摘要：Artificial intelligence (AI) technology is evolving rapidly and has great potential. Howcan we leverage this to improve production

Artificial intelligence (AI) technology is evolving rapidly and has great potential. Howcan we leverage this to improve production efficiency? How should AI be integrated with industry?

LI Zixue

Chairman, ZTE Corporation

At present, AI technology is rapidly iterating, and its general-purpose applications are spreading. However, challenges remain in deeper, industrial integration, including issues like poor usability, complexity, and a lack of proper training.

For the issue of poor usability, AI should evolve from general-purpose to specialized applications. On top of general models, industry-specific needs should drive bespoke customization and deep training, turning large models into "skilled artisans" that understand the specificities of each industry. For instance, ZTE has specialized models tailored to water conservancy and city infrastructure scenarios that are based on its Nebula general-purpose model. The models' monitoring and early warning accuracy have improved by 35% and their scene recognition precision has reached over 90%.

To address complexity, the focus should be on enabling low-cost, efficient, and secure deployment of large models. This requires providing open platforms compatible with GPUs from various manufacturers, deeply integrating hardware and software to maximize computing power, and supporting localized deployments to ensure data security. ZTE's AiCube integrated machine features a high-performance architecture and is fully compatible with leading models like DeepSeek, with factory-integrated solutions that allow devices to be used straight out of the box, greatly lowering enterprises' AI adoption threshold.

To address a lack of proper training, companies should start small, expand gradually, and then scale across their entire operations. Initially, firms should focus on specific scenarios, familiarize their staff with AI, cultivate talents, optimize processes, then gradually extend AI to production, R&D, and other areas. ZTE began with smart manufacturing and later expanded into operations, R&D, and marketing, improving overall operational efficiency by 15% and R&D efficiency by over 10%, while accumulating more than 110 demonstration applications and multiple industry-specific large models. Through these methods, AI can truly be integrated into firms and accelerate intelligent upgrades.

来源：澎湃新闻客户端