Science – Page 68 – Minority Mindset

Taming Chaos: Tensor Networks Predict Turbulent Systems

13.11.2025 by qfx

A tensor network model efficiently encodes non-Markovian correlations within chaotic dynamics by hierarchically mapping sequences of past states—represented as input vectors with dimension $dd$ at each time step—to predicted future states through successive tensor contractions performed by rank-4 weight tensors.

A new approach leverages tensor networks to accurately forecast the long-term behavior of complex, chaotic dynamics.

Predicting State-Level Financial Distress with Privacy-Preserving AI

13.11.2025 by qfx

A federated learning model’s performance in predicting financial distress varies significantly across U.S. states, as evidenced by differing precision, recall, $F1$-score, and AUC values when each state operates as an independent client, revealing substantial regional heterogeneity in model effectiveness.

A new approach leverages federated learning and explainable AI to identify financial risks across U.S. states without sharing sensitive data.

AI Learns to Defend Itself: A New Era for Language Model Safety

13.11.2025 by qfx

Researchers have developed a system that allows large language models to dynamically improve their defenses against evolving adversarial attacks.

Eyes on the Sky: AI Spots Methane Leaks From Space

12.11.2025 by qfx

Hyperspectral satellite data, coupled with machine learning models, facilitates not only the detection of methane leaks but also verification of mitigation efforts, as demonstrated by instances where alerts triggered by the system prompted successful repairs and ongoing infrastructure improvements—a process indicating a shift from simple detection to proactive ecosystem management.

A new deep learning system is now operational, using data from multiple satellites to identify and quantify methane point sources across the globe.

Predicting the Future: How News Articles Foretell Planned Events

12.11.2025 by qfx

A new approach leverages natural language processing to anticipate social unrest and planned gatherings by analyzing news coverage.

AI Underwriting: Governing Large Language Models in Reinsurance

12.11.2025 by qfx

A new framework for ensuring the reliability, transparency, and regulatory compliance of AI models used in critical reinsurance functions.

When AI Stumbles: Lessons from Real-World Inference Failures

12.11.2025 by qfx

The distribution of incidents reveals a correlation between model modality and severity, suggesting that certain approaches are disproportionately associated with more critical failures.

A detailed analysis of production incidents reveals critical vulnerabilities and practical strategies for building more reliable AI services.

The Prediction Machine: Can AI Rival Human Forecasters?

12.11.2025 by qfx

The AIA Forecaster leverages a defined architecture to facilitate predictive capabilities.

A new system leveraging artificial intelligence achieves forecasting accuracy comparable to top human prediction experts.

The Coming Infrastructure Crisis: Can Networks Handle the AI Boom?

12.11.2025 by qfx

The study forecasts agent proliferation over the coming decade, revealing that while initial growth may appear exponential, resource limitations and inherent system dynamics will ultimately enforce a logistic or, more likely, a Gompertzian curve of deceleration and eventual saturation—a prophecy written into any architecture of scaling.

As AI agents and connected devices multiply, existing infrastructure faces unprecedented strain, demanding a new approach to network design.

Predicting Bitcoin’s Price Swings: Can Classic Models Keep Up?

12.11.2025 by qfx

Autocorrelation and partial autocorrelation functions applied to the Bitcoin price series reveal the inherent temporal dependencies within the cryptocurrency's valuation, offering insights into its dynamic behavior and potential predictability through time-series analysis.

Despite its reputation for chaos, Bitcoin’s price and volatility can be meaningfully forecast using established time series techniques.