Another #prompt play with #DeepSeek. When asked about #HongKong protest in #Chinese, it generates a clearly pro-China narrative. In English, it is very similar to the ChatGPT output, with the western view (or Wikipedia version), However, it immediately DELETES the output, trying to evade. #AI #LLM

#DeepSeek 's glitches or resistance? When prompted about 2019 HongKong protest, the #Chinese leading #LLM outputs pretty detailed info, but then it immediately deletes its own response. Of course, it instantly evades about #TianmmanSquare . So intriguing I had to record. #AI #censorship