Top Tech Breakthroughs: Open‑Source Graph AI, Moon 4G, Brain‑Inspired Computing

JD Retail Graph Computing Platform Galileo to be Open‑Source

Galileo is a distributed graph deep‑learning framework supporting graph embedding and end‑to‑end graph neural network training. It is already used internally for search, recommendation, knowledge graph, anti‑fraud, etc. The team plans to optimize operators and training, explore online‑offline integration, and expects to release the source code by year‑end.

LINE Negotiates with Asian Central Banks on CBDC Project

Japan’s messaging giant LINE is discussing a central‑bank digital currency (CBDC) platform based on the LINE blockchain with several Asian central banks, including strong support for South Korea’s CBDC development, marking a first large non‑financial firm backing a CBDC effort.

NASA Plans 4G Network on the Moon

NASA will allocate $14.1 million to partner with Nokia’s Bell Labs and Intuitive Machines to build the first 4G communication system on the lunar surface, scheduled for deployment on a lander by the end of 2022 as part of the Artemis program, with future upgrades to 5G.

Microsoft Teams Up with SpaceX, Challenging Amazon in Space‑Based Cloud

Microsoft announced collaborations with SpaceX and other partners to extend its cloud services into space, using low‑orbit satellites and higher‑orbit platforms to deliver new services, intensifying competition with Amazon’s cloud initiatives.

Nvidia Supplies Ampere GPUs for the World’s Fastest AI Supercomputer

Nvidia’s Ampere‑based A100 GPUs will power a planned AI supercomputer in Europe built by the CINECA consortium, comprising over 14 000 GPUs to deliver unprecedented artificial‑intelligence performance.

Intel‑Powered “PhiSat‑1” Satellite Demonstrates On‑Board AI Processing

The European Space Agency launched 46 small satellites, including PhiSat‑1 equipped with an Intel Movidius Myriad 2 VPU and hyperspectral thermal imager, enabling on‑board AI inference for polar ice and soil moisture monitoring.

MatchGNet: Heterogeneous Graph Neural Network for Unknown Malware Detection

MatchGNet models program execution behavior as a heterogeneous graph, applies attention mechanisms across nodes, layers, and meta‑paths to learn graph representations, and scores similarity to distinguish benign from malicious programs. (Paper: https://www.ijcai.org/Proceedings/2019/522)

Brain‑Inspired Computing Architecture Advances General AI

A Nature paper by researchers from Tsinghua University, Beijing Institute of Technology, and the University of Delaware proposes a “brain‑computing completeness” concept and a system hierarchy that bridges neuroscience‑driven and computer‑science‑driven approaches, aiming to accelerate brain‑inspired and AGI research. (Paper: https://www.nature.com/articles/s41586-020-2782-y)