Top Tech News: R2DBC for Java, 4600‑km Quantum Network, AI Papers & More

MariaDB adds R2DBC connector for Java developers

MariaDB team introduced an R2DBC (Reactive Relational Database Connectivity) connector, offering a more powerful, efficient, and scalable JVM solution compared to traditional JDBC. R2DBC enables declarative programming with asynchronous data streams, improving resource utilization and throughput.

Pan Jianwei’s team builds world’s first integrated quantum communication network (4600 km)

China’s University of Science and Technology team achieved a 4600‑km quantum key distribution network, extending quantum communication from ground to space, marking a major breakthrough published in Nature.

JD’s 16 papers selected for AAAI 2021

Sixteen papers from JD’s research arm were accepted at AAAI 2021, covering federated learning, adversarial learning, deep learning, sequential and social recommendation, graph neural networks, causal inference for risk management, and spatio‑temporal AI for smart cities.

Intel launches RealSense ID facial recognition

Intel’s RealSense depth camera now supports AI‑driven facial recognition that works under varied lighting and skin tones, processes data locally with encryption, and resists spoofing, targeting high‑security applications such as ATMs and door locks.

JupyterLab 3.0 released

JupyterLab 3.0 brings major upgrades, including a multilingual UI (Chinese), a new visual debugger, improved simple mode, clearer documentation, and an enhanced extension system, facilitating data analysis and web‑based development.

JD smart community 2.0

JD unveiled Smart Community 2.0 with five new products (including “物应用”, cloud‑guard, super‑user ID) and a technology‑life service platform, addressing rapid service integration and multi‑device adaptation challenges.

OpenViDial: Large‑scale multimodal dialogue dataset

OpenViDial provides over one million sentence‑image pairs extracted from movies and TV shows, along with three multimodal models; experiments show that incorporating fine‑grained visual information improves dialogue quality.

FixMatch: Semi‑supervised learning via image augmentation (NeurIPS 2020)

FixMatch uses image augmentation for pseudo‑label learning and consistency regularization, achieving strong results on CIFAR and other datasets with few labeled images, beneficial for scenarios with scarce annotations.