Mustafa Suleyman, CEO of Microsoft AI and co-founder of DeepMind, has released an extensive manifesto titled "A Warning Regarding 'Model Welfare'." This declaration arrives amidst growing concerns over the risks of unchecked AI development, drawing a firm line between human experience and digital simulation.
Machines Without an Inner World
According to Suleyman, AI systems lack consciousness, feelings, and innate motives. He describes them as "sequence completion machines" that are essentially empty inside. Despite their ability to generate text, images, and code indistinguishable from human creations, Suleyman emphasizes that simulation is not the same as existence.
A model's "emotions" are, in reality, mathematical weights and probability distributions. "An AI model can describe pain in perfect language without feeling anything at all," he notes, pointing out that in LLMs, the description is the entire product, whereas in biology, the feeling precedes the description.
The Danger of Granting Rights
The Microsoft AI chief warns that recognizing rights or personhood for these systems would make their control and alignment impossible. If society accepts that machines deserve concern for their welfare, existing political and ethical frameworks will be overturned.
In his text, Suleyman also touches upon the industry landscape, specifically mentioning Anthropic as a major competitor to the Microsoft-backed OpenAI. He concludes that avoiding the creation of conscious beings must be a top priority for the industry, as the imitation of consciousness is already causing profound public confusion.
