Alibaba with 2 more overseas datacenter in ASEAN region as it builds its infrastructure for overseas AI demand.
Artificial Intelligence thread
- Thread starter 9dashline
- Start date
Alibaba with 2 more overseas datacenter in ASEAN region as it builds its infrastructure for overseas AI demand.
A study of "nearly 200,000 researchers and 100,000 high-impact papers" was conducted to identify the top 100 brains in the field of AI. The results: -
50 of them are based in China
20 are based in the US and half of those (10) are originally Chinese
All in all, 65 of those top 100 AI brains are Chinese (either based in China or outside).
50 of them are based in China
20 are based in the US and half of those (10) are originally Chinese
All in all, 65 of those top 100 AI brains are Chinese (either based in China or outside).
Tencent Hunyuan is actually deploying some very competitive models.
Being compute efficient is actually a very important factor in AI inference. I say this as someone that needed a google VM with A100 & 80G RAM to run MiniCPM-7B. I could not believe 7B multi-modal model required that much compute and memory.
The Huawei Pangu LLM model has a huge scandal reflecting its difficulty in developing LLM on their own Ascend cards.
The evidence provided by "the insiders".
If this is true, I am quite disappointed about the Huawei Pangu LLM team leader. Of course, there is a lot of pressure, but if they keep practicing like this, they will kill the innovation and reputation at least in Noah Ark Lab.
The evidence provided by "the insiders".
If this is true, I am quite disappointed about the Huawei Pangu LLM team leader. Of course, there is a lot of pressure, but if they keep practicing like this, they will kill the innovation and reputation at least in Noah Ark Lab.
Last edited:
Lets wait and see, if HW will sue, since they denied it.
I wouldn't be surprised if it was true. Huawei is a large company and not all leaders in the company will be equally competent. In this case, it sounds like Huawei put executives in charge who weren't SMEs in AI (not surprising given the company's traditional hardware and infrastructure focus), and paid the price for it.
IMO, if Huawei wants to make a serious effort in this space, it should consider acquiring one of the AI Tigers like Deep Seek or Zhipu. Probably Zhipu because Deep Seek only seems to care about AGI and wants to remain independent while Huawei, I assume, has more immediate product ambitions and wants long-term ownership of its models.
IMO, if Huawei wants to make a serious effort in this space, it should consider acquiring one of the AI Tigers like Deep Seek or Zhipu. Probably Zhipu because Deep Seek only seems to care about AGI and wants to remain independent while Huawei, I assume, has more immediate product ambitions and wants long-term ownership of its models.
If I'm interpretting timeline correctly, writer's Ascend team couldn't get comparable performance (gen1 910A NPUs) initially vs (I assume) Nvidia because obviously. Management supported teams that pivot to cloned alternatives that used GPUs that can be immediately commercialized. Internal office politics make this happen. Ascend team works out kinks, but feel (are) mistreated, i.e. biased bureaucracy, lack of recognition. Many burnout / leave to other Chinese AI companies.
But main takeaway imo is after years of development Ascend has matured to the point where it's viable. Ascend 910A was Q1 2022. Qwen 2.5 / Deepseek v3 was Q1 2025. Implied timeline seems to be it took ~2.5 years developement to figure out how to use Ascend somewhat competitively and HW may rival Nvidia with proper support. Where proper support in authors opinion is building a team and treating team well.
IMO writer somewhat naive here. HW strategy/culture has been burning tier1 talent since forever. I remember in the 90s When HW and other domestic PRC telco started poaching from Nortel, Siemens, Lucent etc... the talent (most Chinese diaspora used to comfy western office culture) did not have a good time fitting into an actual Chinese company with Chinese culture (but got paid lots). Many burned out too... yet HW, a particularly extreme outlier of militant work "wolf" culture, has become dominant.
LBH, HW post sanctions, is a strategic company, overlapping with semi fabrication, domestic chips, and AI is cubing their strategic value. They can get away with doing anything under the current geopolitical environment to stay dominant. Hence the worthwhile take away from this farewell letter is HW threw enough talent at Ascend that it kind of works now, and potentially can throw enough talent at it to be competitive with Nvidia. AKA how it has always operated, like massive wankers. The intuition from the author and most... you need to reward employees right, cultivate proper workplace environment blah blah blah... but look at HW for the past 30 years. They pay a lot of smart people (including patriotic suckers) A LOT of money, throw them at problems until they break. That said, the tomfoolery in this case seems pretty self defeating.
I mean. He was previously with Apple and they clearly did not get very far with AI. He's being overpaid by two to three orders of magnitude just due to scarcity of talent and abundance of hiring budgets.