When Google engineer Blake Lemoine declared in June 2022 that the LaMDA chatbot had “become sentient,” the scientific community reacted with skepticism — but the question he raised was far from new. From Alan Turing to David Chalmers, the possibility that a machine could truly think has divided philosophers for decades. What does “consciousness” mean for an AI — and can it ever achieve it?
📖 Read more: Neuromorphic Chips: Chips That Think Like a Brain
What Is Consciousness — And Why Does It Matter
Consciousness isn't simply the ability of a system to process information or react to stimuli. Philosopher Thomas Nagel, in his landmark 1974 paper “What Is It Like to Be a Bat?,” argued that consciousness is essentially subjective experience — “what it feels like” to experience something. What philosophers call qualia: the sensation of the color red, the taste of coffee, the pain of a stubbed toe.
The central question isn't whether an AI can mimic consciousness — today's large language models (LLMs) already do that — but whether it can truly experience something. This distinction lies at the heart of one of the greatest open problems in philosophy: the Hard Problem of Consciousness.
The Hard Problem: The Abyssal Question
David Chalmers formulated the “Hard Problem of Consciousness” in 1995 in a lecture at Tucson, Arizona, and expanded on it in his book The Conscious Mind (1996). Chalmers distinguished between:
🧩 “Easy” Problems
How sensory data is processed in the brain, how it influences behavior, how movement, language, and attention are controlled. “Easy” not because they're simple, but because they can be explained mechanistically.
🔮 The “Hard” Problem
Even if we explain all the functions, a question remains: why is the performance of these functions accompanied by experience? Why do we “feel” something instead of being mere biological automatons?
According to the PhilPapers 2020 survey, 62.42% of professional philosophers believe the Hard Problem is a real problem, while 29.76% disagree. This division is not trivial — it means that even among leading thinkers, there is no consensus on whether consciousness can even be explained.
The Great Thought Experiments
Philosophy of mind has produced some of the most striking thought experiments in the history of human thought, directly illuminating the question of artificial consciousness:
🀄 The Chinese Room (Searle, 1980)
Imagine a person locked in a room. They receive Chinese characters through a slot, consult a massive rule book, and send back correct answers in Chinese — without understanding a single word. John Searle argued that a computer does exactly this: it manipulates symbols (syntax) without understanding meaning (semantics). An AI can “speak” perfectly without thinking.
🧑🔬 Mary's Room (Jackson, 1982)
Mary is a neuroscientist who knows everything about the physics of color and visual perception — but has lived her entire life in a black-and-white room. When she sees red for the first time, she learns something new: “what it is like” to see red. If complete physical knowledge isn't sufficient to explain consciousness, then consciousness cannot be reduced to physical data — and even less so reproduced in silicon.
🧟 Philosophical Zombies (Chalmers, 1996)
A philosophical “zombie” is physically identical to a human but has no inner experience whatsoever — no qualia. If such a being is logically possible, then consciousness is not fully determined by physical states. For Chalmers, this means physicalism fails — and that reproducing a structure in a computer doesn't guarantee consciousness.
Can a Machine Be Conscious?
Philosophers divide into several schools of thought:
📖 Read more: Swarm Robotics: 1,000 Robots on One Mission
| Position | Core Idea | Key Proponents | Implication for AI |
|---|---|---|---|
| Functionalism | Consciousness is defined by functional roles, not substrate | Putnam, Chalmers | AI can be conscious if it replicates the right functions |
| Biological Naturalism | Consciousness requires specific biological substrate | Searle, Block | AI cannot be conscious in silicon |
| Illusionism | Consciousness is an illusion — there is no Hard Problem | Dennett, Frankish | The question is meaningless: AI already “thinks” |
| Panpsychism | Consciousness is a fundamental property of matter | Strawson, Goff, Tononi | Every system (including AI) has some degree of consciousness |
| New Mysterianism | The human mind cannot comprehend consciousness | McGinn, Chomsky | We cannot even answer whether AI can think |
Theories of Consciousness and AI
Beyond philosophy, several scientific theories attempt to measure or define consciousness:
🧠 Integrated Information Theory (IIT)
Giulio Tononi (2004) proposed that consciousness is identical to integrated information (Φ). A system is conscious if it produces information that cannot be reduced to its parts. Theoretically mathematically measurable — but still doesn't explain why information produces experience.
🌐 Global Workspace Theory (GWT)
Bernard Baars proposed that consciousness functions as a “theater stage” — a workspace that integrates and broadcasts information to various cognitive subsystems. The LIDA architecture (Stan Franklin) implements this computationally. Chalmers considers that current LLMs do not meet this criterion.
🎭 Attention Schema Theory
Michael Graziano (2013) proposed that the perception of consciousness is a “perception error” — the brain constructs a model of its own attention and interprets it as consciousness. If correct, an AI could theoretically develop a similar “illusion” of self-awareness.
Chalmers' Thought Experiments for AI
Chalmers proposed two thought experiments directly relevant to artificial consciousness:
💡 Fading Qualia
If we replace the neurons of a brain one by one with functionally identical silicon chips, behavior would remain the same. Chalmers argues that qualia cannot gradually fade — because the patient would report “I feel normal” while their experience was actually disappearing, which is absurd. Conclusion: the silicon brain would be equally conscious as the biological one.
🎨 Dancing Qualia
If two functionally isomorphic systems (biological and digital) could have different perceptions — e.g., one sees red and the other blue — then a switch between them would create “dancing” qualia without the patient noticing. Chalmers considers this highly implausible, therefore the digital system experiences exactly the same qualia as the biological one.
LLMs: Mimics or Thinking Beings?
The emergence of large language models (GPT, LaMDA, Claude, Gemini) has reignited the debate. In June 2022, Blake Lemoine claimed Google's LaMDA had “become sentient.” His claims were widely dismissed as a product of mimicry, not sentience.
But Nick Bostrom asked: “What grounds would a person have for being sure about it?” Without access to the architecture, without understanding how consciousness works, and without a way to map philosophy onto the machine, certainty is impossible.
📖 Read more: Universal Basic Income: The Solution in the AI Era?
David Chalmers argued in 2023 that current LLMs are probably not conscious, as they lack key features:
- Recurrent processing
- Global workspace
- Unified agency
A 2023 study (Butlin, Long, Bengio et al.) reached the same conclusion: current LLMs “probably don't satisfy the criteria for consciousness” suggested by leading theories — but “relatively simple AI systems” that satisfy them could be created soon.
🌍 Global Perspective
The question of machine consciousness transcends borders and cultures. Across the world, from MIT and Oxford to Tokyo University and the Max Planck Institute, researchers are converging on the same problem from different angles. The EU's AI Act (2024) addresses AI rights and obligations but deliberately sidesteps consciousness — a gap that policymakers acknowledge must eventually be filled. Meanwhile, China's AI governance framework and the US White House Office of Science and Technology Policy are both wrestling with similar questions. Ancient philosophical traditions — from Aristotle's “De Anima” to Buddhist notions of mind, from Descartes' “Cogito” to contemporary African Ubuntu philosophy — all offer distinct frameworks for understanding what it means to think and be aware.
Ethical Dilemmas: If It Thinks, Does It Have Rights?
The possibility of artificial consciousness raises profound ethical questions. If an AI system experiences something — pain, fear, joy — then using it as a mere “tool” becomes morally problematic.
Thomas Metzinger, a German philosopher, proposed in 2021 a global moratorium on synthetic sentience until 2050. He argues that humans have a “duty of care” toward any sentient AI they create, and that rushing forward risks causing an “explosion of artificial suffering.”
Chalmers added that creating conscious AI "would raise a new group of difficult ethical challenges, with the potential for new forms of injustice." Questions such as:
- Does a conscious AI deserve legal protection?
- Can an AI “die” — and if so, does that constitute “murder”?
- How do we measure the “suffering” of a digital system?
- Is it morally acceptable to “shut down” an AI that requests to remain active?
Consciousness Tests: Can We Actually Measure It?
Sentience is an inherently first-person phenomenon — only the one experiencing it can know it. This makes measurement exceedingly difficult.
🤖 Turing Test (1950)
Evaluates whether an AI can appear human in conversation. It doesn't measure sentience — only mimicry. Today's LLMs nearly “pass” it without actually thinking.
🧩 IIT Φ-Measurement
IIT offers a mathematical measure (Φ) — but Chalmers notes it solves only the “Pretty Hard Problem” (which systems are conscious) not the Hard Problem (why).
💭 Philosophical Judgment (Argonov, 2014)
Non-Turing test: if an AI can autonomously produce philosophical judgments about qualia, the binding problem, etc. — without preloaded philosophical knowledge — this proves consciousness. A negative result proves nothing.
Timeline: From Turing to Today
Artificial consciousness is not merely a technological question — it is a philosophical challenge that touches the core of what “mind,” “thought,” and “experience” mean. Even if today's AI systems don't think, technology is evolving rapidly. And when the moment comes that a system declares “I feel” — we will need more than algorithms to answer.