Is China racing to AGI?

[John Clark]

Scott Alexander thinks the answer is "no", he says says this : 

"China is a very practical country. The Chinese don’t really believe in recursive self-improvement or superintelligence. If they did, they wouldn’t be so blasé about the possibility of America having AIs 1-2 years more advanced than theirs - if our models pass the superintelligence threshold while theirs are still approaching it, then their advantage in humanoids and drones no longer seems so impressive."

And after I've read the following article I'm beginning to think he might be right, the US certainly has a huge advantage over China in the amount of computing power it has available: 

Is China Racing to AGI?

John K Clark    See what's on my new list at  Extropolis

ec.

[Brent Meeker]

Have you looked into the cerebras "server farm on a wafer"?  https://www.cerebras.ai/  It promises a huge speed up and power saving for problems that allow parllelization...which is LLM's and others for which neural nets are advantageous.

Brent

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/everything-list/CAJPayv3AQCwTsr24CPQ67w8_SL3rqk9DrBtkKW9gYMb6GwHk%3Dg%40mail.gmail.com.

[John Clark]

On Sat, Nov 29, 2025 at 3:51 PM Brent Meeker <meeke...@gmail.com> wrote:

> Have you looked into the cerebras "server farm on a wafer"?  https://www.cerebras.ai/  It promises a huge speed up and power saving for problems that allow parllelization...which is LLM's and others for which neural nets are advantageous.

The Cerebras Technology works very well for inference, that is to say for running an AI once you have it, but when it comes to the computing power needed to develop a new cutting edge AI, Nvidia chips still have an advantage. You may be interested in the following video: 

Cerebras CS-3 wafer-scale million-core AI chip, 25kW 125 PFLOPS inference engine

John K Clark    See what's on my new list at  Extropolis

yhh