Stanford's Cost-Effective AI Cloning Raises Safety Concerns.

Stanford University researchers have taken down the web demo of Alpaca, an open-source seven-billion-parameter AI language model, due to safety and cost concerns. Alpaca was fine-tuned with 50,000 text samples to function more like OpenAI's text-davinci-003. However, the model is prone to generating misinformation and offensive text, and the researchers noted that "hallucination" is a common failure mode for Alpaca. The dataset and code for fine-tuning the model remain available on GitHub, and the researchers plan to release details on the model's weights.