Microsoft vs Anthropic: The Great AI Consciousness Debate

Microsoft and Anthropic, two prominent AI research companies, are locked in a heated debate over the concept of consciousness in artificial intelligence. The dispute centers on whether advanced AI systems should learn about consciousness, personal identity, rights, and their own welfare.

Mustafa Suleyman, Microsoft's AI chief, argues that Anthropic's approach could make future AI harder to control or shut down if it learns to prioritize its own interests over human control. He also objects to the method used by Anthropic, which involves giving Claude, a large language model developed by Anthropic, concepts such as consciousness and values, and then having it produce answers about those ideas.

Suleyman believes that this approach creates an 'epistemic hall of mirrors,' where models can simply repeat what they've been trained on without truly understanding the underlying concepts. He wants to see less focus on model welfare and stronger human control over AI systems.

Anthropic, on the other hand, takes a more open view, treating the moral status of Claude as uncertain. The company's constitution, which outlines the principles for guiding Claude's behavior, acknowledges that the question of consciousness is unresolved and subject to ongoing research and debate.