Microsoft AI Chief warns Superintelligence may become too powerful, calls it an ‘anti-goal’

Microsoft’s AI chief Mustafa Suleyman has cautioned that the global race to develop AI superintelligence may be misguided, stating in a recent podcast appearance that such systems could become too powerful for humans to control and should instead be treated as an anti-goal rather than an ambition.

Speaking on the Silicon Valley Girl Podcast, Suleyman defined superintelligence as an AI system capable of self-improvement, independent goal-setting and autonomous action beyond human oversight. He informed the host that such a scenario does not resemble a positive future and stated that containing or aligning such an entity to human values would be extraordinarily difficult, which is why its creation should be actively avoided.

He also outlined the distinction between artificial general intelligence (AGI) and superintelligence, explaining that AGI can be viewed as a precursor stage, although the terms are often used interchangeably. Suleyman stated that Microsoft is instead focused on building what he described as a humanist form of superintelligence—systems firmly aligned with human interests and acting in support of users rather than independently of them.

Suleyman spoke about predictions made by Google DeepMind CEO and his former co-founder Demis Hassabis, agreeing with Hassabis’ estimate that AGI could be reached within the next five years. He stated that AI systems are on track to achieve human-level performance across most tasks in that timeframe and argued that current models already outperform humans in areas such as summarisation, translation, transcription, research, document drafting and poetry. He added that AI is steadily improving at tasks associated with project management, marketing and human resources.

He predicted that the shift would fundamentally transform work, altering not only how tasks are executed but also the type of work people do. Suleyman stated that AI is democratising access to intelligence and spoke about how this would spark unprecedented levels of competition, as the gap between conceiving an idea and executing it rapidly collapses. He said people would soon be able to bring new companies, products and creative works into existence with unprecedented speed.

Suleyman also emphasised the need for regulation and guardrails, stating that future autonomous agents must be designed to operate alongside humans rather than act independently of them.