We will never know if AI becomes conscious, according to Cambridge University philosopher

While AI rights issues are often linked to consciousness, McClelland argues that consciousness alone is not enough to make AI ethically relevant. What matters is a particular type of consciousness, known as sentience, which includes positive and negative feelings.

“Consciousness would cause the AI ​​to develop perception and become self-aware, but this may still be a neutral state,” McClelland says. “Sensitivity implies conscious experiences, good or bad, which is what makes an entity capable of suffering or enjoying. This is where ethics comes into play,” he clarifies. “Even if we accidentally create conscious AI, it is unlikely to be the kind of consciousness we are worried about.”

For example, autonomous cars that experience the road would be a breakthrough. But ethically, it doesn’t matter. If they start to have an emotional response to their destinations, that’s another thing.

Companies are investing huge sums of money in the pursuit of Artificial General Intelligence: machines with human-like cognition. Some claim that conscious AI is just around the corner, and that researchers and governments are already considering how to regulate AI consciousness.

McClelland points out that we don’t know what explains consciousness, and therefore we don’t know how to test AI’s consciousness. If we accidentally create conscious or sentient AI, we must be careful not to cause harm. But treating what is actually a toaster as sentient when there are actual sentient beings that we harm on an epic scale also seems like a serious mistake.

In the debates over artificial consciousness, there are two main sides, McClelland says. Proponents argue that if an AI system can replicate the “software” (the functional architecture) of consciousness, it will be conscious even if it runs on silicon chips instead of brain tissue.

On the other hand, skeptics argue that consciousness depends on proper biological processes in an “embodied organic subject.” Even if the structure of consciousness could be recreated in silicon, it would simply be a simulation that would run without the AI ​​perceiving consciousness.

In this new study McClelland analyzes the positions of each side, showing how both take a “leap of faith” that goes far beyond any body of evidence that currently exists or is likely to be developed.

LOOK: Severe anxiety and obsessive-compulsive disorder: ChatGPT, Grok, Gemini and other AI tools underwent psychological evaluation

“We don’t have a deep explanation of consciousness. There is no evidence to suggest that consciousness can arise with the right computational structure, nor that it is essentially biological,” McClelland says. There is also no indication of sufficient evidence on the horizon. At best, we are one intellectual revolution away from any viable proof of consciousness.

“I think my cat is conscious,” McClelland exemplifies. “This is based not so much on science or philosophy, but on common sense; it’s just obvious.” However, common sense is the result of a long evolutionary history during which artificial life forms did not exist, so it cannot be trusted when it comes to AI. But if we look at the evidence and the data, that doesn’t work either. If neither common sense nor rigorous research can give us an answer, the logical position is agnosticism. We can’t know, and we may never know.”

McClelland tempers this by declaring himself a fairly hardline agnostic. “The problem of consciousness is truly formidable. However, it may not be insurmountable.” Thus, he argues that the way the technology industry promotes artificial consciousness is more like a brand strategy. “There is a risk that the AI ​​industry will take advantage of the inability to demonstrate consciousness to make wild claims about its technology. It becomes part of advertising, so companies can sell the idea of ​​a higher level of artificial intelligence.”

According to McClelland, this enthusiasm around artificial consciousness has ethical implications for the allocation of research resources.