engblogs

Exploring Pinterest's implementation of offline Approximate Nearest Neighbors (ANN) for ad retrieval, highlighting cost efficiency, performance improvements, and suitable use cases compared to online ANN methods.

Discover how Salesforce Data Cloud's Bring Your Own Code (BYOC) architecture enables secure, scalable Python execution with tenant isolation, Kubernetes orchestration, and AI-enhanced developer productivity.

MCP servers standardize integration between AI models and external tools by enabling multiplexed, introspective APIs that enhance LLM capabilities beyond traditional request/response interactions.

Analyzing inductive biases in random neural networks through parameter-function maps and local volume measures to understand generalization and training dynamics.

Google DeepMind and Google Research launch Weather Lab with an AI-powered tropical cyclone model that improves prediction accuracy for cyclone formation, track, and intensity, collaborating with the U.S. National Hurricane Center and other meteorological experts to enhance early warnings and disaster preparedness.

A novel AI-driven technique enables rapid physical restoration of damaged paintings by printing a precisely aligned, removable polymer mask that applies digital restorations directly onto original artworks.

MIT researchers have developed a scalable and energy-efficient photonic processor that accelerates AI-based wireless signal processing for real-time 6G applications by performing machine-learning computations at light speed.

MIT's Social and Ethical Responsibilities of Computing initiative showcases interdisciplinary research advancing ethical AI, fair healthcare technology, digital democracy, and inclusive tech governance through innovative projects and community engagement.

Exploring the development and deployment of LACE, an LLM-based automated evaluation framework that enhances customer support chatbot performance through multidimensional assessment, iterative refinement, and agentic evaluation techniques.

A meta-analysis revealing that people’s acceptance of AI depends on the AI’s perceived capability surpassing humans and the need for personalization in different decision contexts.

Munther Dahleh's book details the creation of MIT's Institute for Data, Systems and Society, emphasizing a transdisciplinary approach to tackling complex societal challenges through the integration of data, physical systems, policy, and human factors.

A hybrid AI framework combining large language models and satisfiability solvers dramatically improves personalized trip planning by efficiently handling complex constraints and providing logical, verifiable travel itineraries.