This Week’s Must‑Read Tech & AI Highlights: From Digital Currency to Cutting‑Edge Research

Developer Community Tech Weekly brings the latest news and research highlights for developers.

Digital Currency 200 CNY Red Packets on JD.com

Digital RMB is now available in Beijing via the JD app. Users can search “digital RMB” and register for a lottery from Feb 7‑8, limited to 50,000 red packets of 200 CNY each. Winners will be notified by SMS and can spend the red packets online from Feb 10‑17. Supported banks: ICBC, ABC, BOC, CCB, BoCom, PSBC.

AI Data Collection & Annotation Industry Trends 2021

YunCe Data released “2021 AI Data Collection and Annotation Industry Four Trend Forecast”. It predicts high‑precision data will remain a hot demand in AI training, scenario‑driven data will grow over the next 3‑5 years, and “underlying technology + service capability” will become core competitive factors, requiring integrated data solutions.

Google Drive Merges Enterprise and Personal Services

Google plans to integrate its enterprise and personal Drive services into a unified desktop client later this year, simplifying file synchronization and management for users.

Microsoft Launches Windows 10 Cloud Config

Microsoft announced Windows 10 Cloud Config, allowing unified configuration deployment for Windows 10 devices, supporting Win32 and line‑of‑business apps, Teams, Edge, and OneDrive for Business.

NVIDIA Achieves Real‑Time SDF Rendering Speedup

NVIDIA and collaborators introduced a neural geometry detail method that enables real‑time rendering of implicit 3D shapes via signed‑distance fields, achieving a 2‑3 × 10⁰ speed increase while maintaining state‑of‑the‑art geometry reconstruction quality.

NVIDIA A100 Deep Learning Performance Benchmarks

Benchmarks show the A100 accelerator can train up to 3.5 × faster than V100, with FP16 mixed‑precision delivering up to 2 × performance gains and TF32 offering 20 × over Volta.

AAAI 2021: DeepAligned for New Intent Discovery

DeepAligned uses a small set of known intent priors to transfer knowledge via deep alignment clustering, producing high‑quality intent representations and improving new intent discovery on benchmark datasets.

ACM MM: SST‑EmotionNet for EEG‑Based Emotion Recognition

SST‑EmotionNet employs a dual‑stream architecture to capture spatial, frequency, and temporal EEG features with a spatio‑frequency‑time attention mechanism, achieving superior performance on SEED and SEED‑IV datasets and offering a general framework for multivariate physiological time‑series analysis.