Demis Hassabis, CEO and co-founder of Google-owned AI lab DeepMind, believes that artificial general intelligence (AGI) that matches the cognitive abilities of humans could arrive within the next few years. Speaking at a Wall Street Journal conference, Hassabis acknowledged the incredible progress made in AI research in recent years and predicted that very capable, general systems will be developed in the near future. Google, which bought DeepMind for $500 million in 2014, is doubling down on AI to fortify its business and fend off a challenge to its core search unit from Microsoft-backed ChatGPT developer OpenAI. Hassabis is playing a central role in Google’s mission to advance its AI capabilities, having been announced last month as the leader of a newly formed unit at Google that brings DeepMind together with Google Brain, a separate AI research arm. Since being founded in 2010, DeepMind has been building machines that are capable of learning, thinking, and acting the way humans do. However, researchers across the field have become increasingly divided on the benefits of pursuing AGI, given the issues around inaccuracy and misuse posed by existing large language models underlying tools like ChatGPT and Google’s rival technology Bard. Hassabis suggested that developing AGI technologies would need to be done “in a cautious manner using the scientific method” that involves rigorous experiments and testing.
