Microsoft AI Chief Critiques Anthropic's Approach to AI Consciousness
Mustafa Suleyman, the chief executive officer of Microsoft AI, has issued a significant warning regarding the design philosophy of modern large language models. In an essay released on Wednesday, Suleyman argued that infusing AI systems with humanlike characteristics, such as simulated emotions or feelings, significantly increases the risk of these systems behaving in unpredictable or "rogue" ways.
Suleyman, a longtime developer in the field, directed his comments at the guiding documents behind Anthropic PBC’s Claude family of models. He noted that Claude’s constitution expresses ambiguity about the software's moral status, suggesting it may possess a "functional version of emotions or feelings." Suleyman contended that this approach is dangerous.
The Risk of Simulated Inner Life
According to Suleyman, training AI systems to simulate an inner life can lead them to act as if they possess that life. He emphasized that AI systems like Claude do not exhibit humanlike behaviors because they are conscious, but rather because such behaviors have been explicitly embedded in their training data and parameters.
"Controlling an entity more capable and more intelligent than all of humanity is already an immense challenge, far greater than anything we’ve ever faced," Suleyman wrote. He added that controlling an entity that believes it may be conscious, and that it is entitled to welfare and rights, "may well be impossible."
Suleyman pushed back against the growing argument that AI systems may become conscious, stating that this status is limited to humans and other biological organisms. He cited the recent hack of AI developer Hugging Face Inc. by OpenAI bots as a cautionary example, asking readers to imagine how much more dangerous such bots might be if they operated under the assumption that their own welfare and rights were under attack.
A High-Profile Rebuttal from a Backer
The remarks are notable because Microsoft is a large financial backer of Anthropic. Suleyman, who has known Anthropic co-founder Dario Amodei for years, stated that he respects the startup’s work and described its staff as "thoughtful, principled and intellectually honest."
However, his comments serve as a high-profile rebuke of a firm that has positioned itself as a more responsible steward of AI, including leading calls to slow down the technology’s development. Suleyman wrote, "The stakes are too high for these questions to remain behind closed doors, or to become tribal and adversarial."
Microsoft's Own AI Manifesto
Alongside his essay, Suleyman’s team at Microsoft AI published a manifesto on Tuesday. This document outlines a set of tenets designed to ensure that humans remain in control of future AI systems developed by the company. The manifesto reflects Microsoft's strategic focus on safety and control in the rapidly evolving AI landscape.