This Week’s Tech Highlights: AI Research Breakthroughs, 5G Surge, Multi‑Cloud DB & More

JD AI Research Papers Selected for Interspeech 2020

Four papers from JD Intelligent Cloud were accepted at the top‑tier speech conference Interspeech 2020, covering sound event detection, speech dereverberation, speaker verification, and neural vocoder technologies. Interspeech is the premier global forum for speech signal processing, held annually and this year marked its 21st edition in China with the theme "Cognitive Intelligence for Speech Processing".

Shenzhen Supercomputing Center Plans 1000× Power Increase

The National Supercomputing Shenzhen Center announced a plan to boost its computing capability by at least a thousandfold within two years, targeting 2022. The upgrade will support Bay Area scientific research, cloud computing, big data, and AI workloads, and includes development of an E‑level supercomputer and expansion of its campus and ecosystem.

T‑Mobile Expands Mid‑Band 5G Coverage

T‑Mobile announced a near‑doubling of its 2.5 GHz mid‑band 5G network coverage, now available in about 410 U.S. cities and towns. The carrier combines low‑band, mid‑band, and mmWave 5G to improve speed and reach, aiming to serve 100 million users by the end of 2020.

Apple Developing A14T Processor for iMac Using 5 nm Technology

Apple revealed plans to launch an A14T processor for the next‑generation iMac, fabricated on TSMC's 5 nm process. This follows the company's broader transition to custom ARM‑based silicon across its Mac lineup, with the new processor and GPU expected to debut in early 2021.

MongoDB Atlas Adds Multi‑Cloud Cluster Support

MongoDB Atlas now enables fully managed multi‑cloud database clusters, allowing users to run MongoDB instances across multiple cloud providers simultaneously. This feature enhances high availability, backup, and analytical workloads, addressing enterprise demand for flexible, cross‑cloud deployments.

Wikimedia Foundation Migrates Codebase to GitLab

The Wikimedia Foundation decided to move its code repositories from Gerrit to a self‑hosted GitLab Community Edition, citing developer dissatisfaction with Gerrit and a desire for a more open, flexible platform that aligns with the foundation’s open‑source principles.

GPT‑GNN: Generative Pre‑Training for Graph Neural Networks

GPT‑GNN addresses the scarcity of labeled data by leveraging massive amounts of unlabeled data through self‑supervised pre‑training. This approach enables graph neural networks to learn useful representations from raw graph data, reducing the need for extensive manual annotation.

Deep Clustering in the Era of Graph Neural Networks

Clustering, a fundamental task in machine learning and data mining, has evolved from traditional methods to deep clustering and now to graph‑neural‑network‑enhanced clustering. GNNs capture structural information, offering significant performance gains for clustering tasks involving relational data.