North America Minder Forum Debates AI: Moving from Model Capability to Commercial Value, Human-Machine Collaboration Still Needs Clear Boundaries

Deep News11-15

On November 14 local time, the 12th North America Minder Forum 2026 Annual Meeting, hosted by the Renmin University of China North America Alumni Association, was held at Columbia University in New York, with the theme "New Intelligence Order: The World, Capital, and Humanity." Sina Finance was a partner of this year's forum.

During the "Artificial Intelligence" roundtable discussion, Wang Hanrui, Senior Vice President of Nebius, Kathryn Zhao, Global API Product Management Lead at OKX, Qin Yongming, AI Product Lead at SiteIQ Technologies, and Wang Danye, Founder and Managing Partner of UniGlobe Capital, exchanged views on AI commercialization, agent applications, AI in the physical world, and entrepreneurial investment. The roundtable was moderated by Zhang Shu, COO of Yinglian Capital and Founder and CEO of Allin AI.

How can continuously improving model capabilities be translated into corporate returns? Wang Hanrui believes that enterprises need to choose appropriate models based on specific tasks; applications such as daily summarization and email processing may not require the most powerful general-purpose models. Adapting and optimizing models for specific scenarios can reduce usage costs. He particularly noted that understanding the decline in AI costs should compare the expense required to achieve the same level of intelligence; as frontier model capabilities strengthen, they will also open up new application spaces and drive more demand. He stated that AI services are extending from providing data centers and computing power to models, software, and inference services, and what customers ultimately need is usable intelligence. He is also optimistic about optimization approaches involving multi-model division of labor with repeated generation and evaluation, applying them to tasks such as low-level computing programs to improve R&D efficiency.

From the perspective of financial markets and corporate practice, Kathryn Zhao pointed out that AI agents are gradually becoming more active market participants, and financial infrastructure needs to serve both humans and agents. The significance of application programming interfaces (APIs) has also extended from being connection tools to platforms supporting automated research, decision-making, and execution. In this process, permission controls, accountability attribution, and risk constraints must be established simultaneously. According to her, her team introduced agent workflows in API product management, shortening delivery cycles by approximately 50% to 60%. This change comes from a complete redesign of the process: based on a knowledge base, different agents handle requirements analysis, product documentation, code development, and testing, with human review retained at key stages. She emphasized that the value of AI should be measured by what problems it actually solves, not just by how much code or documentation is generated or how many tokens are consumed.

Qin Yongming steered the discussion toward the physical world. He believes that compared with processing language, robots also need to understand environments, perceive force and contact, and take action under real-world constraints, thus requiring richer perceptual capabilities. As software development efficiency improves, challenges in hardware and the real environment become more prominent. He looks forward to AI and robots entering scenarios such as home care, daily chores, and physical space operations to help people reduce their burdens.

From a capital perspective, Wang Danye believes that AI tools enable startup teams to advance product development with fewer resources and complete more customer and market validation before fundraising. Investment judgment still needs to return to the entrepreneur's understanding of the problem, execution capability, and long-term vision. For venture capital funds, she particularly focuses on their ability to access excellent talent, win quality projects, and continuously support portfolio companies. She also cautioned that the timeline for breakthroughs in areas such as physical AI remains uncertain, and investors need to match corresponding fund terms. Family offices participating in AI investment should establish a professional asset allocation framework, weighing security, liquidity, and returns, and avoid being driven by market enthusiasm, since venture capital also requires long-term capital patience.

When asked what things they are currently still unwilling to hand over to AI, the guests gave their own boundaries: Kathryn Zhao is temporarily unwilling to let AI directly operate personal bank accounts and make investment decisions on her behalf; Wang Hanrui emphasized that high-risk tasks such as chip design still require repeated review; Qin Yongming hopes to retain the choice of communicating with others in person. Wang Danye stated that nothing can replace face-to-face communication, and building trust and relationships still requires personal involvement.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

We need your insight to fill this gap
Leave a comment