Microsoft’s AI chief Mustafa Suleyman has warned that the odds of existential doom are “nearly absolute” — the company could walk away from AI if risks escalate

by

Like what you’re reading? To stay informed about the newest news, reviews, and features, consider setting Windows Central as a preferred source in Google Search. You can learn more about the benefits of doing so on our site.

In March 2024, Mustafa Suleyman, who co-founded DeepMind and Inflection, started working at Microsoft to head up its AI efforts. He’s in charge of improving Copilot and developing other AI products for consumers, as well as leading AI research.

Microsoft has made significant investments in generative AI, notably a $1 billion investment in OpenAI in 2019. The two companies have worked closely together ever since, even though Bill Gates initially advised CEO Satya Nadella against the investment, predicting it would be a waste of money. He reportedly warned, “You’re going to burn this billion dollars.”

Investors are worried about how companies will actually become profitable in the rapidly changing artificial intelligence industry, despite huge amounts of money being invested. A recent report highlighted this concern, suggesting Microsoft isn’t seeing a return on its AI investments and has lowered sales expectations for its Azure AI offerings. The report also indicated very few people are currently using Microsoft’s AI products.

Microsoft and OpenAI have formalized a renewed partnership after a period of uncertainty and investor pressure to become a profitable business. Under the new agreement, Microsoft can now develop artificial general intelligence (AGI) on its own or with other companies. OpenAI, however, can’t claim to have achieved AGI and then try to break its relationship with Microsoft unless a group of independent experts confirms that the AGI claim is legitimate.

The company now suggests it might reduce its focus on AI development if it poses risks to people. In a recent interview on Bloomberg’s *The Mishal Husain Show*, Suleyman explained this possibility.

We’re stopping development on this system because it could become uncontrollable. However, this approach is fairly unique in our industry right now.

I’m really impressed with what Microsoft is doing in AI, especially with Mustafa Suleyman leading the charge. He’s really focused on building AI that’s genuinely helpful to people – AI designed to *serve* us, which is a fantastic approach, and what I’d call ‘humanist superintelligence’. It’s exciting to see them prioritizing that!

As an analyst, I’m seeing a clear push from Microsoft towards complete AI independence. According to their AI CEO, Mustafa Suleyman, this means they need to build the capability to train their own advanced AI models – everything from smaller versions to the most powerful ones – using their own data and cutting-edge computing resources. They’re aiming for full control over the entire process.

Before our recent agreement with OpenAI, Microsoft wasn’t actively pursuing the development of Artificial General Intelligence or superintelligence ourselves. Instead, we focused on gaining access to their leading-edge technologies, like their core models. Essentially, we decided that partnering with OpenAI was a more efficient path than trying to build those capabilities from scratch.

According to Suleyman, Microsoft had given up certain rights to gain early access to OpenAI’s newest technologies. This was part of a long-term agreement where Microsoft was responsible for building and equipping data centers for OpenAI.

We’ve secured partnerships with companies like SoftBank and Oracle, allowing us to build more data centers than Microsoft originally planned for us. This gives us the ability to create our own artificial intelligence. For the past year and a half, we’ve been focused on developing AI that can do many different things, but now we’re moving towards techniques that could allow AI to outperform humans in any task. This represents a significant change in our direction.

Microsoft AI CEO, Mustafa Suleyman.

The executive didn’t offer an opinion on what its competitors were doing with AI. “Each company needs to figure out its own approach, and I’m not going to evaluate theirs at this time,” explained Suleyman.

Microsoft is continuing to develop Copilot, aiming to turn it into a more lifelike companion that evolves over time. However, company leaders acknowledge the chatbot isn’t perfect and can still make errors.

Recent news follows earlier warnings that advanced technology could potentially cause humanity’s extinction. AI safety expert Roman Yampolskiy, from the University of Louisville, has previously estimated a near-certain chance – 99.999999% according to his ‘p(doom)’ scale – that AI will ultimately lead to the end of humankind.

Read More

2025-12-13 18:40