(Yicai) March 10 -- Artificial intelligence will branch out into differentiated models, and finance will become a hub of early adopters of industry-specific AI, Chinese experts discussed during the Lujiazui Financial Salon.

The financial sector, with its high digitalization, strong acceptance of new technologies, and willingness to pay for innovation, is expected to emerge as a key arena for AI adoption, said Li Jing, vice president of Shanghai-based AI startup Stepfun, during a roundtable discussion at the new event series hosted by Shanghai's Pudong New Area.

Automobiles and smartphones will become the core drivers of innovation in AI applications and devices, Li noted, adding that generative AI is predicted to significantly improve professional content production.

Industry-specific vertical AI applications will continue to grow, according to Wei Zhongwei, board secretary of Shanghai-based MetaX Integrated Circuits. Demand in finance, transportation, education, and scientific research is rising rapidly, especially in transportation, where automakers' growing investments in data centers are expected to spur substantial AI application needs in the coming years.

Against this backdrop, Wei emphasized that the next two to three years will be critical for AI to accelerate its integration across industries. The versatility, stability, and reliability of AI technologies will become key benchmarks, requiring infrastructure providers to deliver high-quality products and services.

The financial sector prefers vertical models, said Yu Feng, chief information officer of Guotai Junan Securities. By leveraging proprietary data, fine-tuning strategies, and adjusting training objectives, firms can achieve differentiated competition, avoid homogenized investment strategies caused by general-purpose models, and mitigate risks of amplified market volatility, he added.

However, AI demands profound industry changes. Device makers need to grant greater access to deeply integrate AI capabilities, while third-party service providers need to redesign their frameworks under agent architectures. Policy support will also be crucial, Li from Stepfun said.

Regarding data privacy, Li stated that privacy protection is not an unsolvable challenge. "Technologically, we've already identified promising directions to explore."

Editor: Emmi Laine