TLDR: U.S. AI growth is hitting a wall because the power grid is weak and slow to expand. China has huge surplus power and uses data centers to soak it up. State-led, long-term planning gives China an edge, and the U.S. will fall behind unless it changes.
The whole AGI discussion is deeply misguided in my opinion. What we’ll actually see is a gradual increase in capabilities of AI system across different domains. While LLMs get most of the attention right now, they’re far from the only technique we know about, and people are already combining many different approaches together. For example, you have stuff like neurosymbolic systems where LLMs are used for classifying noisy input data, and a symbolic logic engine is used to reason about it. We’re also seeing a move towards world models as the basis for reasoning systems, and I think that’s where you start getting the ability to do genuine judgements. A system which has an internal persistent model of the world can reason about it in a similar way to a human.
In my opinion, it is highly likely we will be seeing automation replacing much of the labour that can only be done by humans within a few decades. However, I also expect that this will be happening in China where there’s actual industry to automate.
I very much agree with the points Alibaba cloud founder makes in this interview https://www.youtube.com/watch?v=X0PaVrpFD14
I found a YouTube link in your comment. Here are links to the same video on alternative frontends that protect your privacy: