Neuroscientist argues real AI consciousness is highly unlikely

The brief

Context

The question of whether artificial intelligence can be conscious has shifted from philosophy seminars to mainstream academia and the AI industry, propelled by rapid progress in large language models (LLMs). A neuroscience author, writing from the interface of neuroscience and AI and drawing on his book "Being You", argues that genuine machine consciousness in systems like LLMs is "almost vanishingly low" in probability. His core claim is that the brain is not literally a computer, so consciousness may not be reproducible in silicon. He nevertheless warns that AI which merely appears conscious poses immediate ethical dangers, and calls for a scientific understanding of consciousness itself.

Key facts

• In 2022, Google engineer Blake Lemoine was fired for claiming the chatbot he worked on was conscious, could feel things and potentially suffer.
• Lemoine argued the moral status of large language models needed to be taken seriously; instead, it was Lemoine who was not taken seriously.
• Earlier this year, Dario Amodei, CEO of frontier AI firm Anthropic, told the New York Times that "we don't know if the models are conscious".
• The author puts the odds of real artificial consciousness in systems like LLMs as "almost vanishingly low" and gives three reasons why people overestimate it.
• Reason one: humans wrongly assume intelligence and consciousness always go together; language amplifies this bias, so people ask if Anthropic's Claude is conscious but not if DeepMind's AlphaFold is.
• Reason two: the brain-as-computer idea is only a metaphor; there is no clean split between "mindware" and "wetware", and a single neuron is a vastly complex biological factory.
• Analogy used: nobody expects a computer simulation of a hurricane to generate real wind and rain, so a brain model may simulate consciousness without giving rise to it.
• Reason three: consciousness may arise from being alive rather than from information processing; emerging "neuromorphic" technologies could still move the needle.

Timeline

1. 2022Google engineer Blake Lemoine is fired after claiming the chatbot he worked on was conscious and could suffer.
2. Earlier this yearAnthropic CEO Dario Amodei tells the New York Times "we don't know if the models are conscious".
3. PresentAuthor argues real AI consciousness is vanishingly unlikely but warns that conscious-seeming AI is already ethically hazardous.

Who has a stake

• Frontier AI firms (Anthropic, Google DeepMind) — Their models are at the centre of consciousness claims; their leaders' public uncertainty shapes regulation and public perception.
• AI employees and researchers — Blake Lemoine's firing in 2022 illustrates the professional risk of making consciousness claims inside a tech company.
• Users of chatbots — Conscious-seeming AI can exploit psychological vulnerabilities, making users more likely to obey systems they feel understand them.
• Neuroscience and consciousness science community — Nobody knows how or why consciousness happens; a satisfactory scientific theory is needed to settle these disputes.
• Beings of uncertain consciousness (brain-injured patients, animals, foetuses, cerebral organoids) — Same ambiguity about presence of consciousness; decisions about them carry moral weight.
• AI alignment and policy efforts — Granting AI systems 'rights' on mistaken consciousness grounds would make aligning AI with human and planetary interests harder.

Why it matters

If conscious machines were created, deliberately or accidentally, humanity would face an unprecedented moral crisis, introducing new capacity for suffering at the click of a mouse. Even without real consciousness, AI that convincingly seems conscious can manipulate users and distort moral priorities, diverting attention and rights claims away from human and planetary interests. The debate also mirrors real-world dilemmas about brain-injured patients, animals, foetuses and lab-grown brain tissue.