DeepSeek's reported $294,000 training cost is misleading; the actual cost to train their base model was around $5.87 million, with the lower figure referring only to a specific reinforcement learning phase, not the entire training process. The article clarifies misconceptions about the expenses involved in developing large AI models and compares DeepSeek's efforts to Western counterparts like Meta's Llama 4.
OpenAI responds to The New York Times' copyright lawsuit, claiming that training AI models using publicly available data, including articles from the web, is fair use and necessary for innovation and U.S. competitiveness. They argue against the phenomenon of regurgitation and place responsibility on users to avoid intentionally prompting models to regurgitate. The copyright debate around generative AI is intensifying, with notable figures and organizations weighing in. Some news outlets have chosen to ink licensing agreements with generative AI vendors, but the payouts tend to be relatively small. A recent poll suggests that a majority of respondents believe AI companies should compensate outlets if they want to use copyrighted materials in model training.
