
2026上海苏河湾大会圆桌论坛探讨AI产业进展的衡量标准与投资逻辑
全球财富管理论坛·2026上海苏河湾大会在沪举行。与会专家探讨AI产业化进展,指出行业已从“讲故事”进入价值兑现阶段,判断标准正从单一榜单转向市场化评估与产业链深度融合。
AI-generated summary
全球财富管理论坛·2026上海苏河湾大会于10月16日至17日召开,重点讨论AI技术与产业应用的深度结合。
中新网上海10月17日电 (高志苗)全球财富管理论坛·2026上海苏河湾大会16日到17日召开。在17日举办的“AI——从技术前沿到产业应用”圆桌论坛上,嘉宾们的讨论没有停留在“AI能做什么”,而是反复追问一个更根本的问题:判断AI产业进展的标准究竟是什么?
上海人工智能研究院院长、联合国人工智能独立国际科学小组成员宋海涛以“加速、知识平权、全球化、适度性”四个关键词概括过去一年AI的变化,并指出检验方式本身也在改变。“过去我们还需要通过榜单去自我证明,现在大量的市场和客户实际上已经学会了用模型来评估我们的方案和技术能力。”在他看来,这种从“自我证明”到“市场用脚投票”的转变,本身就是AI走向产业化的一个信号。
君联资本总裁李家庆则从产业与投资视角提出另一重检验尺度。他认为,讨论“大模型能不能挣钱”本身是伪命题,因为大模型会贯穿整个产业,是端到端产业链条中必须具备的要素,脱离这一要素去谈企业应用或基础设施没有意义;产业结构本身在分层演化,“今年你多赚点钱,明天你的那份钱就得交给我”,各方在分层结构中动态调整、向下游延伸。
面对“大家All in会不会太拥挤”的问题,李家庆主张以投资组合方式覆盖芯片、基础设施、大模型、产业应用,并以“养成类”心态长期陪伴。他特别提醒,非市场化因素会导致长期逻辑被短期化,使原本波动不大的事情剧烈震荡,因此“长期资本”“耐心资本”之外更要强调“求真务实”,不要把长跑当成短跑。
面壁智能联合创始人、董事长、首席执行官李大海以“技术大爆炸”描述当前阶段:大模型知识密度约每100天提升一倍,端侧模型一年内从1B跃升至35B可上手机运行,AI开始参与训练AI、解决人类尚未解决的数学问题。他认为模型能力“发散和收揽同步发生”,技术长期价值容易被低估、短期预期容易被高估,正是行业焦虑的根源。“大模型给整个社会带来的增幅是高度确定性以及在加速的,我认为它是一个持续在发生的过程。”
AI产业化已从“讲故事”阶段进入可检验的价值兑现阶段,而落地的检验标准并非单一维度,是多重尺度相互校准的结果。

Clark, co-founder of Anthropic, said at the IMF and World Bank annual meetings in Bangkok that the rapid development of AI technology and the improvement of scientific exploration capabilities are expected to push the US GDP to US$44 trillion by 2030, and emphasized the need to establish international norms.

The South China Sea Navigation Support Center of the Ministry of Transport of China carried out real-ship sea trials of the S-100 maritime data model at the Pearl River Estuary, and jointly tested a number of data products with the Shenzhen and Hong Kong maritime departments, aiming to promote the upgrade of shipping in a safe, green and intelligent direction and prepare for the 2029 International Maritime Organization standard switch.
Ulanqab, Inner Mongolia, has successfully transformed into an important AI computing power cluster in the Asia-Pacific region by taking advantage of its wind, solar, green power resources and climate advantages. The city has attracted leading companies such as Huawei and Apple to settle in, and is moving from simple computing power leasing to the high end of the industrial value chain of the "Token Capital".

China's Ministry of Commerce and Ministry of Science and Technology have proposed export restrictions on 23 AI-related technologies. The move aims to tighten state control over the sector and impacts the global expansion plans of domestic AI developers.

Apple's new iPhone is about to be launched, and fraud groups are taking advantage of the opportunity to set up fake pre-orders, lottery draws and malicious websites. Trend Micro warns that some malicious websites will take advantage of known iOS vulnerabilities. Users who just open the web page without entering data may be attacked, and urges the public to update their systems and be vigilant.

Cisco's latest "Continuous Defense Report" points out that about 60% of information security professionals believe that insufficient organizational collaboration capabilities are the main obstacle to defense. The average score of Taiwanese companies is 59 points, and only 3% are "continuous defenders", facing multiple challenges such as cross-department coordination, data access, and AI introduction.