Resources

OOD and Malicious Behavior Detection

• Deep Anomaly Detection with Outlier Exposure
• A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks
• ViM: Out-Of-Distribution with Virtual-logit Matching
• VOS: Learning What You Don’t Know by Virtual Outlier Synthesis
• Scaling Out-of-Distribution Detection for Real-World Settings
• A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks

Interpretable Uncertainty

• On Calibration of Modern Neural Networks
• Can You Trust Your Model’s Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift
• PixMix: Dreamlike Pictures Comprehensively Improve Safety Measures
• Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
• Posterior calibration and exploratory analysis for natural language processing models
• Accurate Uncertainties for Deep Learning Using Calibrated Regression

Transparency

• The Mythos of Model Interpretability
• Sanity Checks for Saliency Maps
• Interpretable Explanations of Black Boxes by Meaningful Perturbation
• Locating and Editing Factual Knowledge in GPT
• Acquisition of Chess Knowledge in AlphaZero
• Feature Visualizations and OpenAI Microscope
• Exemplary Natural Images Explain CNN Activations Better than State-of-the-Art Feature Visualization
• Network Dissection: Quantifying Interpretability of Deep Visual Representations
• Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead
• Convergent Learning: Do different neural networks learn the same representations?

Trojans

• Poisoning and Backdooring Contrastive Learning
• Universal Litmus Patterns: Revealing Backdoor Attacks in CNNs
• Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks
• TrojAI
• Detecting AI Trojans Using Meta Neural Analysis
• STRIP: A Defence Against Trojan Attacks on Deep Neural Networks
• Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
• BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain

Detecting and Forecasting Emergent Behavior

• The Effects of Reward Misspecification: Mapping and Mitigating Misaligned Models
• The Basic AI Drives
• Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets
• Optimal Policies Tend to Seek Power
• The Off-Switch Game

Honest AI

• TruthfulQA: Measuring How Models Mimic Human Falsehoods
• Truthful AI: Developing and governing AI that does not lie

Machine Ethics

• What Would Jiminy Cricket Do? Towards Agents That Behave Morally
• Ethics Background (Introduction through “Absolute Rights or Prima Facie Duties”)
• Aligning AI With Shared Human Values
• Avoiding Side Effects in Complex Environments
• Conservative Agency via Attainable Utility Preservation
• The Structure of Normative Ethics

‍

Forecasting

• Forecasting Future World Events with Neural Networks
• On Single Point Forecasts for Fat-Tailed Variables
• On the Difference between Binary Prediction and True Exposure With Implications For Forecasting Tournaments and Decision Making Research
• Superforecasting – Philip Tetlock

ML for Cyberdefense

• Asleep at the Keyboard? Assessing the Security of GitHub Copilot’s Code Contributions

Cooperative AI

• Uehiro Lectures 2022
• Open Problems in Cooperative AI
• Cooperation, Conflict, and Transformative Artificial Intelligence: A Research Agenda