American companies can't compete with China on foundational models; Musk may have already begun abandoning the Grok path.
Musk just said it's nearly inevitable that China's LLMs will eventually become the strongest globally.
To keep getting stronger, models face two major constraints: chips and electricity.
China's power generation exceeds that of the United States, Europe, and India combined. Although it still lacks chips now, Musk believes chips will eventually be available.
I think this situation will end up being similar to manufacturing. Once production technology, machinery, supply chains, and capital are all widely accessible, Western companies will find it increasingly difficult to maintain an advantage simply by claiming "my product is better." In the end, they'll have to shift focus toward branding, design, software, and other areas.
Musk's assessment is sound. The main reason America no longer holds a competitive edge is its incomplete manufacturing supply chain.
The same could be true for AI. As chips, computing power, and training technologies become more widespread, the gap between models will shrink further. American AI companies will gradually shift their focus from foundational models to computing power and application layers.
If you can't outperform China in applications built on models, then applications are what truly matter for users—no matter how good the model is, without practical applications it’s all for nothing.
That said, today’s AI situation resembles the late 1990s internet landscape—nobody knows where the real applications will emerge, especially those that haven’t even appeared yet. Back then, many people made predictions, and most were wrong.
Today is no different—nobody can know which new application areas AI will bring in the future.
Even the current internet leaders didn’t know back then—they just experimented and iterated along the way, like Amazon, which initially sold only books.
Now, even someone as influential as Musk cannot predict the future. The future is something we discover through trial and error.
Original: toutiao.com/article/1872945491627018/
Disclaimer: The views expressed in this article are solely those of the author.