Special Topic: Global Wealth Management Forum · 2025 Shanghai Suhe Bay Conference On October 18, financial news reports that the Global Wealth Management Forum · 2025 Shanghai Suhe Bay Conference is taking place in Jing'an District, Shanghai, from October 18 to 19. Xia Lixue, co-founder and CEO of C3.ai, Inc., attended and spoke at the conference.
Xia Lixue reflected on his journey since 2013, when he started researching the combination of artificial intelligence and hardware at Tsinghua University, during which he has witnessed rapid advancements in the AI field. “The developments in AI over the past decade can rival the entire 100-year history of AI, much like the technological advancements over the past 200 years have surpassed millennia of human history.” He believes that support from the capital market is a crucial condition for the rapid growth of the technology sector. Today, AI not only effectively serves industrial development but has also become a popular arena in the capital market. For industries to thrive, a closed-loop must be formed: “funding into technology—science transforming into technology—technology creating quality products and industries.” AI is providing robust support for industrial upgrades, and its value potential is highly anticipated.
Discussing the current explosion of AI technology, Xia pointed out that the previous two AI booms laid the foundational scientific principles. Current widely used technologies like large models and Attention derive their underlying methods—such as backpropagation and neural networks—from earlier waves. The key to today's breakthroughs lies in the collaborative drive of three core elements of AI development: algorithms, data, and computing power. The earlier waves struggled due to a lack of GPU chips, limiting large-scale model exploration and only showcasing the “tip of the iceberg” of algorithm capabilities. Now, vast amounts of data act as “fuel,” while powerful computing and systems serve as the “engine,” collectively facilitating a closed-loop growth for AI. This phenomenon is especially evident in the third AI boom.
Xia further illustrated this by referencing the development of OpenAI: After the release of the paper on Attention in 2017, it wasn't until the introduction of GPT-3.5 in 2022 that AI truly garnered widespread public attention and significant capital investment. During this period, OpenAI continuously invested in building larger clusters and conducting large-scale training, ultimately discovering scaling laws and emergent capabilities. “Currently, major companies are still focused on building larger clusters, attempting to achieve significant leaps in intelligence, underscoring the importance of foundational computing power for AI development.”
Regarding the ultimate vision for artificial intelligence, Xia stated that the industry aims to create “workers” that resemble humans, but current models are still two generations away from achieving this goal. Referencing OpenAI’s five-tier classification of AI (comparable to the stages of cultivating an excellent expert: learning professional knowledge, undergraduate and graduate education, internship, industry creation, and management collaboration), current AI technology is situated between levels L2 and L3.
This phase holds significant importance—the AI technology has reached a turning point in commercialization. “It's akin to hiring employees; at this juncture, AI possesses some capability to independently complete tasks and has the foundational elements to generate more value for industries, representing an excellent opportunity for application across various sectors.” Xia emphasized that the exploration from reasoning models to agent models is ongoing, but challenges remain: when AI needs to coordinate resources, schedule tasks, iterate thoughts, and even call tools or write code, it incurs higher costs. The “token assassin” phenomenon reflects the pressures of cost budgeting, so the industry urgently requires an optimization capability of tenfold or more to achieve cost control and closure throughout the supply chain.