Will 2019 Be the Golden Harvest for AI? Insights from iQiyi’s VP

iQiyi’s Vice President Xie Danming was interviewed by China Electronics Daily about the outlook for artificial intelligence in 2019. He asserts that AI will enter a "golden harvest" period, with widespread adoption across diverse application scenarios and continued empowerment of enterprises.

Key technology drivers

5G will accelerate the explosion of AR/VR use cases, enabling ultra‑high‑definition video, immersive interaction, and new entertainment‑technology convergence.

Voice AI is reaching maturity, allowing comprehensive deployment in video entertainment through improved speech recognition, synthesis, and semantic understanding.

Graph neural networks are emerging as a hot research focus, promising advances in relational reasoning despite current challenges in interpretability.

Investment trends

Capital is expected to shift toward application‑oriented AI startups. After years of building foundational layers (compute, data, sensors) and technical layers (frameworks, models), many AI capabilities are now mature enough for industry deployment. While infrastructure‑focused startups face limited space, there remains abundant B2B and B2C opportunity for companies that combine deep industry knowledge with AI.

Challenges for AI algorithms and hardware

Balancing performance and cost: High‑accuracy models often require powerful GPUs, making deployment on mobile or edge devices expensive. Future work should emphasize model lightweighting, quantization, and compression to bring deep models to embedded platforms.

Lack of generality: Most AI solutions are task‑specific, requiring separate models and large labeled datasets for each application. Transfer learning, multi‑task learning, and reinforcement learning are needed to reduce this fragmentation.

Insufficient multimodal deep analysis: Current AI systems do not fully exploit multimodal data (vision, audio, text). Integrating these modalities can unlock richer video understanding for platforms like iQiyi.

AI‑generated content: While generative models (e.g., GANs) have produced impressive results, they still lack universal applicability and often generate artifacts. Ongoing research aims to improve quality and controllability.

iQiyi’s AI applications in entertainment

The company leverages AI across the content lifecycle:

Content creation: AI‑assisted script analysis, smart casting ("YiHui" system), and AI‑generated hand‑sign language avatars for the hearing‑impaired.

Production efficiency: Automated video editing, subtitle generation, and scene detection reduce post‑production time by roughly 20%.

Content understanding: Facial and object recognition, scene segmentation, and user behavior analysis enable personalized recommendations and interactive advertising.

Copyright protection: An AI‑enhanced digital rights management (DRM) system identifies infringing material and automates takedown.

Viewing experience: Features such as "only watch this idol," masked danmu, and real‑time up‑scaling (540p to 1080p, 8K VR playback) improve engagement.

These initiatives illustrate how AI is transitioning from experimental prototypes to core business capabilities, driving both technical innovation and commercial value in the entertainment sector.