Artificial Intelligence thread

[9dashline]

Thoughts about the writing of this intellectual Hanjian?

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

View attachment 173425

Mythos is a 10T model.... ~1 Trillion activated.... and its about 50% saturated already... which means the ceiling is visible....
Once a watered down version (capability gated not just intent gated) is made public, it will likely be 2x to 3x the api charges of OPUS

China will catch up with the likes of Kimi K3 later this year

 

[tphuang]

Please, Log in or Register to view URLs content!

note on China's scnet.cn and its AI4S (AI for science) computing cluster has 3 million CPU cores and 200k GPU card to support science model development. This particular cluster by Sugon in Zhengzhou's Supercomputing center has 60000 card

 

[Proteus]

https://twitter.com/i/web/status/2044038044437668259

Update on the popularity of Dola/Cici in global markets. quickly gaining from where I can see.

I'm still cannot log in with the PC app; they always show "We are experiencing high demand right now. Please try again later."

 

[Michael90]

I'm still cannot log in with the PC app; they always show "We are experiencing high demand right now. Please try again later."

Compute capacity is getting limited as demand keeps skyrocketing. So will take time for most companies to catch up

 

[bsdnf]

Both countries' top model providers are experiencing severe computing shortages. Opus and GLM have both been complained about as becoming stupid, slow, and unresponsive.

And I still couldn't snag a GLM Coding Plan; it sells out in seconds every single day.

 

[HighGround]

Sounds like prices have to rise then.

 

[tphuang]

I get the same problem with Kimi and GLM, but I just try a few more times. I normally get responses.

Ever since agentic AI and openClaw took off, the compute resources are facing huge constraints.

see this, now CPUs are in shortage.

https://twitter.com/i/web/status/2044038072107434251

 

[meedicx]

Both countries' top model providers are experiencing severe computing shortages. Opus and GLM have both been complained about as becoming stupid, slow, and unresponsive.

And I still couldn't snag a GLM Coding Plan; it sells out in seconds every single day.

You still have the chance to buy a Kimi coding plan in CNY before prices rise. I bought an annual plan to lock in the price. I've been using the latest K2.6-code-preview and it's very good.

 

[Maikeru]

Stanford Annual AI Report. China/US AI gap "effectively closed" as AI capability accelerates:

Please, Log in or Register to view URLs content!

 

[shiftenter]

11:05:49 | National Bureau of Statistics: Average Daily Token Calls Surpassed 140 Trillion, Up 40% from the End of Last Year


April 16, Cailian Press — At a press conference held by the State Council Information Office on April 16, Mao Shengyong, Deputy Director of the National Bureau of Statistics, said that the leading role of intelligent development continues to strengthen, and China has achieved phased breakthroughs in the commercialization and large-scale operation of artificial intelligence. As of March this year, the average daily number of token calls had exceedebe end of last year. The development of artificial intelligence is empowering a wide range of industries and driving rapid growth in related sectors.