The Self-Fulfilling Prophecy of AI
Machine learning models aren’t passive observers – they’re increasingly active participants in the world they predict, creating feedback loops that reshape reality.
Science
AI’s Rising Tide: Can Generative Models Deliver Social Good?
This review explores how advances in generative AI are poised to overcome critical hurdles in deploying artificial intelligence for positive social impact.
Predicting Market Moves: A New AI Approach to Stock Correlation
Researchers have developed a deep learning model that leverages the power of graph neural networks and transformers to more accurately forecast how stocks move in relation to each other.
Beyond the Metrics: Testing the Limits of Machine Learning Fairness
A new framework systematically evaluates how changes in underlying data relationships impact the reliability of fairness interventions in machine learning systems.
Modeling the Human Response to Pandemic Threats
Researchers are using the power of artificial intelligence to simulate how individuals adopt preventative behaviors during an epidemic.
Closing the Security Gap: AI-Powered Post-Incident Analysis
New research demonstrates how artificial intelligence can automate the critical process of identifying security policy weaknesses after a cyberattack.
Seeing Beyond the Flood: AI-Powered Crop Damage Mapping
![The FLNet model establishes a novel architecture for [latex]f(x) = w^T x[/latex], enabling efficient and scalable feature learning through a learned network of weights, <i>w</i>, and input features, <i>x</i>.](https://arxiv.org/html/2601.03884v1/x1.png)
A new deep learning pipeline leverages freely available satellite data to pinpoint flood damage at the individual farm level, offering a cost-effective solution for disaster response and agricultural monitoring.
Sensing Mental Wellbeing: A New Era of Forecasting
Research reveals how data from our smartphones, combined with advanced artificial intelligence, can predict changes in mental health with increasing accuracy.
Unraveling Language Model Collapse
![The study demonstrates that compounding data and weight recursion-where each generation of synthetic text refines training from the previous generation’s weights-results in measurable drift, quantified as the change in [latex]\Delta U_{\mathrm{LLN,cov}}(\delta)[/latex] and [latex]\Delta G_{\mathrm{KF}}(\delta)[/latex], relative to a baseline checkpoint.](https://arxiv.org/html/2601.03385v1/x2.png)
New research offers a powerful framework for understanding and predicting when large language models begin to lose coherence during prolonged, synthetic data training.
The Tipping Point of Finance: How Network Structure Drives Systemic Risk
New research reveals that financial networks exhibit surprisingly fragile stability, with even modest increases in interconnectedness potentially triggering cascading defaults.