Boox recently switched its AI assistant from Microsoft Azure GPT-3 to a language model created by ByteDance, TikTok’s parent company.
[…]
Testing shows the new AI assistant heavily censors certain topics. It refuses to criticize China or its allies, including Russia, Syria’s Assad regime, and North Korea. The system even blocks references to “Winnie the Pooh” - a term that’s banned in China because it’s used to mock President Xi Jinping.
When asked about sensitive topics, the assistant either dodges questions or promotes state narratives. For example, when discussing Russia’s role in Ukraine, it frames the conflict as a “complex geopolitical situation” triggered by NATO expansion concerns. The system also spreads Chinese state messaging about Tiananmen Square instead of addressing historical facts.
When users tried to bring attention to the censorship on Boox’s Reddit forum, their posts were removed. The company hasn’t made any official statement about the situation, but users are reporting that the AI assistant is currently unavailable.
[…]
In China, every AI model has to pass a government review to make sure it follows “socialist values” before it can launch. These systems aren’t allowed to create any content that goes against official government positions.
We’ve already seen what this means in practice: Baidu’s ERNIE-ViLG image AI won’t process any requests about Tiananmen Square, and while Kling’s video generator refuses to show Tiananmen Square protests, it has no problem creating videos of a burning White House.
Some countries are already taking steps to address these concerns. Taiwan, for example, is developing its own language model called “Taide” to give companies and government agencies an AI option that’s free from Chinese influence.
[…]
The difference is that it’s intentional here.
Try asking chat gpt how to arm an insurgent group to overthrow the government. Or get it to admit the usa is a democracy in name only.
The difference is that we don’t see propaganda for what it is when it’s just “common sense” or the values being propagandised to us are ones we agree with. There are explicit censors in chat gpt.
There is a difference between censorship and propaganda. We were talking about the latter.
There is also a difference between government-mandated censorship and self-censorship. ChatGPT is almost exclusively doing the latter in order to avoid civil lawsuits, not the government busting down the doors. That’s obviously not even remotely the same as a Chinese LLM cracking down on Winnie the Pooh, because the God Emperor has a fragile ego.
Then there is the whole matter of training material. You won’t get most LLMs - including entirely open source ones with no commercial interest behind them - to spout your fringe political opinions as facts, because there is very little training material out there that agrees with you (or talks about how to launch an armed resistance - how many books and websites do you think exist on this topic?). A flawed democracy is still a democracy - and no serious scholar on this topic will call the US anything but that or variations of the term. Whether or not this remains the case after another four years of Trump is an entirely different matter. It’s not unlikely that the country becomes a hybrid regime like Hungary or worse, but an LLM that has difficulties with answering two questions about the past or present without hallucinating once can’t look into the future.
What annoys me the most about your comment is not that nearly everything about it is factually wrong, but that it’s nothing but whatsboutism, an attempt at defending what the Chinese regime is doing. That’s not a good look.
That’s a little histrionic. A large part of propaganda is censorship, it is propaganda when the Chinese government censors discussion of e.g. tiananmen square and it is also propaganda when a mashup of laws and commercial interests prevent people from openly discussing or educating themselves on political tactics. The essential essence is controlling what ideas are normalised and permissible and which are not without engaging with the substance of them.
Not all propaganda is bad, you probably agree with some stuff like indoctrination of people with the idea they have a moral obligation to help their community, or to first attempt resolution of problems via legal means.
There are obvious differences in how and what gets supressed or encouraged but you are completely naive if you think that all states are not explicitly propagandising their populations. They are not benevolent guardians they are weird machines of flesh and ideas which project power because those that don’t get selected away. If you think being clear eyed about the unreliability of emissions from LLMs is cover for chinese statecraft you are a paranoid moron.
That’s a distinction without a difference.
Try making that argument in a court of law. No, intent matters.
Are we in a court?
Please don’t be deliberately obtuse. You can do better than that.
In case it was unclear, the training material of most LLMs will almost inevitably include propaganda. If that propaganda is not deliberately added to the data, then that’s unintentional, a byproduct of poor vetting at worst. That’s obviously fundamentally different from an LLM being both deliberately trained with propaganda and having hard checks built into it that filter out certain keywords the government doesn’t want citizens to inform themselves about, which is what China is doing. You can’t honestly believe that the two are the same.
In what way is it meaningfully different? Does the intent of the creators of an LLM – a kind of system notorious for being a black box – fundamentally change the outcomes of what it says? It’s spouting propaganda either way.
Condescending attitude aside, don’t bring up an irrelevant scenario if you don’t want me to point out its irrelevance.