‘Artificial General Intelligence isn’t coming, it’s already here’

Artificial General Intelligence isn't coming - it's already here, according to one of Google's most senior AI researchers.

- Vishal Mathur

Blaise Agüera y Arcas, vice president, fellow and chief technology officer of Technology & Society at Google, has broken ranks with Silicon Valley's obsession over AGI timelines. In an exclusive interview with Hindustan Times, he declared definitively: "We already have AGI."

The statement marks a sharp departure from industry consensus. While Tesla chief Elon Musk predicts AGI by next year, Nvidia CEO Jensen Huang targets 2029, and OpenAI's Sam Altman forecasts human-level intelligence by 2026, Agüera y Arcas insists the milestone has already passed.

Microsoft's AI chief Mustafa Suleyman anticipates an AGI timeline of around 10 more years. The reason Blaise believes most people don't realise that we've already crossed into that territory, is due to the lack of a "moment" that would signify that transition.

"The 'G' in AGI means general," explained Agüera y Arcas, who leads Google's Paradigms of Intelligence team. "The moment we began training models like LLMs in an unsupervised waynot for narrow tasks like face recognition or handwriting-we got rid of the narrow. That's when they became general."

The Google executive, recognised as one of the inventors of federated learning in 2016, argues in his latest book What is Intelligence? that the transition occurred around 2019. That year, sequence models began defeating the Winograd Schema challenge, a test introduced in 2011 as an alternative to the Turing benchmark.