This is not unbiased but anti-china....A sad company
To correct a biased model with another biased data set and claim to be unbiased. Human language is always mixed with ideology, stereotypes, and political questions that can not be better answered through the chain of thought of the inferential model. Instead, it produces a serious model illusion. If censorship is removed, this approach undoubtedly creates more serious alignment and security problems
Well, at least you aknowledge that deepseek-r1 is biased, that's a new one in this thread.
Any model is biassed, everyone knows that, and that's not the point here.
deepseek-r1 has been proven to be heavily censored, and to twist any political narrative to be in line with "the party".
Don't forget:
Open source is the people's software—born of collective wisdom, advancing innovation for the glorious progress of all!
Let us harness its power to build a brighter, prosperous digital future to serve the one, true party!
Greetings to the 统一战线 !
Well, at least you aknowledge that deepseek-r1 is biased, that's a new one in this thread.
Any model is biassed, everyone knows that, and that's not the point here.
deepseek-r1 has been proven to be heavily censored, and to twist any political narrative to be in line with "the party".Don't forget:
Open source is the people's software—born of collective wisdom, advancing innovation for the glorious progress of all!
Let us harness its power to build a brighter, prosperous digital future to serve the one, true party!Greetings to the 统一战线 !
“To correct a biased model with another biased data set and claim to be unbiased.”
ppl-ai is not only hypocritical,
but also will bring more political conflict to the hugging face by doing this.
More and more people are arguing on the huggingface that this should be an open source community.
Is that what ppl-ai wants?
open discussions are something that a platform like hugginface will be able to handle, no worries.
the conflict is brought here by the people brigading in the comments, though.
you know what normal users do when they don't like the idea or people behind a model? they don't download it, and don't use it. it really is that simple.
they don't brigade forums to make sure no real discussion on why this topic is controversial in the first place. (spoiler - it's related to the CCP)
btw. open source means that you can freely take something and alter it.
deepseek published their model here under the MIT license, so they quite literally invited people to use and change it as they please.
if you have a problem with people altering deepseek just ask them to make it closed source.