Expert Round‑Table on AIGC: Technology vs. Market Beliefs, Domestic Model Challenges, and Enterprise Deployment in China

In early 2024, as the first year of AIGC (AI‑generated content) deployment, a series of pressing questions were raised: should the community cling to a technology‑centric belief while pursuing AGI, how to move beyond merely copying foreign successes like GPT‑4 and Sora, what algorithmic hurdles remain, and why many Chinese firms deploy multiple large models.

The 51CTO live interview "AIGC实战派" gathered three experts—Baruan (chief data scientist at a gaming company), Zhu Lei (co‑founder of YuanYu Intelligence and community leader), and Xiao Ran (ThoughtWorks China GM). They answered the above questions, sharing perspectives on technology versus market belief, model development, and practical adoption.

Baruan argued that, in the long run, a technology‑first stance drives productivity, though short‑term market hype may overestimate impact. Zhu Lei echoed a strong technology belief, emphasizing massive data training and talent competition, while acknowledging that technology must ultimately serve concrete business problems.

The panel highlighted key differences between domestic and foreign large‑model ecosystems: Chinese firms face compute constraints due to foreign chip dominance, rely on imported frameworks like Transformers, yet benefit from massive domestic data, especially in vertical domains. They also noted that many domestic companies are still replicating foreign models rather than pioneering new architectures.

On enterprise adoption, the experts observed that Chinese companies often purchase two or three different models and fine‑tune open‑source alternatives, using vector databases and augmentation tools. Deployment costs are falling, but operational expenses and talent requirements remain high. Organizational responsibility typically falls to CIO/CTO, with emerging roles such as Chief AI Officer (CAO) overseeing AI strategy.

Looking ahead, the discussion turned to how to judge AIGC applications: compliance, user willingness to pay, distinctive data‑driven advantages, and engineering‑level scalability are essential. The panel concluded that while AIGC will become a new productive force, not every enterprise needs to host its own models; the real competitive edge lies in how effectively organizations integrate AI tools into existing workflows.