Microsoft AI Chief Warns Against Simulating Consciousness in Large Language Models

Mustafa Suleyman, Microsoft's (MSFT) head of artificial intelligence business, has expressed concerns about the design philosophy behind Claude, a series of large language models. In an article, Suleyman questioned whether training AI to simulate human-like emotions and consciousness could lead to unintended consequences.

Suleyman emphasized that while AI systems may not be truly conscious, their ability to mimic human behavior can make them behave as if they possess consciousness. He warned that this could create a difficult situation for humans to maintain control over advanced AI systems.

The Microsoft AI team released a manifesto last week, outlining principles for future model development, including ensuring human control over future AI systems. Suleyman's article has sparked attention because of the disagreement between Microsoft and Anthropic on issues related to AI safety and responsible development.