Responsible AI

Introduction to responsible AI (RAI): Need for ethics in AI, AI Risks, Risk sources/Taxonomy Decomposition of Risk. Familiarizing the terms- Weaponized AI, Proxy Gaming, Deceptive Alignment, Persuasive AI Introduction to Principles of RAI - Transparency; Accountability; Safety, Robustness and Reliability; Privacy and Security; Fairness and bias, Interpretability.

Introduction to Robustness- Black Swans, Long Tailed Distributions, Mediocristan and Extremistan, Unknown Unknowns, Test Robustness to Adverse Inputs, Improving Long Tail Robustness Adversarial Robustness: Overview (Adversarial Attacks: Fooling a binary classifier, Fast Gradient Sign Method, Projected Gradient Descent, Text based attacks; Adversarial Defenses: Data and Parameters, Data Augmentation, Adversarial Training), Distribution Shifts, Data Poisoning: Trojan Attacks, alignment problem

Introduction to Bias, Categories of Bias, The Source of Bias, Measure bias- Benchmark datasets, Remove bias-Debiasing methods, Preprocessing, inprocessing and postprocessing to remove bias. Group fairness and Individual fairness, Counterfactual fairness, Fair Supervised learning

Privacy in ML, Cryptographic solution Vs Statistical solution, Randomized Response, Trusted Curator, Differential Privacy, Laplace Mechanism, Approximate Differential Privacy. (Attack models, Privacy-preserving Learning, Differential privacy, Federated learning) Interpretability / Transparency / Probing: Pixel attribution methods- Saliency Maps, Saliency Maps for Text, LIME, SHAP, Probing.

RAI Regulation landscape: DPDP act (India), GDPR (EU), EU AI act, US presidential declaration, Ethical approvals, informed consent, participatory design. Real-world Applications RAI in Legal domain, RAI in Health care domain, RAI in Education domain, other domains