Google DeepMind CEO Demis Hassabis says AI scaling 'must be pushed to the maximum' (www.businessinsider.com)

Demis Hassabis, CEO of Google DeepMind, recently emphasized the importance of scaling AI systems during an appearance at the Axios' AI+ Summit in San Francisco. He argues that maximizing scaling laws—feeding AI models more data and computation power—is crucial for advancing towards artificial general intelligence (AGI), the theoretical AI capable of human-like reasoning. Following the notable success of DeepMind’s Gemini 3, Hassabis stated that scaling might not only accelerate the path to AGI but potentially serve as its foundation, although he acknowledges the need for additional breakthroughs. This perspective comes amidst a broader debate within the AI community about the sustainability and effectiveness of scaling alone, particularly as industry leaders like Yann LeCun of Meta advocate for alternative approaches. Critics argue that the reliance on perpetually increasing data and compute resources may lead to diminishing returns, especially with limits to available data and the environmental costs of expansive data centers. LeCun is pivoting towards developing world models, which focus on spatial data and aim to create AI systems capable of understanding the physical world, reasoning, and planning complex actions. This divergence highlights an ongoing tension in the AI landscape between traditional scaling methodologies and innovative, potentially more efficient approaches to achieving advanced AI capabilities.