In recent years, the rapid advancement of artificial intelligence (AI) technology has sparked a wave of innovation and competition in the global tech landscapeAmong the notable players is a company called Jumpshor, which has made significant strides in AI development, particularly in the realm of intelligent agentsFounded by former Microsoft executive Jiang Daxin, Jumpshor has rapidly established itself as a key contender in the race to achieve Artificial General Intelligence (AGI).

During Jumpshor's recent Step UP open day, Jiang shared insights into the company's roadmap toward AGIHe revealed that the company is currently focusing on the development of intelligent agents, which he believes will play a central role in the practical application of large AI modelsThis strategic pivot toward embedding intelligent agents into various sectors, including automotive, mobile technology, and the Internet of Things (IoT), indicates a broader ambition to transform everyday interactions with technology.

The excitement surrounding intelligent agents echoes the popular cultural desire for a personal assistant akin to Jarvis from the Marvel Cinematic UniverseJiang articulated his vision, suggesting that soon every device will have an intelligent agent facilitating a more seamless, multi-modal interaction between humans and machinesThis, he argues, will fundamentally change how individuals carry out complex tasks in their daily lives.

Jumpshor is not alone in its pursuit; the competitive landscape is rife with companies striving to harness the power of AI to enhance user experiencesThe breaking ground of each significant technological advancement creates ripples across industries, further validating the worth of AI unicorns like JumpshorThe company's decisive emphasis on the application of intelligent agents is both a test of its technological capabilities and a metric for its potential for commercial success.

Since its inception in April 2023 in Shanghai, Jumpshor has quickly attracted attention

Advertisements

The company boasts a team comprising more than 80% research and development personnel, highlighting its commitment to innovationFollowing a successful Series B funding round at the end of last year that raised hundreds of millions of dollars, its core investors—including state-owned Shanghai Capital Investment Company and tech giant Tencent—provided a solid foundation for its ambitious plans.

According to the AGI development roadmap outlined by OpenAI, the progress toward AGI is divided into five distinct stages: conversational AI, reasoners, agents, innovators, and organizersAt present, Jumpshor is positioned in the third stage—intelligent agents—signifying its pivotal role in the overall AGI developmental journey.

Jiang emphasizes that the hallmark of an intelligent agent lies in its autonomy to assist humans with complex tasksThis autonomy manifests in two primary ways: the agent should ideally carry out tasks independently without requiring human intervention, and it should proactively initiate tasks based on the user's context, shifting from a reactive to a proactive mode.

To realize this vision, Jumpshor has embarked on collaborations with various partners to develop a diverse ecosystem of intelligent agentsFor example, they have deepened technological ties with Geely Automotive and Qianli Technology to advance the integration of AI in automotive technologiesJiang identifies the automotive sector as a crucial arena for the deployment of intelligent agents, seeing potential innovations that can greatly enhance vehicle intelligence.

In the domain of smartphones, Jumpshor is actively working with manufacturers to integrate its agent technology deeply into mobile devicesThis effort includes innovative functionalities developed under their Step series of multi-modal models, such as OPPO’s “one-click question screen” and “one-click omni-search,” which aim to revolutionize user interaction with smartphones.

Moreover, Jumpshor has announced a strategic partnership with Zhiyuan Robotics to explore applications of AI-powered embodied robots

Advertisements

The company is also engaging with various IoT platforms and manufacturers, including TCL, to comprehensively enhance device connectivity and user experience.

This thriving ecosystem illustrates Jumpshor's ability to foster advancements in AI agents within a remarkably brief period, a feat attributed to its extensive technical groundwork and expertiseJiang identifies two critical factors driving the development of intelligent agents: multi-modal capabilities that allow agents to perceive and understand the world and reasoning abilities that enable agents to conduct long, reflective thought processes for precise task execution.

As of now, Jumpshor has launched 11 multi-modal large models across various functionalities, including speech recognition, generation, and image and video synthesis, indicating robust technical capabilitiesThe recent release of Step R-mini, the first reasoning model in the Step series, demonstrates further expansion of its AI model boundaries.

Additionally, Jumpshor is committed to fostering an open-source ecosystem, sharing technological innovations with developers worldwideRecently, in collaboration with Geely, the company released the Step-Video-T2V video generation model and the first product-grade open-source voice interaction model, Step-Audio, both of which are available for experience on the Yuedong app.

However, entering the agent phase is merely the beginningJiang’s projection outlines a path toward the development of intelligent agents that will go through five levels of complexityLevel 1 involves simple tasks handled by a single application, while Level 2 progresses to combined tasks that span multiple applicationsLevel 3 encompasses intricate tasks requiring substantial cognitive effort, and Level 4 highlights a shift where agents transition from being reactive to proactive, prompting or completing tasks autonomouslyThe pinnacle, Level 5, envisions an intelligent agent that possesses emotional intelligence, capable of providing emotional value and companionship.

Yet, achieving this ambitious Level 5 requires ongoing technological breakthroughs, particularly in enhancing the multi-modal and reasoning capacities of foundational models to maintain competitiveness among a plethora of global AI startups

Advertisements

Advertisements

Advertisements