While Elon Musk has consistently projected a utopian vision for the future of artificial intelligence, numerous other technology leaders and governments have been raising alarms about the technology's potential dangers.

Google DeepMind's chief AI readiness officer, Lila Ibrahim, was among the signatories of a 2023 statement declaring that mitigating the risk of human extinction from AI should be treated with the same gravity as nuclear war. Now, she is offering her own assessment of how probable that worst-case scenario might be.

"It's not zero," Ibrahim told Fortune. "Anything beyond that is just me guessing… that's the engineer in me."

Although three years have passed since Ibrahim signed the Center for AI Safety's statement—alongside OpenAI's Sam Altman and Anthropic's Dario Amodei—she said it would still "be irresponsible of us not to talk about existential risk."

Geoffrey Hinton, widely referred to as the "godfather of AI," has estimated the odds of human extinction at 10% to 20% within just three decades. Amodei has said there is a 25% chance that the future—meaning human existence—will go "really, really badly." Even the late physicist Stephen Hawking, who relied on AI to communicate his ideas to the world, warned that the technology could end mankind nearly a decade before ChatGPT emerged.

"The two things that keep me up at night are: Are we doing everything we can to mitigate the risks? And are we doing everything we can to maximize the opportunity? Those two, in my mind, go hand in hand," Ibrahim said.

Ibrahim, who was the company's first chief operating officer, noted that the COVID-19 pandemic served as a reminder that catastrophic risks are not always the ones you anticipate. She also pointed out that businesses have historically failed to adequately discuss potential downsides during previous technological shifts, such as the rise of the internet and later social media.

"We didn't sit around the table when we were making investment decisions and talk about the risks in the way that we should have. Sometimes we can be so much in the here and now," Ibrahim said, adding that apocalyptic science fiction films should be treated as cautionary tales.

"If we don't do anything to mitigate that or to invest in the good things along the way, then we're ending up exactly where we don't want to go," she added. "There's a lot of things that could go wrong."

Musk Says Money Won't Matter by 2036—Ibrahim Disagrees

If Ibrahim is cautious about assigning a precise probability to AI causing humanity's downfall, she is equally skeptical of the opposite extreme: a post-scarcity utopia.

Musk has repeatedly promoted a future in which AI and robots generate such abundance that work becomes optional and wages lose their relevance. He recently placed a 10-year timeline on robots taking over the workplace and rendering salaries obsolete.

"Money won't matter in 2036," Musk told The Economist, arguing that goods will become so inexpensive that society will enter a period of deflation and governments will distribute a universal high income to citizens freed from the daily demands of employment.

Similarly, Amazon founder Jeff Bezos has predicted that by 2045, "millions of people" will be living in space, with robots managing the daily commute—since, in his vision, people will be traveling to other planets rather than to offices. OpenAI's Altman has said that in less than 10 years, college graduates will be working "some completely new, exciting, super well-paid" job in space.

Ibrahim, however, is not convinced. Despite holding a front-row seat to AI's most significant breakthroughs at Google DeepMind—an advanced AI research lab focused on solving major scientific problems—she argues that the technology is evolving too rapidly for anyone to predict what comes next with meaningful accuracy. That uncertainty is part of why she frames the debate less as a choice between utopia and catastrophe and more as a practical question about governance, safety, and real-world use cases.

"I don't think we can really predict what a future state looks like," she told Fortune.

Her focus, she explained, is less about which planet humans might inhabit two decades from now and more about how these emerging technologies can be practically deployed to improve the world as it currently exists—addressing concrete challenges such as extreme weather and industrial waste, while ensuring that the benefits extend beyond the small group of executives running powerful AI companies. For businesses and policymakers, that means the discussion is not only about frontier models, but also about who gets access to them, which risks are monitored, and where the returns from productivity gains are likely to show up.

"I would rather we put our attention towards how we use technology as a tool to maximize human potential," Ibrahim said. "That's going to be where we need to be investing and spending our time—having the debates around what are the diseases that we can solve by having this new tool that can find patterns that we as humans haven't."

"I think a lot about how we get to a future versus what the future is," she added.