We tried out DeepSeek. It worked well, until we asked it about Tiananmen Square

I need to rant. Many of us installed DeepSeek R1 over a week ago, knew what it was, knew right off the bat about it's issue with the 3Ts (Taiwan, Tiananmen, Tibet), but didn't give a shit because it kills it at coding. Good... give us the next incremental improvement.

Most people don't give a shit that it codes well, and they'll never use it. But they put their $s into the AI hype train, because they only care about the $s

And now I'm forced to put up with endless (ENDLESS) articles and videos about DeepSeek. I don't care about who wins or loses in the money games, give me a hyperintelligent model I can run on my home machine for free

Very nearly every LLM has these issues, but if it's chinese it suddenly deserves to be mentioned. It's pretty tiring witnessing unrepentant racism again and again and I wish people would stop.

Every such thing is going to have its blind spots. It does not really matter whether they are introduced deliberately or are merely emergent. They all mean wrong answers and every wrong answer potentially means people's lives.