"Technology is advancing rapidly, and human lifestyles are rapidly changing, bringing a series of incredible singularities to human history. Everything we once knew has begun to become unfamiliar," said John von Neumann, the father of computers.

In the midst of the full-scale explosion of AI, every day is filled with uncertainty, and the emergence of every new technology drives the entire market into a frenzy. Nevertheless, the ultimate destination of the road ahead remains unknown, making it a topic of great concern and ongoing debate among everyone.

Image source: Pixabay

Until recently, Barclays released a research report providing an answer, a clear "AI roadmap" outlining the future evolution path of AI technology. It encompasses three stages: the prevalent era of chatbots known as the "AI Assistant Era," the gradually emerging "AI Agent Era" from 2025 to 2026, and the impending "Digital Employees and Robots Era" in 2027.

So, what epic history will the market continue to write in this rhythm?

The enduring 'Assistant Era,' where some have already begun 'reaping their harvest'

Looking back, with the astonishing success of ChatGPT and Sora in 2023, the incubation boom of chatbots quickly spread across the global market. Foreign countries accelerated technological iteration, while China continued to stage a "hundred-model battle," creating a lively atmosphere.

As we all know, chatbots have already achieved widespread adoption both domestically and internationally. Especially in the domestic market, large models based on chatbots have undergone countless iterations, penetrating into various industries.

The 'AI Assistant' has evolved from the earliest chatbots into numerous specialized types, and the increasingly refined AI large model track has gradually formed a 'dumbbell-shaped' bipolar characteristic.

At one end are technology giants and basic model vendors represented by OpenAI, who focus on underlying research and development and are highly dependent on funds and resources. Once they master groundbreaking technology, they can create a 'stronger becomes stronger' advantage.

At the other end are represented by Baidu's ERNIE Bot, Aliyun's Tongyi Large Model, and Tencent's Hunyuan Large Model, which approach from the application path and focus on the integration of large models with scenario and industry knowledge. Although reliant on the capabilities provided by basic models, their ultimate application in specific industries or scenarios can provide users with solutions or services to real-world problems.

Among them, Baidu, which entered the market earliest, has already achieved market coverage and begun 'reaping its harvest.' IDC China data shows that in 2023, the Chinese large model platform market reached 1.765 billion yuan, with Baidu Intelligent Cloud capturing 19.9% of the market share to rank first in the Chinese large model platform market, while ERNIE Bot boasts China's largest industrial deployment scale.

In 2024, while large models underwent a major shakeup, industrial deployment also accelerated significantly. According to statistics, in the first half of 2024, the total winning bid amount for large models was approximately 938 million yuan, far exceeding the entire year of 2023. In terms of winning bids, Baidu had the highest number of winning projects (17), the largest total winning bid amount (92.97 million yuan), and the most diverse industries covered (up to six).

Taking the Qianfan platform as an example, the platform has already launched solutions for five major industries and over 20 scenarios, including education, human resources, enterprise services, entertainment, and healthcare.

Overall, during the past two years of the 'AI Assistant Era,' the market not only experienced a major industrial explosion but also underwent a significant shakeup. Domestic AI large models have a clearer direction forward, with the 'assistant' attribute penetrating into more corners of the market. Pioneers like Ali and Baidu, who entered the market earliest, have already achieved substantial results.

As time passes, the 'AI Assistant Era' is transitioning to the 'AI Agent Era,' and the strategies and pace of major platforms and enterprises have not lagged behind.

Who is 'sharpening their swords and feeding their horses' for the highly anticipated 'Agent Era'?

The emergence of large language models and AI's enhanced speech interaction capabilities and comprehension have provided a powerful boost to the development of 'AI Agents.'

Facing the upcoming 'AI Agent Era,' AI agents capable of autonomously completing tasks will usher in a new stage of widespread adoption. Unlike chatbots and assistants, AI agents can not only accomplish complex tasks through multiple request transfers but also reduce direct human intervention.

Currently in the preheating stage, many leading enterprises have taken the lead in action. On October 21, Microsoft announced the launch of 10 'Artificial Intelligence (AI) Agents' aimed at helping staff in sales, customer support, finance, and other fields better accomplish tasks while promoting the widespread application of AI in the business sector.

Simultaneously, Anthropic, an American AI startup, also announced the release of its latest generation of 'Claude' models, including an 'AI Agent' capable of using computers. In September of this year, Salesforce, the largest provider of customer relationship management software services in the United States, also launched its own 'AI Agent' – Agentforce.

At present, the AI agents launched by foreign giants are mostly computer applications. However, there are many other application scenarios for AI agents. For instance, they can be used in information communication, smart homes, healthcare, and financial services.

Image source: Pixabay

Among them, autonomous vehicles are nothing more than a tangible manifestation of 'AI Agents.' An autonomous driving system can process sensor data independently, navigate, and make driving decisions without relying on human input, ensuring travel safety.

Taking Tesla as an example, its pure vision-based autonomous driving solution embodies this approach, emphasizing perception over mapping, and entrusting route decisions entirely to the FSD system. Although Tesla's recent Robotaxi Day demonstration disappointed the market, it nevertheless reflects the market's high expectations for the application of autonomous driving.

Apart from Tesla, there is another surprising domestic enterprise, Baidu, which seems to have hit the 'sweet spot' every time, from 'assistant' to 'agent.' With ERNIE Bot, it seized the initiative in the AI Assistant Era, and with its autonomous vehicle Luobo Kuaipao, it placed a bet on the 'AI Agent Era.'

This was further confirmed by Baidu's first-quarter performance this year. Multiple businesses experienced significant AI-driven growth, with Baidu Intelligent Cloud revenue increasing by 12% year-on-year, and AI revenue accounting for 6.9%. AI reshaped Baidu Search, with 11% of search results generated by AI. Additionally, Luobo Kuaipao doubled its operational area in Wuhan, with fully autonomous driving accounting for over 55% of rides in the first quarter, and successfully achieved round-the-clock operations in March.

In today's rapidly changing technology and market landscape, being able to enter the market early and seize opportunities represents not only extraordinary courage but also leading strength. This spirit is precisely what is needed in the upcoming 'AI Agent Era' and 'Robot Era.'

Approaching the 'Robot Era,' opening the door to a new world

With the further upgrading of AI agents, AI may usher in a concrete and fully intelligent presentation, thereby entering the 'Digital Employees and Robots Era.'

Specifically, in enterprise applications, AI agents may evolve into 'digital employees' capable of independently completing tasks. In the consumer market, intelligent robots will gradually integrate into household life, undertaking simple and repetitive daily tasks.

While the concept of 'digital employees' is already well-known. Early on, virtual anchors such as CCTV's Little C, Xinhua News Agency's Xiao Zheng, People's Daily's Guo Guo, and Hunan TV's Xiao Yang set the tone for the application of digital employees. Now, represented by AIYA from Baixing Bank, Harbin Beer's Ha Jiang, and OPPO's Xiao Bu, 'digital employees' are gradually maturing in terms of technology and application.

However, there is a significant difference between these and the currently mentioned 'digital employees.' The key lies in whether they are 'autonomous' and 'independent,' with the latter clearly representing the 2.0 version.

With the continuous upgrading of large models, the practical capabilities of digital humans (including digital employees) are expected to evolve from assistants to agents and even 'managers.' Therefore, as manufacturers accelerate their exploration of large model deployment paths, digital humans are viewed as pioneering scenarios by many.

In reality, this is also the common direction of industry efforts at present. The practicality and cost-effectiveness of AI digital humans have become the focus of market attention, and reducing product usage costs through relevant technologies has become one of the critical factors in future competition.

At this year's first China Digital Human Conference, Zhipu presented its 000th digital employee 'Hua Tongxue' and virtual host 'AI Ma Fanshu,' making a stunning debut. Especially in September of this year, Baidu's Xiling digital human underwent a 4.0 upgrade, reducing production thresholds, optimizing efficiency and costs, and significantly enhancing overall capabilities.

The performance of digital humans is moving towards surpassing that of real humans, a manifestation of comprehensive intelligence. Looking further, Tesla and Xiaomi have successively launched humanoid robots this year, undoubtedly another tangible expression of AI intelligence.

Xiaomi Cyberone Humanoid Robot (Image source: Xiaomi official website)

Currently, both Tesla's Optimus and Xiaomi's Cyberone prioritize industrial manufacturing as their initial deployment scenarios. Compared to service or household scenarios, industrial scenarios are safer, more controllable, offer more substantial ROI, and have a clearer and more focused business model.

The first step towards concretization is already clear, leaving only the pioneering journey ahead for the era's forerunners.

In the future, if we can achieve the dual visions of comprehensive intelligence and concretization and integrate them, the bustling streets of the 22nd century may surpass our wildest imaginations.

Source: Hong Kong Stocks Research Society