Explore the potential pitfalls and biases in AI decision-making systems in this thought-provoking conference talk from NDC Oslo 2021. Delve into six categories of common problems in AI development, from confusing causation with correlation to the subtle effects of feedback loops in algorithms. Learn concrete steps and tools to mitigate these issues, ensuring more objective and fair AI systems. Discover the importance of responsible AI development and its impact on people's lives through real-world examples and practical solutions. Gain insights into concepts such as conditional probability, data sampling, bias recognition, and distributionally robust optimization to create more ethical and effective AI models.
Overview
Syllabus
Introduction
Why does this matter
Consequences
Correlation
Conditional Probability
Data vs Sample Data
Bias
What can you do
Pricing algorithms
Shadow of understanding
What can we do
Distributionally Robust Optimization
Feedback
Taught by
NDC Conferences
