Boosting AI Efficiency in Digital Content Risk Control: Insights from QCon

As AI technology expands, reducing costs and simplifying deployment become critical challenges. At QCon 2022 Shanghai, NetEase Zhiji algorithm specialist Li Yuke discussed how his team enhances AI service efficiency in digital content risk control.

Interview Introduction

Li Yuke : I lead AI‑based digital content risk control at NetEase Yidun and head the AI sub‑committee of the NetEase Zhiji Technical Committee.

Role of the Technical Committee

The committee breaks down technical roles across business units to improve alignment and combat duplication. It promotes API reuse, shares solutions, drives department‑wide initiatives, facilitates knowledge sharing, and provides a holistic view of technology adoption.

AI Solutions Across Three Domains

In content risk control, algorithmic tuning improves recognition of complex scenarios. For audio‑video, AI‑driven denoising, scene analysis, and super‑resolution enhance user experience. In natural language dialogue, feature‑retrieval optimization and domain‑specific frameworks address intricate customer needs.

Industry‑Specific Risk Control

NetEase Yidun supports varied industries by gathering sector data, building specialized capabilities (e.g., for minors or fraud detection), and forming cross‑functional teams that stay close to customers, enabling flexible, informed solutions.

Core Advantages of NetEase Yidun

Deep industry expertise, long‑term commitment, and a customer‑centric approach that goes beyond POC to continuous collaboration ensure robust, adaptable AI services.

Data Standards in Risk Control

Standardizing data helps machines learn objective patterns; over time, models can handle complex standards, reducing reliance on manual labeling and improving scalability.

Future AI Frontiers

Multimodal AI can boost performance but adds cost, so practical timing is key. Unsupervised and large‑scale models aid data generation and downstream tasks but still need careful integration. Engineering automation improves efficiency but must balance flexibility and require skilled users.

Personal Outlook

Li plans to focus on data science to reduce manual labeling, streamline training for higher efficiency, and enhance service‑level AI efficiency to support diverse AI offerings.