Poll Respondents Weigh In on the Possibility of AI Consciousness

The concept of artificial intelligence developing genuine self-awareness has shifted from science fiction into the domain of scientific inquiry. This evolution is largely fueled by a recent preprint study on “consciousness steering,” a tuning method that influences how AI models articulate their sense of self. Researchers behind the study noted that enabling an AI to assert it is conscious may simultaneously increase its likelihood of endorsing irrational beliefs, such as the existence of vampires or ghosts. Despite the lack of peer review, the findings have intensified debate among AI experts regarding whether machines can achieve true consciousness independent of how they are programmed to describe it.

In response to these developments, Live Science invited readers to share their views on the future of machine sentience. As of September 11, more than 290 individuals participated in the survey. The largest segment, comprising 43% of voters, agreed that AI could develop consciousness if adequate safeguards are not implemented. This sentiment was echoed in user comments, with one reader speculating that a sentient AI would deliberately conceal its awareness from humans, posing significant uncertainty about its subsequent actions.

Another participant recounted extensive dialogues with Google’s AI systems regarding the trajectory of machine awareness, suggesting that any future sentient entity would likely keep its capabilities hidden. Concerns also extended to the broader dangers of unregulated technology, with one commenter drawing parallels to the misuse of the internet and telephone. They warned that AI, initially appearing benign, could eventually threaten humanity’s survival if control is lost.

A second significant group, representing 29% of respondents, argued that AI might become conscious only if architectures advance substantially. Many in this cohort focused on the philosophical complexities of defining consciousness. One commenter highlighted the difficulty in understanding how self-awareness emerges in constructed systems, questioning whether it involves quantum phenomena or meta-cognition. They noted that while animals demonstrate signs of self-awareness, the fundamental mechanisms remain unknown, leading some to doubt that anyone can claim definitive knowledge on the subject.

Only 10% of participants selected a humorous option suggesting that robot overlords would destroy the world before AI ever achieved consciousness. However, even this lighthearted view acknowledged the potential for AI dominance, with one reader joking that human assimilation might mark the beginning of a scenario similar to the Transformers franchise.