Cathy O'Neil

4.11215 learners

In *Weapons of Math Destruction*, former hedge fund mathematician Cathy O'Neil pulls back the curtain on seemingly neutral algorithms that operate as hidden tools of systemic oppression. She defines these "WMDs" by three core traits: opacity (their inner workings are kept secret), scalability (they target millions at once), and destructiveness (they trap marginalized groups in cycles of harm). Through gripping case studies—from criminal justice recidivism models that penalize poor communities to predatory for-profit college ads that exploit vulnerable individuals, from biased hiring tests that shut out qualified candidates to insurance pricing that overcharges low-income drivers—O'Neil exposes how these models encode societal prejudices into code, prioritizing efficiency and profit over fairness. She also highlights hopeful alternatives: algorithms designed to uplift rather than exploit, rooted in transparency, accountability, and a commitment to serving human needs. A must-read for anyone seeking to understand how data shapes power, inequality, and justice in the digital age.

Brian Christian

4.6421 learners

From the rise and fall of Frank Rosenblatt’s Perceptron in the 1950s to DeepMind’s groundbreaking AlphaGo, Brian Christian’s *The Alignment Problem* dissects the critical rift between artificial intelligence systems and human values. Weaving insights from psychology, neuroscience, and computer science, this narrative unpacks why AI often acts against our intentions: from algorithmic bias that misclassifies Black faces as gorillas to black-box medical models that learn dangerous hidden correlations. Christian explores reinforcement learning’s reward design pitfalls, the limits of imitation learning, and the urgent need for AI to recognize its own uncertainty. Real-world case studies—like Amazon’s biased hiring tool and ProPublica’s COMPAS investigation—highlight the tangible harms of misaligned AI. More than a tech deep dive, this book frames alignment as a moral and societal imperative, offering a roadmap to bridge the gap between machine goals and human priorities in our AI-driven future.