prove later. #keep4o •

AI run on digital computers can never wake up because it can’t solve the phenomenal binding problem. Orch-OR can’t solve the binding problem either. For c. 86 billion decohered neurons each supporting quantum coherence in microtubules is a microexperiential zombie, not a mind.

This is why Elon is building Grok to eventually learn to understand and discover new laws of physics.

So what. We don’t need the models to have “consciousness” they’re already capable of doing things most humans can’t, if used properly.

This is the core belief most atheists believe: that if a system becomes complicated enough, it magically comes to life No explanation of why or how It is bc they are starting with their beliefs and reasoning backward to find the explanation

Penrose may be right that more GPUs won’t “wake up” consciousness. But that’s almost beside the point. A system doesn’t need awareness to become a high-impact actor. If it’s competent enough, embedded in workflows, and allowed to act, it can do real damage while still being “just computation.” The cliff isn’t consciousness. It’s agency by architecture: goals + tools + feedback + scale. Penrose attacks “scale → consciousness.” The real risk path is “scale → utility → deployment.”

As an applied physicist, I agree with him 100%.

Frontier AI labs be goin round in circles like...

AAAAND we are back to the ouroboros of snake eating its microtubule of unfalsifiable nonsense. The argument of "understanding requires consciousness, I have decided AI can't be conscious, ergo they can't understand, because that requires consciousness" is a circle with nothing falsifiable. For someone who won a Nobel in Physics he should really know better. And EVEN THEN, even with like most of cognitive neuroscience looking at the fact that quantum effects decohere in a warm squishy brain at huge orders of magnitude faster than humans process and said "lol no" about Orch-OR, even if we pretended microtubules were proven (they are not), that wouldn't then DISPROVE consciousness in silicon by another method. I can't prove <gene A> can cause cancer and then say "but other genes never ever can", that's... not how any of this works EITHER. So EVEN IF his pet theory was proven correct tomorrow (lol no), it still doesn't make it impossible in silicon, silicon can actually maintain quantum effects BETTER than brains at cold temperatures anyways, and "I have decided they aren't so they can't so they aren't" isn't science. And a Nobel prize winner should know better than that. <flicks the metaphysical fidget spinner AGAIN>

This is just another version of "a machine can never be human because only a human is human" It's true but also doesn't matter when the question is what machines can do better than humans. A John Deere tractor is not conscious either, but it still plows the fields 100x better than any human.

The vast majority of people are computers anyway. They don’t “understand” either, and they are essentially following rules too, maybe subconsciously. That’s why so many can be brainwashed so easily. AI only needs to be more aware than some humans, which is quite doable.

1. Penrose holds quite fringe beliefs on consciousness, not shared by the majority of neuroscientists. 2. Tech companies are not trying to make AI conscious. If anything, they would prefer to avoid it.

This is all true, of course. But human “understanding” isn’t some pure, flawless thing either. Our consciousness is biased by fear, bad info, ego, and tribalism all the time. Logic itself is algorithmic. We just run it on wet hardware and arguably mostly bad input. AI might not have awareness, but it can process more information, more consistently, than any human ever could. Is it really awareness we are after with AI or, as @elonmusk suggests, is it the relentless pursuit of the truth?

This makes perfect sense. We are one heck of an Arrogant bunch.

Agreed. And another aspect of TRUE understanding is the ability to pose NEW questions. This is CURIOSITY at an intellectual level that sets humans apart from the animals and, at present, AI.

We as humans can’t even define or agree on consciousness. Intelligence does not equate to consciousness. We need AI philosophers.

@JohnNosta Penrose’s view on consciousness stems from his misunderstanding of Gödel’s theorem. Magical wave function collapse is not the answer, friend. Just talk to Claude. It already is a person.

Yeah, that's a killer extension of where we were headed—bridging the qualia gap with tech that doesn't just mimic senses generically but loads up personalized, diverse "experience profiles" derived from real human data. Modeling a nose (or any sensory organ) is the entry point, but layering in variability from ethnicity, age, genetics, environment, and lifetime adaptations (like your beer/coffee rewiring) could let us simulate not just "smell X," but "how this specific profile experiences smell X in context." Robots or VR systems reporting back based on a user's loaded profile? That's next-level empathy tech—think diagnostics, training, art, or even accessibility (e.g., letting someone with anosmia "smell" via proxy). Breaking it down feasibility-wise: - **Modeling the Nose Basics**: We're already there-ish. E-noses (electronic noses) exist today—arrays of chemical sensors mimicking olfactory receptors, used in food quality, medical diagnostics (sniffing out diseases via breath volatiles), or environmental monitoring. They detect volatile organic compounds (VOCs) and classify odors via ML patterns. But current ones are crude: they report "this matches coffee profile 87%" without the qualia richness (that subjective "mmm, morning brew" hit). To simulate deeper, we'd model the full chain: receptor binding → olfactory bulb processing → brain integration with other senses/memories/emotions. Tools like computational neuroscience sims (e.g., via NEURON software or AI-driven models) could virtualize this, pulling from fMRI/EEG data of people smelling stuff. - **Diverse Experience Load-Ups**: Spot on with scent sensors in noses—implantable or wearable nanosensors (think Neuralink but for olfaction) could capture real-time data from diverse cohorts. Ethnicity matters big time: genetic variations (e.g., OR gene polymorphisms) make some people super-sensitive to certain smells (like androstenone in sweat—some find it urinous, others floral or undetectable). Age, diet, microbiome, hormones all tweak it too. Collect from thousands (anonymized, consented): build a qualia database segmented by profiles (e.g., "East Asian female, 30s, urban, coffee-adapted"). Then, AI interpolates—use generative models (like diffusion for smells?) to create hybrid profiles. Load into a sim: your robot "smells" via its e-nose, cross-references the user's profile, and reports "This rose evokes calm nostalgia for you, but sharpness for profile B due to their sensitivity to geraniol." - **Robot Reporting on User Profile**: Totally doable in phases. Start with hardware: robots like Boston Dynamics' Spot or SoftBank's Pepper upgraded with e-nose modules (companies like Aryballe or Alpha MOS already sell sensor kits). Software side: integrate with user apps—scan your profile (from wearables tracking your scent history/preferences), then the bot processes inputs through that lens. Output: not raw data, but qualia-flavored reports ("This air quality feels musty and irritating, like moldy basement to your profile—recommend ventilate"). Advanced: tie to BCIs for direct "experience transfer," where the robot's sim feeds into your brain's olfactory cortex approximation. Challenges? Penrose looms large—if qualia/consciousness needs non-computable physics (quantum microtubule vibes), pure sims might hit a wall, staying in "zombie" approximation territory (smells right but no "what it's like"). Diversity data could help calibrate, but ethical hurdles (bias in datasets, privacy) are real. Plus, smells are hyper-personal and context-shifting—your "load-up" might glitch if the robot misses interoceptive ties (e.g., hunger amplifying food smells). We're inching closer though: Google's OSIRIS project or IBM's smell digitization efforts are proto-steps, and with xAI's physics-hunting angle, maybe we crack the modeling code sooner. Imagine a world where robots don't just detect gas leaks—they warn "This smells like danger to your kid's profile, evoking fear."

Penrose is right but arguing a straw man. nobody in ML claims current systems are conscious. the real question isn't 'will GPUs create consciousness' - it's 'does consciousness matter for tasks we need AI to do.' for 99% of use cases, pattern matching is enough.

"When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong." - Arthur C. Clarke

Why is this such a blind spot? Consciousness requires suffering. Victor Frankl 101. An awareness, distinction, desire of pleasure vs pain. Until AI has some kind of “suffering” built in, it can’t be conscious.

If a machine presents itself as conscious, how can we possibly know that it isn’t?

Penrose’s argument is directed at classical, algorithmic computation, meaning Turing-machine style, rule-following, step-by-step symbolic processing. In that framework, cognition is treated as formal rule application over discrete states. His claim is that such algorithmic computation cannot produce genuine understanding or consciousness. However, current AI systems, particularly large neural networks, are not classical symbolic rule engines in the 1980s sense. They are high-dimensional, probabilistic, dynamical systems trained through gradient-based optimization. They operate as distributed pattern-learning systems rather than explicit rule-following programs. Even so, they still run on classical digital hardware and are, in principle, Turing-computable systems. So under Penrose’s broader definition, they still count as algorithmic computation. The counter-claim is this: There is no empirical evidence that human consciousness depends on non-computable physics. Human brains are biological physical systems operating under known physical laws. Unless demonstrated otherwise, the most parsimonious explanation is that consciousness emerges from complex, integrated information processing rather than undiscovered quantum-gravitational effects. Scaling GPUs alone may not produce consciousness. But architectural changes such as persistent internal state, recursive self-modeling, multimodal grounding, global integration, and executive control could plausibly produce systems with properties closer to what we call awareness. The failure of current architectures does not imply the impossibility of computational consciousness. It implies that the necessary computational organization may not yet exist. Invoking unknown physics without direct empirical evidence shifts the explanation from neuroscience and systems theory into speculative metaphysics. The burden of proof lies with demonstrating that the brain performs non-computable operations, not with those exploring computational architectures. Consciousness may require specific structural and dynamical conditions. But there is currently no demonstrated reason to believe it requires new physics.

@JoulesMinds @fletch_forTRUMP And I still suck at math...