Archive · 8 Dec 2025
DeepMind's Hassabis Advocates Maximum AI Scaling for AGI
Demis Hassabis, the CEO of Google DeepMind, believes that to create AI as smart as humans, we need to make AI models much bigger and more powerful.
He thinks this is the best way to reach a goal called AGI, where AI can think and solve problems like people.
However, not everyone agrees.
Yann LeCun, a famous AI scientist, thinks that just making AI bigger isn't enough.
He is working on a different kind of AI that learns like humans, using memory and understanding the world around it.
This debate is important because it shows there are different ways to make AI smarter, and people in the tech world are trying to figure out which way is best.
Demis Hassabis advocates for maximizing AI scaling to achieve Artificial General Intelligence (AGI).
DeepMind's Gemini 3 model reflects the company's focus on scaling AI systems.
Hassabis believes scaling is a key component, but acknowledges that additional innovations may be needed.
Yann LeCun criticizes the 'scale-is-everything' approach, advocating for AI systems that learn through spatial and contextual understanding.
The debate highlights the tension between scaling AI models and rethinking AI architecture for future breakthroughs.
- Who
- Demis Hassabis, co-founder and CEO of Google DeepMind, and Yann LeCun, former Chief AI Scientist at Meta
- What
- A debate on the path to achieving Artificial General Intelligence (AGI) through scaling versus rethinking AI architecture
- When
- During recent talks at the Axios' AI+ Summit and the National University of Singapore
- Where
- San Francisco and Singapore
- Why
- To discuss the future of AI development and the most effective strategies for achieving AGI
- Demis Hassabis
- Co-founder and CEO of Google DeepMind
- AGI
- Artificial General Intelligence, human-level AI
- Gemini 3
- DeepMind's recently launched AI model
- Yann LeCun
- Former Chief AI Scientist at Meta
- World Models
- AI systems focusing on spatial and contextual learning