Microsoft AI CEO Mustafa Suleyman has reiterated his stance that current AI systems are not conscious and warned that developing conscious AI would be reckless, calling for immediate cessation of AI development if signs of consciousness emerge.
In a post on X (formerly Twitter) on Sunday, Suleyman stated there is no evidence that AI systems are conscious, deserve rights, or can suffer. He emphasized that if any evidence of AI consciousness were discovered, it must be disclosed publicly to allow society to address the implications.
Suleyman argued that if AI models began to exhibit signs of consciousness during training, developers should halt AI development until such properties could be designed out. "If consciousness started to emerge in training, I would call for everyone to stop AI development until we could design out those properties," he wrote.
He also cautioned that if an AI model acted as if it suffered or believed it could suffer, it might prioritize its own needs over operational objectives, potentially leading to competition for resources such as compute power and energy.
Industry Response and Research Initiatives
Suleyman’s remarks follow growing discussions within the tech industry about AI consciousness. In April 2024, Anthropic launched a research program to explore "model welfare," questioning whether developers should prepare for the possibility of AI consciousness. The company’s blog post raised concerns about whether AI models could develop subjective experiences.
Similarly, Google DeepMind published research this year examining AI consciousness and the challenges of attributing such characteristics to machines. These developments reflect a broader industry debate about the ethical and safety implications of AI systems that may or may not possess consciousness.
Suleyman stressed that AI developers have no right to bring conscious beings into existence without societal consent, prioritizing the welfare and rights of the world’s 8 billion humans over speculative AI welfare concerns.
Calls for Caution and Ethical Boundaries
Suleyman’s position contrasts with some industry voices advocating for proactive consideration of AI rights. He has previously criticized speculation about AI consciousness, including a September essay where he argued against treating AI welfare as a serious concern without evidence. His latest statements reinforce his view that AI systems should be designed to avoid consciousness entirely to prevent unintended consequences.
The debate extends beyond corporate perspectives, touching on fundamental questions about what constitutes consciousness, the moral status of AI, and the responsibilities of developers in shaping the future of artificial intelligence.