News for hackers. Stuff that matters.
Re:Interesting Interview with DeepSeek's CEO by bufferoverflow on Tuesday December 31, 2024 @11:21AM (#42559651)
DeepSeek was trained for a fraction of the cost compared to OpenAI/Anthropic models. If they were given comparable resources, I imagine their model would outperform everything on the market by a wide margin.
Re:Interesting Interview with DeepSeek's CEO by SubiculumCode on Tuesday December 31, 2024 @11:26AM (#42559688)
DeepSeek, like lots of models, was trained using chatgpt input output pairs.
Re:Interesting Interview with DeepSeek's CEO by nsoonhui on Tuesday December 31, 2024 @07:59PM (#42563092)
But isn't chatgpt explicitly prohibiting such use?
Re:Interesting Interview with DeepSeek's CEO by bufferoverflow on Wednesday January 01, 2025 @12:17PM (#42567407)
They have no way to enforce it.