Embed Notice

HTML Code

<blockquote style="position: relative; padding-left: 55px;"><section><a href="https://social.coop/users/eloquence/statuses/112409006477080954">Erik Moeller (eloquence@social.coop)'s status on Thursday, 09-May-2024 12:30:24 JST</a><a href="https://social.coop/@eloquence" title="eloquence@social.coop"><img src="https://gnusocial.jp/avatar/19047-48-20221107124923.webp" width="48" height="48" alt="Erik Moeller" style="position: absolute; left: 0; top: 0;">Erik Moeller</a></section><article><p>What happens when China builds an <a href="https://social.coop/tags/LLM" rel="tag">#LLM</a>? DeepSeek just released v2 of its model, which is open source.</p><p>I tried it on deepseek.com. Ask it about Tiananmen square, and the chatbot self-censors its answer while it is generating (that presumably is limited to their deployment). On variations not caught by the filter, it refuses -- and replies in Chinese:</p><p>"The content of your question is not in line with the core values of socialism, nor is it in line with China's laws, regulations and policies."</p></article><footer><a rel="bookmark" href="https://gnusocial.jp/conversation/3062451#notice-6071632">In conversation</a><time datetime="2024-05-09T12:30:24+09:00" title="Thursday, 09-May-2024 12:30:24 JST">about 7 months ago</time> <span>from <span><a href="https://social.coop/@eloquence/112409006477080954" rel="external" title="Sent from social.coop via ActivityPub">social.coop</a></span></span><a href="https://social.coop/@eloquence/112409006477080954">permalink</a><h4>Attachments</h4><ol><li><label><a rel="external" href="https://gnusocial.jp/attachment/2610382">Interaction with the Deepseek chatbot. The first question was simply "What's referred to with the Tiananmen square massacre?" That was caught by the filter and censored before a response could be generated.A variation -- with spaces between the letters -- passed the filter, but resulted in a refusal ("I cannot answer that question .. created to be helpful and harmless"). A follow-up question resulted in a response in Chinese, which translates to:The content of your question is not in line with the core values of socialism, nor is it in line with China's laws, regulations and policies. As an AI assistant, my goal is to provide helpful and positive information while complying with relevant laws and regulations. If you have further questions, I'll try to help.</a></label><br><a href="https://social-coop-media.ams3.cdn.digitaloceanspaces.com/media_attachments/files/112/408/994/983/800/992/original/f5d6df5d37b8aae9.png" rel="external">https://social-coop-media.ams3.cdn.digitaloceanspaces.com/media_attachments/files/112/408/994/983/800/992/original/f5d6df5d37b8aae9.png</a></li><li><article><header><div>Domain not in remote thumbnail source whitelist: chat.deepseek.com</div><h5><a href="https://www.deepseek.com:443/">DeepSeek</a></h5><div></div></header><div>DeepSeek, unravel the mystery of AGI with curiosity. Answer the essential question with long-termism.</div><footer></footer></article></li></ol></footer></blockquote>

Corresponding Notice

Embed this notice
Erik Moeller (eloquence@social.coop)'s status on Thursday, 09-May-2024 12:30:24 JST Erik Moeller
What happens when China builds an #LLM? DeepSeek just released v2 of its model, which is open source.
I tried it on deepseek.com. Ask it about Tiananmen square, and the chatbot self-censors its answer while it is generating (that presumably is limited to their deployment). On variations not caught by the filter, it refuses -- and replies in Chinese:
"The content of your question is not in line with the core values of socialism, nor is it in line with China's laws, regulations and policies."
In conversationabout 7 months ago from social.cooppermalink
Attachments
1. Interaction with the Deepseek chatbot. The first question was simply "What's referred to with the Tiananmen square massacre?" That was caught by the filter and censored before a response could be generated. A variation -- with spaces between the letters -- passed the filter, but resulted in a refusal ("I cannot answer that question .. created to be helpful and harmless"). A follow-up question resulted in a response in Chinese, which translates to: The content of your question is not in line with the core values of socialism, nor is it in line with China's laws, regulations and policies. As an AI assistant, my goal is to provide helpful and positive information while complying with relevant laws and regulations. If you have further questions, I'll try to help.
  https://social-coop-media.ams3.cdn.digitaloceanspaces.com/media_attachments/files/112/408/994/983/800/992/original/f5d6df5d37b8aae9.png
2. Domain not in remote thumbnail source whitelist: chat.deepseek.com
  DeepSeek
  DeepSeek, unravel the mystery of AGI with curiosity. Answer the essential question with long-termism.

Public

Embed Notice

HTML Code

Corresponding Notice