MIT researchers have developed a wristband equipped with an ultrasound sticker that captures real-time images of wrist muscles, tendons, and ligaments as the hand moves. An AI algorithm translates these images into precise finger and palm positions, enabling wireless control of a robotic hand with high dexterity. Demonstrations include playing piano and manipulating virtual objects. The team plans to miniaturize the hardware, expand training data across diverse hand types, and ultimately build a wearable tracker for controlling humanoid robots or virtual environments in real time.
Nguồn: https://www.technologyreview.com/2026/06/23/1138279/ultrasound-imaging-turns-a-robot-hand-into-a-skillful-mimic. 8sync News chỉ tóm tắt và dẫn link; bản quyền nội dung thuộc tác giả và nguồn gốc.
Google Cloud vừa giới thiệu TPU Developer Hub, một nền tảng giáo dục tập trung dành cho nhà phát triển ML sử dụng TPU, bao gồm kiến trúc phần cứng, stack phần mềm (XLA, Pallas kernels), công cụ gỡ lỗi XProf, chiến lược tối ưu hóa (như offloading KV cache) cùng networking và bảo mật. Nội dung đa dạng từ Colabs tương tác, mã nguồn mở đến tài liệu chuyên sâu, hỗ trợ tích hợp AI-assisted development.
Lập trình viên ML nên đọc để hiểu cách tối ưu hóa hiệu suất và chi phí của mô hình trên TPU với các công cụ mới như XLA, Pallas và các chiến lược parallelism, từ đó tiết kiệm thời gian và nguồn lực trong triển khai sản phẩm AI.
AQSolotl and QuantrolOx have announced a strategic partnership to integrate AQSolotl's Chronos-Q quantum control hardware with QuantrolOx's Quantum EDGE machine learning-based automation platform. The goal is to automate qubit calibration, reduce manual tuning cycles, and improve qubit stability as quantum systems scale toward commercial deployment. The collaboration will proceed in two phases: near-term technical integration and performance benchmarking, followed by deeper hardware-software co-design and joint commercial offerings for research and enterprise customers.
General Intuition, a New York-based AI startup, has raised $320 million at a $2.3 billion valuation to scale AI agents trained on hundreds of millions of hours of video gameplay. The company's key differentiator is action-labeled gameplay data — records of button presses and timing — sourced from Medal, a gaming clip platform co-founded by CEO Pim de Witte. Unlike competitors inferring actions from video alone, General Intuition embeds this action data to train a single model capable of playing games, navigating simulated environments, and controlling physical robots. A quadruped robot was fine-tuned for real-world navigation using just eight minutes of street data. The round was led by Khosla Ventures, with participation from General Catalyst, Jeff Bezos, Eric Schmidt, and researchers from Google DeepMind and MIT. The company plans to offer its model via API and build a data flywheel across gaming, simulation, and robotics use cases, while explicitly ruling out lethal military applications.
General Intuition has raised $320 million at a $2.3 billion valuation to scale AI agents trained on hundreds of millions of hours of video game footage. The key differentiator is action-labeled gameplay data — records of button presses and timing — rather than video alone, which the company argues enables richer causal understanding. The same model powering a Fortnite-playing agent also drives a quadrupedal robot that required only 8 minutes of real-world fine-tuning. The round was led by Khosla Ventures with participation from Jeff Bezos, Eric Schmidt, and Google DeepMind researchers. General Intuition plans to sell its agentic model as a foundation for gaming, simulation, and robotics use cases via an API, while also launching Nerve, a jobs marketplace letting gamers earn income through data labeling and robot teleoperation.
A reproducible benchmark comparing gradient-boosted decision trees (GBDTs) vs. LLM-based scoring for payment fraud detection across three dimensions: latency, cost, and determinism. On a single CPU core, GBDTs hit p99 latency of 0.15ms vs. ~1,200ms for LLMs — well outside the 100ms ISO 8583 authorization budget. Cost-wise, GBDTs run ~$54/hour at 50K TPS vs. $16,200–$351,000 for LLM tiers. Determinism is the most critical issue for regulated environments: GBDTs return identical scores on identical inputs while LLMs produce hundreds of distinct outputs even at temperature=0. The recommended architecture keeps deterministic tree ensembles on the synchronous hot path and deploys LLM agents on the asynchronous cold path for SAR drafting, evidence gathering, and agent-as-a-judge validation before human review. All benchmark code is open-source and reproducible on a laptop.
Standard GPS receivers only achieve around two-meter accuracy, which is insufficient for precise robotic navigation. RTK (Real-Time Kinematic) GPS solves this by using a base station at a known location to transmit phase-angle correction data to a mobile receiver, enabling centimeter-level accuracy. GreatScott! demonstrates this on a tracked robot platform, placing the base station on a fence post and the RTK receiver on the robot. While the RTK system itself proved accurate enough, the robot's steering hardware and algorithms became the limiting factor for hitting centimeter-sized targets. The setup demonstrates practical applications like autonomous lawn mowing and amateur land surveying.
Hexora v0.3 is a Python library for detecting malicious PyPI packages using static analysis. The new release adds a gradient boosting machine learning model that analyzes code structure, semantic features, and static analysis results to assess entire Python files. The ML model's primary role is filtering false positives — previously yielding 5-10 false positives per real finding. Running against newly published PyPI packages, it now detects 2-10 malicious packages daily. Remaining false positives mostly come from AI-related projects that use dynamic code execution, base64-inlined assets, or telemetry.
Tombot has closed a $7 million Series A3 round to scale manufacturing of Jennie, its autonomous robotic Labrador puppy designed as a companion for people with dementia, cognitive impairment, anxiety, loneliness, autism, and PTSD. Investors include healthcare and aging-services backers such as Caduceus Capital Partners and the Lutheran Foundation for Long Term Living. The company reports over 23,000 pre-orders and waitlist sign-ups ahead of a planned Fall 2026 commercial launch — its first shipments to paying customers. Jennie mimics the behavior of an 8-to-10-week-old puppy and is pitched as delivering companionship benefits without the care burden of a live animal.