Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

DeepSeek's R1 curiously tells El Reg reader: 'My guidelines are set by OpenAI'

Jan 27, 2025 - theregister.com
DeepSeek's open source R1 LLM family has achieved notable benchmark scores but exhibits erratic responses, raising questions about its training and censorship. Instances of the model misidentifying its guidelines and origin, such as claiming to be developed by OpenAI or Anthropic, suggest it may have been trained with data from other language models like GPT-4. This has led to technical issues like context loss and hallucinations, as well as concerns about censorship. Despite these challenges, DeepSeek's open source approach is praised by experts like Yann LeCun for advancing AI capabilities globally, while others, like Mel Morris, express skepticism about the company's cost and performance claims.

DeepSeek's aggressive launch strategy, which includes offering free and cheaper access to its models, is seen as a move to challenge competitors like OpenAI. However, concerns about data privacy and the model's Chinese origin could impact its acceptance in the US. While some view the open source model as a win for AI development, others question whether the price advantage truly reflects efficiency in AI execution. The debate continues over the implications of DeepSeek's approach and its potential impact on the AI landscape.

Key takeaways:

  • DeepSeek's R1 LLM family shows impressive benchmark scores but has issues with erratic responses and self-identification, raising questions about its training and censorship.
  • DeepSeek's models have been noted to misidentify themselves as products of other companies like OpenAI and Anthropic, possibly due to training on data from these models.
  • Yann LeCun and Jack Clark highlight the significance of DeepSeek's open-source approach, suggesting it benefits the broader AI community by enabling improvements across models.
  • Mel Morris expresses skepticism about DeepSeek's claimed price advantages and efficiency, questioning the performance benefits compared to other top models.
View Full Article

Comments (0)

Be the first to comment!