Robotics

Building Privacy-Preserving Federated Pipelines for LLMs with LoRA

Key TakeawaysFederated learning enables privacy-preserving collaborative learning across distributed systems. LoRA can effectively fine-tune large language models, improving efficiency and accuracy. Using Flower enhances the construction...

Microsoft’s OrbitalBrain: Revolutionizing Space-Based Machine Learning

OrbitalBrain — Key TakeawaysOrbitalBrain employs nanosatellite constellations for efficient model training in space. Traditional BentPipe satellite systems are limited by data relay to ground stations. The...

Smart Synthetic Skin: A Breakthrough in Adaptive Camouflage

Key TakeawaysResearchers developed a smart synthetic skin capable of hiding and revealing images, inspired by cephalopods. This skin utilizes hydrogel, enabling shape and texture changes...

Discover OAT: Revolutionizing Robotics with Action Tokenization

Key TakeawaysOAT, developed by researchers from Harvard and Stanford, enhances robotics with scalable actions. It converts continuous robot movements into discrete tokens using a transformer...

Harvard Dropouts Launch AI Smart Glasses to Record Conversations

Key TakeawaysTwo former Harvard students introduced AI-powered smart glasses for $249 that can transcribe conversations. The glasses utilize Soniox technology for audio transcription, promising not...

Exploring Project Genie: Create Infinite Interactive Worlds

Key TakeawaysProject Genie allows users to create, explore, and remix interactive worlds. Genie 3 generates dynamic environments as users interact, setting it apart from static...

Exploring D4RT: AI’s Breakthrough in 4D Scene Reconstruction

Key TakeawaysD4RT is a unified AI model that excels in 4D scene reconstruction and tracking. The architecture is based on a transformer, which enhances processing...

Waymo Unveils Waymo World Model for Advanced Autonomous Driving Simulation

Key TakeawaysWaymo's World Model leverages Genie 3 to create realistic driving simulations. It generates high-fidelity camera images and lidar point clouds dynamically. The model exposes systems...