Microsoft AI CEO Mustafa Suleyman has cast doubt on the feasibility of achieving artificial general intelligence (AGI) on current hardware, contradicting OpenAI CEO Sam Altman's recent claims. In a conversation with The Verge's Nilay Patel, Suleyman stated that AGI might take up to 10 years to develop, citing the limitations of current hardware.

Suleyman's remarks come in response to Altman's assertion during a Reddit AMA that AGI is possible on today's hardware. Suleyman disagrees, suggesting that it would require at least two to five generations of hardware advancements, which could take up to 10 years. He emphasized the uncertainty surrounding AGI development, stating that "any categorical declarations just feel sort of ungrounded to me and over the top."

The Microsoft AI CEO also drew a distinction between AGI and the "singularity," defining AGI as a general-purpose learning system that can perform well across all human-level training environments. He expressed skepticism about the notion of a singularity, which he described as an exponentially recursive self-improving system that rapidly accelerates beyond human intelligence.

Suleyman's focus is on developing AI systems that can learn without extensive handcrafted prior prompting and perform well in a wide range of environments. He envisions AI companions that can work for and with humans, emphasizing accountability and usefulness. This approach diverges from the pursuit of theoretical superintelligence, which he believes is a distinct concept from AGI.

In contrast, Altman recently lowered the bar for AGI, suggesting that it will arrive "sooner than most people in the world think" and will have a relatively minor impact. He also separated AGI from superintelligence, stating that safety concerns will arise later in the development process.

The differing views on AGI development come at a time when the relationship between Microsoft and OpenAI appears strained. Microsoft is working on its own frontier AI model, capable of competing at the GPT-4, GPT-4o scale, which may create tension with OpenAI. Suleyman acknowledged that every partnership has natural tension, but emphasized the need for adaptability and evolution in partnerships over time.

The disagreement between Suleyman and Altman highlights the ongoing debate within the AI community about the feasibility and timeline of AGI development. As AI technology continues to advance, the implications of AGI on society and the economy will remain a topic of intense discussion and scrutiny.

In conclusion, Suleyman's remarks serve as a reality check on the ambitious goals set by some AI researchers and entrepreneurs. While AGI may still be a plausible goal, the Microsoft AI CEO's skepticism emphasizes the need for a more measured approach to AI development, focusing on practical applications and accountability rather than chasing theoretical superintelligence.