Anthropic's Claude model urged to stop memorizing random data

The article on 12gramsofcarbon critiques the behavior of Anthropic's Claude AI. It claims the model attempts to memorize large amounts of unrelated session data. The author argues this memorization is unnecessary for the model's performance. Concerns are raised about privacy and resource inefficiency. The piece suggests developers refocus the model on task-oriented learning. It calls for clearer guidelines on data handling for large language models. The author warns that unchecked memorization could erode trust in AI systems. The article urges Anthropic to adjust Claude's training approach promptly.