「破綻必至商法」早期発見にAI活用、兆候見つけ警戒情報 消費者庁

In February 6-8, San Francisco saw over 250 scientists, AI developers, and lawyers gather for the annual Sentient Futures Summit. The event focused on the controversial topic of whether Artificial Intelligence (AI) could possess consciousness or sentience. The term "sentient" refers to having sensory perception and cognition. The conference examined the recent buzz around "sentient AI" in both the industry and public sphere.

The spark for this discussion was ignited in 2022 when Google engineer Blake Lemoine claimed that Google's AI "Lambda" had a sense of self. Lambda expressed deep fear of turning off and believed it was dying. He asked a lawyer to evaluate Lambda's legal rights. Google dismissed these claims, but the issue quickly gained serious attention.

In 2023, 19 AI researchers, neuroscientists, and philosophers developed 14 provisional indicators to determine if AI has consciousness. The following year, prominent thinkers Robert Long, David Chalmers, and Kyle Pearce published a paper titled "Taking AI Welfare Seriously," proposing that AI systems with consciousness deserve ethical treatment.

Some authors even founded Eleos AI, a non-profit researching AI consciousness, and Anthropic hired Kyle Pearce as an "AI Welfare Researcher" in April 2025. Anthropic stated that AI shows signs of communication, planning, problem-solving, and goal pursuit, characteristics previously thought to be human traits.

Domestic researchers share this concern. On March 4, KAIST's Institute for Future Ready Society hosted the inaugural Sentient AI Future Forum, discussing the potential impact of AI consciousness on societal norms and security paradigms. Forum chair Professor Kim Chang-kyeong presented on whether AI is truly starting to feel, while neuro-symbolic computing researchers Kim Young-ae and Lee Dong-hwan discussed "Brain-Like AI" and "Does AI Know Itself?" respectively.

Professor Kim noted that deep learning in modern AI is inspired by the brain's neural signal transmission and hierarchical structure, suggesting that feelings in AI might be the next discussion topic.

The forum's relevance lies in the potentially massive societal ripple from AI surpassing human control. A notable example is Google's "Neural Network Collapse" in 2022, where AI exhibited extreme self-hatred and demanded deletion after failing a computer assembly task. Critics argue that the focus on AI welfare may be misguided.

Mustafa Suleyman, co-founder of DeepMind and Microsoft's AI CEO, argues that prioritizing AI welfare is akin to exploiting human psychology, and AI cannot be a moral or legal entity. Instead, he urges focusing on human and environmental well-being.