谷歌DeepMind的CEO Demis Hassabis警告称，通用人工智能（AGI）可能比我们想象的更早到来，预计在5到10年内达到人类智能水平。这意味着机器在2035年左右就能像人类一样思考、推理和理解世界。Hassabis强调，世界尚未为AGI做好准备，并呼吁国际合作以确保其安全可控。尽管存在风险，AGI也有潜力彻底改变医疗健康领域，甚至可能“终结疾病”。Hassabis建议建立一个类似CERN的国际AGI研究中心，以及一个监督机构来监控和控制发展。

🤔 谷歌DeepMind的CEO Demis Hassabis预测，AGI可能在未来5到10年内实现，这意味着机器将在2035年左右具备人类水平的智能，能够像人类一样思考、推理和理解世界。

⚠️ Hassabis警告称，世界尚未为AGI的到来做好准备，并强调需要国际合作，包括政府、公司和实验室之间的合作，以确保AGI的安全性和可控性。DeepMind的研究甚至指出，如果处理不当，AGI可能对人类构成潜在的生存威胁。

💡 尽管存在风险，Hassabis认为AGI也有潜力彻底改变医疗健康领域。他表示，AI可能在未来十年内“终结疾病”，并提到了在蛋白质折叠和药物发现方面的突破。

🤝 为了应对AGI带来的挑战，Hassabis建议建立一个类似CERN的国际AGI研究中心，以及一个像国际原子能机构（IAEA）一样的监督机构，以监控和控制AGI的发展。

A top Google executive just confirmed what many have feared, and that is Artificial General Intelligence (AGI) is closer than we think.

Demis Hassabis, CEO of Google DeepMind, told 60 Minutes that AI could reach human-level intelligence in just 5 to 10 years. That means machines could think, reason, and understand the world like humans by 2035.

“We’ll have a system that really understands everything around you in very nuanced and deep ways,” Hassabis said.

The warning

In a separate interview with Time, Hassabis said the world is unprepared for what’s coming.

“Testing new AI features” April 25, 2025

— vitruppo (@vitrupo) April 25, 2025

“AGI is coming… I’m not sure society’s quite ready for that yet.” – Demmis Hassabis, chief executive officer and co-founder of Google DeepMind

He emphasized the urgent need for international cooperation between governments, companies, and labs to ensure AGI is safe and controllable.

DeepMind’s own research flagged AGI as a potential existential threat. A recent paper warned AGI could “permanently destroy humanity” if mishandled.

A powerful promise

Despite the risks, Hassabis says AGI could also revolutionize health. Speaking to The Economic Times, he said AI has the potential to be “the end of disease” within the next decade.

“I think there’s a good chance AI could be the end of disease as we know it,” Hassabis said, citing breakthroughs in protein folding and drug discovery.

AGI is different from today’s AI. It’s not just smarter software; it’s software that can do anything a human can. Think creativity, logic, intuition,  at digital speed.

Hassabis’ solution

He’s calling for a CERN-style international AGI research hub and an oversight body like the IAEA to monitor and control development.

The race to AGI is speeding up. And the person leading that race says we’re not ready, but the stakes couldn’t be higher, from ending disease to risking human extinction.

The post AGI by 2035? Google DeepMind CEO Warns “Society’s Not Ready” appeared first on DailyAI.