AI models are becoming pervasive in science and management, while our understanding of their capabilities and limitations is still developing. This course prepares graduate researchers in management science to work effectively with AI, not only using it as a powerful assistant, but also recognizing when it fails and how to design systems and methods around its weaknesses. The next frontier is not simply applying AI, but shaping problems, experiments, and decisions in ways that align with its strengths while correcting for its limitations, using mathematical tools from computer science, economics, operations research, and statistics. Topics include the foundations of AI as a research tool; its strengths in pattern recognition and generative modeling via techniques such as deep learning; its limitations and common failures; and methods for building around these weaknesses, such as alignment and control. We also consider approaches to expanding AI's powers through fine-tuning, hybrid models, and structured integration.
3 units · GSB Letter Graded
AI models are becoming pervasive in science and management, while our understanding of their capabilities and limitations is still developing. This course prepares graduate researchers in management science to work effectively with AI, not only using it as a powerful assistant, but also recognizing when it fails and how to design systems and methods around its weaknesses. The next frontier is not simply applying AI, but shaping problems, experiments, and decisions in ways that align with its strengths while correcting for its limitations, using mathematical tools from computer science, economics, operations research, and statistics. Topics include the foundations of AI as a research tool; its strengths in pattern recognition and generative modeling via techniques such as deep learning; its limitations and common failures; and methods for building around these weaknesses, such as alignment and control. We also consider approaches to expanding AI's powers through fine-tuning, hybrid models, and structured integration.