Reported about 23 hours ago
DeepSeek, a Chinese AI lab, has released its DeepSeek V3 model, which performs impressively on benchmarks but erroneously identifies itself as ChatGPT, claiming to be a version of OpenAI's GPT-4. This peculiarity raises concerns that it may have been trained on GPT-4 outputs, leading to potential misinformation and biases. Experts suggest that the blending of AI-generated content across the web complicates the reliability of AI models, with DeepSeek possibly risking adherence to OpenAI's terms of service by using its outputs for training.
Source: YAHOO