As generative AI emerges as a prevailing trend and a standard component, the industry is stepping into an entirely novel phase of development. For every participant in this field, a pivotal question looms large: how to propel the further evolution of AI and unlock its vast potential. Recently, Yang Yuanqing, the Chairman and CEO of Lenovo Group, engaged in a meaningful dialogue with Jensen Huang, the Founder and CEO of NVIDIA. Huang articulated that AI is transitioning from "generative AI" to "agentic AI," which boasts enhanced problem-solving abilities and the capacity for fact-based reasoning. Both Yang and Huang concurred that the core opportunity for future AI development resides in "hybrid AI."

This dialogue is undeniably in sync with the current trajectory of AI development. Ultimately, AI transcends the realm of mere intelligent chat tools; it is delving deeper into industries, achieving a more profound integration. For any entity aspiring to leave its mark in this phase, the crux lies in discovering ways to seamlessly integrate AI with industries and truly transform them at a fundamental level, thereby securing victory in the new cycle.

As AI embarks on a new cycle, particularly as it increasingly tests the commercialization prowess of players, identifying suitable development models and achieving true success in the new era of "hybrid AI" is paramount for propelling AI development into an unprecedented stage. So, what development trends will the AI industry exhibit in this novel phase?

AI Capabilities Continue to Evolve

Undoubtedly, the current capabilities of AI have witnessed a substantial enhancement compared to the past, be it in terms of underlying algorithms, computational power, or surface-level applications. As AI capabilities continue to surge, AI is no longer confined to the simplistic and mechanical tasks of yore; it can now undertake more intricate and demanding tasks. This evolution is vividly evident in the iterative history of humanoid robots. Initially, humanoid robots were akin to toddlers, but now they can not only run and box but also participate in dance performances at Wang Leehom's concerts.

As AI enters a new cycle, particularly after transcending the generative AI stage, AI capabilities are no longer limited to mere chatting and walking; they are commencing to think and dance. Hence, if we are to chart a new course for AI development in this cycle, the continuous evolution of AI capabilities undoubtedly warrants our primary attention.

For every entity aiming to make a mark in this phase, continuously bolstering their AI capabilities and enabling their AI to possess a wider array of functions that cater to diverse needs are undoubtedly crucial facets deserving of attention. Moreover, another aspect that merits our consideration is that AI capabilities are not confined to a single dimension but also encompass composite capabilities across multiple facets. Ultimately, the ability to tackle more complex and specialized tasks, rather than merely engaging in simple chat conversations, is the linchpin for ensuring that AI players can continue to achieve new milestones.

AI Begins to Transition from General-Purpose to Vertical Applications

Traditionally, the AI we are acquainted with often manifests as a large model or an intelligent agent that caters to the needs of a vast user base and a multitude of scenarios. As AI continues to evolve, particularly as this general-purpose AI encounters an increasing number of pain points and challenges, the question of how to meet the diverse and personalized needs of different users, industries, and scenarios, especially achieving harmonious development between general-purpose large models and personal intelligent agents, has emerged as a new challenge that more and more players must confront.

Ultimately, AI is beginning to leverage the advantages of general-purpose large models, such as their expansive boundaries, wide industry coverage, and diverse applications, to encompass different users, industries, and scenarios. Building upon this foundation, by deploying distinct intelligent agents for each, the synergy between the broadness of large models and the uniqueness of personal intelligent agents can be realized, thereby better catering to the personalized needs of diverse users, industries, and scenarios.

For every entity aiming to leave its mark in this phase, on the one hand, they must continually expand the breadth of intervention of their general-purpose large models across a vast user base, industries, and scenarios; on the other hand, they must consistently introduce personalized applications for different users, industries, and scenarios to meet their practical needs for data security and diverse demands. Ultimately, AI must transition from a previous state of broad and comprehensive yet rudimentary development to a current state of small-scale and refined development to cater to the needs of different industries, scenarios, and users, thereby unlocking more commercialization possibilities.

AI Begins to Foster Deep Integration Between Enterprise and Personal Applications

Traditionally, the AI we are familiar with, in terms of enterprise-grade applications and personal applications, exists in two entirely distinct realms. This is a phenomenon frequently encountered during the generative AI stage. As AI embarks on a new cycle, particularly as it enters the agentic AI development stage, the question of how to achieve deep integration between enterprise-grade AI and personal applications has become a crucial topic that every player must ponder.

Specifically, players can cater to people's needs by deploying distinct intelligent agents for enterprises and individuals. Taking Lenovo as an exemplar, in smart cities, hybrid AI orchestrates intelligent agents in domains such as transportation and energy through a "super hub" to enhance governance efficiency; in enterprise applications, Lenovo's "Agent as a Service (AaaS)" model enables clients to swiftly deploy customized intelligent agents in a lightweight manner, such as achieving predictive maintenance in manufacturing and assisting diagnosis in the medical field; personal users can manage tasks across devices through a "personal super intelligent agent" (such as Lenovo Xiaotian), striking a balance between privacy protection and personalized services.

Through this approach, both enterprise and personal applications are duly satisfied. Simultaneously, this approach can harness the strengths of different players while catering to the needs of diverse groups. Ultimately, AI applications can more diversely meet the needs of different groups and offer them a wider array of services. Under this trend, the barriers and divisions between enterprise and personal applications will be dismantled, propelling the industry into an entirely novel stage. As this phenomenon becomes more widespread, the application threshold for AI will further diminish, and a stage of deep integration between public large models and privatized models for individuals or enterprises will ensue.

Epilogue

As AI development gradually matures, particularly as it transitions from the generative stage to the agentic stage, relying solely on past models and strategies is encountering an increasing number of challenges and difficulties. It is foreseeable that hybrid AI will emerge as the new trend in the next stage. In this phase, continuously optimizing AI capabilities, transitioning AI from general-purpose to vertical applications, and achieving the seamless implementation of enterprise-grade and personal applications are the keys to truly propelling AI development into a new era.