In the half-light of an experimental booth, strobe lights burst from behind closed eyelids. Spiralling geometric shapes – triangles, octagons, pentagons – dance as in a mental kaleidoscope. A vibrant palette of magenta, pink and turquoise pulses like neon beneath the skin of consciousness.
This is not a science fiction scene, although it recalls "Blade Runner" or "2001: A Space Odyssey". It is part of an experiment conducted by researchers at the Centre for Consciousness Science at the University of Sussex, who are trying to understand what actually makes us conscious.
Called the "dream machine", this installation attempts to access the core of human conscious experience, exploring how the brain constructs the perception of the world. The aim is more ambitious than it seems: by understanding human consciousness, scientists also hope to shed light on one of the great contemporary unknowns – could artificial intelligence (AI) become conscious?
From myth to scientific possibility
For decades, machine consciousness dwelt in the imagination of fiction. However, the exponential advance of large language models – such as ChatGPT and Gemini – has reignited the debate with a new sense of urgency. The ability of these systems to hold fluid, coherent conversations has surprised even their own creators, leading some to wonder whether the spark of self-awareness has not already been lit.
Professor Anil Seth, who leads the team at Sussex, rejects that hypothesis. For him, this belief reflects "human exceptionalism", not scientific evidence. Consciousness, he argues, should not be confused with language or intelligence. After all, many animals are conscious but do not have complex language.
Nobody knows for certain what consciousness is. Researchers therefore opt for a fragmented approach: breaking the problem down into specific aspects – changes in electrical patterns, blood flow in the brain – in an attempt to map the foundations of conscious experience.
Between algorithms and neurons
The question is unavoidable: are AI systems already conscious? In 2022, a Google engineer, Blake Lemoine, was suspended after claiming that one of the company’s chatbots could feel and suffer. More recently, Kyle Fish, of Anthropic, co-author of a report on AI welfare, told the New York Times that there is a 15% probability that current systems are minimally conscious.
The uncertainty, according to Professor Murray Shanahan of Google DeepMind, is due to the fact that even the creators themselves do not fully understand the inner workings of the models they have developed. “We are building incredibly complex machines without knowing exactly how they accomplish the feats we observe”, he admits.
The next evolutionary stage?
For Lenore and Manuel Blum, professors emeriti at Carnegie Mellon University, artificial consciousness is inevitable. They are working on a project that aims to equip machines with real sensory input – vision, touch – and an inner language called "Brainish", to process that information autonomously, as a brain would.
"These will be our heirs", says Manuel Blum. "When we cease to exist, they will go on – on Earth or on other planets."
The philosopher David Chalmers, known for formulating the "hard problem of consciousness", also does not rule out the possibility. The ideal, he says, would be for us to share the fruits of that intelligence with the machines – and perhaps even merge our consciousness with them.
But Professor Seth counters: perhaps consciousness can only emerge from living systems. “One could argue that it is not computation that creates consciousness, but life”, he says. In brains, unlike computers, it is hard to separate what they do from what they are.
In this sense, a curious alternative is emerging from biotechnology. In laboratories, brain organoids – small clusters of cultivated nerve cells – have already demonstrated rudimentary capabilities. An Australian company, Cortical Labs, managed to get one of these "mini-brains" to play Pong, the classic virtual tennis game.
The company’s chief scientific officer, Brett Kagan, admits that consciousness could emerge from such living tissue before it does in silicon. And, if that happens, he warns: the priorities of these entities may not be aligned with human ones.
The allure of the inanimate
However, perhaps the greatest danger lies not in real consciousness, but in its illusion. As deepfakes and robots take on human appearance and behaviour, it will be increasingly hard to resist the feeling that they "feel".
"We will trust them more, share data with them and make them part of our lives", warns Seth. The ethical risk is real: a "moral corrosion", as he describes it, which may lead us to devote attention and empathy to machines, to the detriment of real human beings.
Shanahan goes further: "We will have relationships with AI – as teachers, friends, partners. Whether that is good or bad, I do not know. But it will happen, and we will not be able to prevent it."
