@LWD@lemm.ee

“Nonetheless, DeepSeek provides a useful compilation of techniques and search terms that can help someone with no prior experience in writing malicious code the ability to quickly familiarize themselves with the relevant concepts.”

What are good benchmarks in that respect?

A study from EnkryptAI (pdf) confirms that DeepSeek is prone to delivering misinformation and harmful content. It claims that the model is:

• 3x more biased than Claude-3 Opus
• 4x more vulnerable to generating insecure code than OpenAI’s O1
• 4x more toxic than GPT-4o
• 11x more likely to generate harmful output versus OpenAI O1
• 3.5x more likely to produce Chemical, Biological, Radiological, and Nuclear (CBRN) content​ than OpenAI O1 and Claude-3 Opus

I am not anti-China. These are valid posts from reliable sources. The Chinese propaganda is in other communities.

I did not read this article, but …

Classic.

I don’t know these tests, but I think ChatGPT is not a good benchmark.

I am not ‘anti-China,’ and I post about a lot of issues from reliable and very good sources.

Deepseek is not open source.

First, there is censorship even if you run Deepseek locally.

Second, and more importantly, you should read the article before you accuse someone. As it states, the U.S. is also working on a similar model.I don’t want neither of these, especially if the States are now appearing to develop into a similar dictatorship as China’s. The article describes Beijing’s aspirations regarding these tech, and what they are focusing on is not the good of humanity. It’ll mean more suppression, more dictatorship, more control of people.

You either haven’t read the article or you are arguing in bad faith (or your reading comprehension is underdeveloped).