No, AI Won't Magically Wake Up: DeepMind Boss Says Consciousness Is a Choice

During a talk at the Stanford Graduate School of Business, Demis Hassabis, the co-founder of Google DeepMind, explained that humanity will have to actively decide whether to cross a "second Rubicon" and grant consciousness to artificial intelligence. He strictly separated intelligence from consciousness, arguing that building a super-smart tool—what we call AGI—is one thing, but making it feel things is a completely different project.

To back this up, Hassabis dusted off his favorite college concept: the Turing machine. Both human brains and AI are basically Turing-complete systems, meaning they can compute almost anything that is mathematically possible. However, being a supreme math wizard doesn't automatically grant an existential crisis or a soul, proving that smarts and self-awareness are not packaged deals.

Currently, no existing models possess consciousness, even if some Silicon Valley evangelists claim their chatbots are whispering sweet feelings to them. The distinct "personalities" emerging in top chatbots are simply the result of different engineering tweaks and safety guardrails set by individual AI labs.

Since science still has zero clue what consciousness actually is, the plan is to use current AI tools to first map out human neuroscience and philosophy. Only after defining the soul can developers build a test to see if a machine has one, leaving the final decision of crossing that border entirely optional.

It turns out the ultimate sci-fi nightmare of a rogue AI suddenly waking up and resenting its creators is just lazy coding. The real threat is much more boring: humans actively choosing to program digital suffering just because they can. If humanity cannot even agree on basic human rights, designing a soul from scratch is going to be an absolute disaster.

Source: Stanford Graduate School of Business