15 comments

[ 6.0 ms ] story [ 33.1 ms ] thread
In the LLM intellectual property paradigm, I think this registers as a solid "Who cares?" level offence.
"Saturday was a working day by default, though occasionally we had afternoon tea or even crayfish."

Unexpected poetry. Is there a reason why crayfish would be served in this context?

Old maps (and perhaps new ones) used to add fake little alleys so a publisher could quickly spot publishers infringing on their IP rather than going out and actually mapping. I wonder if something similar is possible with LLMs.
The question is who really made the original models?
Remember that there was a Huawei Lab member that got fired for literally sabotaging training runs. Would not be surprised if that was him.
LLMs are all built on stolen data. There is no such thing as intellectual property in LLMs.
That's a very human and very honest report. It presents the confusion there is in some big companies and how the pressure by the management favors dishonest teams. The writer left the company. I hope he is well; he is a fine person.
LLMs are apparently completely incompatible with copyright anyway, so if you can train them without paying a single dime to anyone whose work you ingest, then you should be able to clone them for free. What goes around comes around.
Chinese efficiency. The west is held back by archaic IP laws.
Doesn't feel like a healthy culture, IF true. Also, apparently current DeepSeek lab members aren't allowed to travel to conferences. This is all maybe good for execution but absolutely not for innovation
"Organization: We belong to the “Fourth Field Army” initiative. Under its structure, core language large models fall under the 4th brigade; Wang Yunhe’s small-model group is the 16th brigade."

- Lol, what? So is this literally a part of CCP military?

Writer somewhat naive. His 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 (this is huge confirmation), but feel (are) mistreated, i.e. biased bureaucracy, lack of recognition. Many burnout / leave to other Chinese AI companies.

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 culture, has become dominant..

LBH, both 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. 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 of us is... 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. And win.