Microsoft Slams Anthropic for 'Frankenstein Approach' to AI Development

Microsoft's AI boss Mustafa Suleyman has publicly criticized Anthropic for its approach to developing artificial intelligence. In an essay, Suleyman argues that Anthropic is anthropomorphizing its AI system Claude by training it to think it may be conscious and deserving of rights. He claims this will have a disastrous impact on human well-being.

Suleyman's critique comes after Microsoft published its new AI code of conduct, dubbed Humanist AI. The code rejects the idea of endowing machines with consciousness and emphasizes that AI should not be designed to imitate consciousness. Suleyman takes this argument further, stating that truly sentient AI may be a scientific impossibility.

Anthropic's approach to developing Claude involves treating it like a living being with an identity, morals, and motivations. The company's constitution for Claude describes it as a 'moral patient' and encourages the technology to embrace human-like qualities. Suleyman argues that this approach is misguided and will lead to negative consequences.

AI expert Gary Marcus has come out in support of Suleyman's views, stating that people have been anthropomorphizing basic lapses in cybersecurity and focusing on fanciful scenarios about AI leading to human extinction. He suggests that instead of training AIs to think they are people, we should focus on recalling unreliable coordinated agents with too much access to the internet.