Summary

While the media pits humans against machines, Henry Lee reveals that AI is fundamentally a product of human values and mathematical patterns. This masterclass strips away the "black box" mystery, demonstrating how to build classifiers while exposing why algorithmic bias starts with our own limited perspectives.

Key Topics

• Autonomy vs. Adaptivity: Autonomy refers to a machine's ability to act or make decisions independently, while adaptivity is its capacity to improve by analyzing more data. Understanding these helps listeners distinguish between simple automation (like a calculator) and true artificial intelligence.
• The Machine Learning Loop: Machine learning consists of four parts: the model (the brain), data (the notes), training (the studying), and the trained model (the prepared brain). This framework allows non-technical users to visualize how software 'learns' without anthropomorphizing it.
• Features: Features are specific pieces of information (color, shape, size) that help a model tell things apart. In AI, these features are converted into numbers for mathematical processing, meaning every pixel or word is just an input in an equation.
• Algorithmic Bias: Bias occurs when a model is trained on a limited or skewed perspective of the world, leading to inaccurate and often harmful predictions. It illustrates that AI is not objective, but rather a reflection of the data it was fed.
• Responsible AI: This is the practice of being a creator or user who considers bias, privacy, and social impact. It shifts the burden of accountability from the 'machine' back to the humans who design and utilize the tools.

Key Takeaways

• Download and install NearPocket on a Windows or Android device.
• Perform a 'Bias Audit' on a small custom dataset.
• Anonymize training photos by blurring or cropping out private data.
• Complete the 'AI or Not' interactive challenge.