Google’s AI CEO Wants To Feed AI So Much Data It’s Practically Bulimic
KEY POINTS
- •Demis Hassabis said at the Axios AI+ Summit that pushing scaling laws to the max is essential for AGI progress.
- •Hassabis believes that massive data and compute feeding could be the core or entirety of theoretical human-level AI.
- •Yann LeCun left Meta in 2025 to launch a startup focused on world models using spatial data instead of text.
At the Axios AI+ Summit last week in San Francisco, Demis Hassabis, Google DeepMind CEO and Gemini 3's proud dad, pushed 'scaling laws' like a gym rat obsessed with data protein shakes. He insists we pump machine learning with relentless amounts of data and compute, hinting that more is not just more but could be ALL that’s needed to birth AGI—artificial general intelligence, aka AI’s glow-up to human-level brainage. Meanwhile, Yann LeCun, Meta’s former AI guru who’s now sprinting to start a spatial-data startup, warns wickedly that ‘most interesting problems scale extremely badly,’ essentially saying you can't just throw supercomputers and expect Einstein 2.0. Good luck stacking data centers, though; they guzzle power like a frat house at midnight, raising environmental and economic eye-rolls across Silicon Valley.
Share the Story
Source: Businessinsider | Published: 12/7/2025 | Author: Lakshmi Varanasi