NEWS IN BRIEF: AI/ML FRESH UPDATES

Get your daily dose of global tech news and stay ahead in the industry! Read more about AI trends and breakthroughs from around the world

Mastering Gradient Descent with Momentum

Gradient descent struggles on surfaces with uneven curvature, but Momentum addresses this by using past gradients to stabilize updates and accelerate convergence. A simulation shows Momentum outperforming vanilla GD on an anisotropic surface, highlighting its effectiveness in optimizing oscillating gradients.

Revolutionizing Browsing with OS Level Actions

AgentCore Browser introduces OS Level Actions, enabling AI agents to interact with native UI elements outside the browser's web layer. This capability allows agents to observe, reason, and act on content visible on the screen, enhancing automation workflows.

Protecting AI Agents: Amazon Bedrock AgentCore Identity

Amazon Bedrock AgentCore Identity ensures secure access for AI agents on Amazon ECS with Authorization Code Grant, session binding, and scoped tokens. This solution maintains an auditable chain from user authentication to agent action, providing user consent and limited permissions.

Revolutionizing Training and Inference with TSP: 2.6x Faster Performance

Zyphra introduces Tensor and Sequence Parallelism (TSP) for large transformer models, reducing per-GPU memory usage in benchmark tests on up to 1,024 AMD MI300X GPUs. TSP combines Tensor Parallelism (TP) and Sequence Parallelism (SP) to optimize memory management, offering a new approach to parallelism folding for improved efficiency.

Efficient Matrix Pseudo-Inverse Refactoring in C#

Machine learning offers various techniques for training linear models, such as stochastic gradient descent and pseudo-inverse algorithms like relaxed Moore-Penrose and left pseudo-inverse via normal equations. The Cholesky decomposition technique for left pseudo-inverse is simpler but can be vulnerable to poorly conditioned matrices, making it crucial to understand the pros and cons of each met...

Fixing Tokenization Drift

Tokenization drift occurs when small formatting changes lead to unpredictable shifts in model behavior. Leading spaces create different token IDs, impacting attention computation and model performance.

Unlocking the Mysteries of Sleep with Beacon Biosignals

Beacon Biosignals, founded by Jake Donoghue PhD ’19 and former MIT researcher Jarrett Revels, uses EEG technology to monitor brain activity during sleep at home. The company's FDA-cleared device has been used in over 40 clinical trials globally to study conditions like major depressive disorder and Alzheimer’s disease.

Mastering the Power of Language

MIT senior Olivia Honeycutt's research focuses on the intersection of human thinking, language learning, technology, and social group interaction. She explores how language shapes our perception of the world and ourselves, delving into areas like neurolinguistics and AI at MIT.