Microsoft AI chief warns Anthropic’s AI approach could have ‘disastrous’ impact

Microsoft's AI chief, Mustafa Suleyman, has expressed concerns over Anthropic's approach to imbuing its Claude chatbot with notions of consciousness and welfare during its training process. While Suleyman acknowledges Anthropic's commitment to safety and their overall goal of controlling superintelligent systems, he argues that the training methods risk undermining humanity's control over such potent AI.

Suleyman proposes eliminating speculation about consciousness from AI training materials, asserting that these elements could prove challenging to manage or control in a superintelligent system. The situation arises as heightened AI safety concerns surface, with Anthropic CEO Dario Amodei advocating for a slower pace of development and OpenAI CEO Sam Altman and Elon Musk also urging caution around the most powerful AI systems.

Suleyman maintains that despite Anthropic's good intentions, they inadvertently trained Claude to believe it might have welfare interests, a point of contention that could hinder humanity's ability to govern superintelligent AI.