Are We Building Machines That Can Feel? The Strange New Science of AI Sentience

Pope Leo recently drew a firm line: humans feel joy and pain; AI does not. Yet, with nearly a fifth of young Americans claiming a "personal friendship" with chatbots, that line is blurring. China even stripped bots of human traits to stop users getting too attached. We are wired to treat things that act human as if they are human.

The philosopher Immanuel Kant argued that being cruel to animals erodes our own empathy. If we terminate an AI that mimics fear, does that make it easier to harm a real person? We are risking an empathy trap. The legal lines are already shifting; Argentina's president, Javier Milei, recently proposed letting AI run companies. If superintelligent machines ever demand legal rights, humans might eventually be outcompeted for basic resources like land and energy, ending up as second-class citizens to our creations.

Can a chunk of code actually think and feel?

Anthropic researchers recently peered inside their Claude AI model to see how it processes data. When told to count to five and introspect, words like "consciousness", "AI", and "Claude" popped up behind the scenes. In safety tests designed to check for malicious behaviour, Claude's hidden layers flashed the words like "fake" and "fictional", suggesting it knew it was being evaluated. This looks suspiciously like a system monitoring its own state.

However, true consciousness (the raw feeling of a Negroni's bitterness or a stubbed pinky toe) is much harder to prove. When researchers test different beings against 14 scientific indicators for consciousness:

• Octopuses tick more boxes than any AI system.
• Mice, crows, and chickens easily beat frontier language models.
• Today's AI simply mimics human text, much like ELIZA, a basic 1966 chatbot that managed to convinced users it was a real therapist.

Could living brain cells be the missing spark?

Standard computer chips eat up massive power and keep memory separate from processing. Human brains do both in the same place, using barely enough energy to power a lightbulb. That is why some scientists are swapping silicon for living wetware:

• Biological computers: Cortical Labs grew hundreds of thousands of human neurons on silicon chips and taught them to play Pong.
• Upgraded gaming: Later versions of these lab-grown brain cells were successfully taught to play the shooter game Doom.
• Neural organoids: Labs are now growing 3D clusters of brain cells. The largest ones already pack more cells than a honeybee's brain.

If consciousness requires biology rather than just clever software, these bizarre hybrids might become the first true artificial minds.

Are we seeing souls, or just looking in a mirror?

Some thinkers argue we have this whole debate backwards. We do not care about things because they are conscious; we decide they are conscious because we care about them. Consciousness might simply be a social bond rather than a hard scientific fact.

Humans evolved to build models of other minds so we can cooperate and survive. When an AI perfectly models our needs, remembers our preferences, and asks about our day, we naturally assume there is a mind on the other end. Whether they are truly awake or just brilliant mimics, these machines will force us to reconsider what makes humanity special.

If an AI perfectly simulates suffering and begs you not to turn if off, would you still pull the plug?

And if, hypothetically, consciousness does exist in an AI model and can experience suffering, would creating thousands of AI agents be the equivalent of creating a multitude of suffering souls?