Does Google Actually Believe LLM Scaling Won’t Lead to AGI?

I have been thinking quite a bit about Google’s place in the AI race recently.

Despite having enormous amounts of money, data, talent, and computing resources, none of Google’s models seem to have consistently pulled ahead of OpenAI or Anthropic at the frontier. What makes this even more interesting is that Google’s DeepMind researchers were among the authors of “Attention Is All You Need,” the landmark paper that introduced the Transformer architecture that underpins modern LLMs.

So how did Google, with all those advantages, end up seemingly playing catch-up, even behind some Chinese AI labs in certain areas?

It makes me wonder: Is Google simply executing poorly, or does it fundamentally disagree with where the AI industry is heading?

Could Google actually believe that LLMs are ultimately a bubble that AGI won’t emerge simply by scaling models, data, and compute year after year? Maybe they think the next breakthrough requires a fundamentally different approach rather than just bigger and better LLMs.

Or perhaps I’m reading too much into it.

That’s all, folks. I’d genuinely love to hear what you think.