Alibaba's Tongyi Qianwen 2.0 Debuts! The 'Third Wave' of Cloud Computing Has Arrived
-
On October 31, the 2023 Yunqi Conference kicked off as scheduled in Yunqi Town, Hangzhou.
According to reports from the opening day, this year's Yunqi Conference features two major 'returns.' First, Alibaba Cloud founder Wang Jian has returned, and second, the conference theme, 'Computing, for the Value Beyond Calculation,' evokes memories of the theme from eight years ago, 'For the Value Beyond Calculation.'
In 2015, cloud computing experienced explosive growth, revolutionizing fields such as social media, e-commerce, finance, and media. In 2023, as AI adoption becomes an unstoppable trend, another leap is underway.
From the perspective of Alibaba Cloud's thematic return, it can be seen as a reaffirmation of its positioning, re-emphasizing cloud computing as the 'infrastructure' for technological development and its future potential value.
As Alibaba Cloud's flagship entry into the large-scale AI model market, Tongyi Qianwen has been upgraded to version 2.0 and was made publicly available for the first time on the opening day. Reports indicate that its overall performance has surpassed GPT-3.5 and is closing the gap with GPT-4.
Wang Jian: The Third Wave of Cloud Computing Has Arrived
Founded in 2009, Alibaba Cloud started from scratch, independently developing China's first true cloud computing operating system. In the same year, the Yunqi Conference was born, initially a grassroots gathering of just over 400 webmasters.
Over the past 14 years, cloud computing has served as a foundational force closely intertwined with three global waves of technological innovation, with the Yunqi Conference reflecting the evolution of China's digital economy.
At this year's Yunqi Conference, Wang Jian, affectionately known as 'The Doctor' and a returning figure, described the current integration of cloud computing and AI as 'the third wave of cloud computing.'
Wang Jian, the founder of Alibaba Cloud and a member of the Chinese Academy of Engineering, is regarded as the pioneer of cloud computing in China.
During the main forum on the opening morning, Wang Jian reviewed the history of cloud computing and innovation waves, dividing it into three distinct phases:
The first wave saw cloud computing transform the global internet, giving rise to 'cloud-native' enterprises. Companies like Netflix and miHoYo represent digital-native businesses that have been 100% cloud-based since their inception.
The second wave witnessed traditional enterprises embracing cloud computing. Pioneers in IT adoption like government and financial sectors began their cloud transformation. The landmark event was the 2022 Beijing Winter Olympics, where cloud computing first supported core competition systems.
As Wang Jian noted, "The relationship between cloud computing and the Olympics isn't simply about using technology or not. Its significance is comparable to the first use of alternating current at the Chicago World's Fair."
The third wave emerged in 2023 with the concentrated explosion of AI models and cloud computing. The advent of GPT models has created profound impacts on technological innovation.
"Looking back at the electrical revolution over a hundred years ago, the relationship between cloud computing and GPT is akin to that of electricity and electric motors. In the future, the computing power of cloud computing will be consumed by these 'electric motors of the intelligent era'—the AI models," Wang Jian stated at the conference.
Years ago, when discussing the rise of AI, Wang Jian mentioned that while artificial intelligence is fascinating, patience is required as it will take a long time to ensure the correctness of the entire business and operations.
At the recent Yunqi Conference, Wang Jian reiterated this perspective, noting, "The integration of cloud computing and artificial intelligence won't be completed in a year or two. It may take a decade or even several decades for enough innovations to emerge in this wave."
For Alibaba Cloud, which has now spun off, the proposed AI + cloud computing integration strategy could provide new fuel for this 'unicorn' to accelerate its growth.
Financial reports show that in the 2023 fiscal year, Alibaba Cloud's revenue reached 77.203 billion yuan, nearly six times the 13.39 billion yuan in the 2018 fiscal year. It has achieved profitability for two consecutive years and has become the second-largest revenue contributor to Alibaba Group, second only to the Taotian Group.
In terms of competitive landscape, Alibaba Cloud remains the leader in the domestic market. According to IDC reports, Alibaba Cloud consistently ranks first in the public cloud market, whether in IaaS or IaaS+PaaS segments.
The rise of large AI models has provided a new approach for global cloud computing companies and a fresh growth opportunity for Alibaba Cloud, potentially allowing it to reap benefits in the third wave of innovation.
At the main technology forum of the Yunqi Conference, Alibaba Cloud CTO Zhou Jingren announced the latest progress of Alibaba's self-developed large model, releasing the Tongyi Qianwen 2.0 with hundreds of billions of parameters.
On October 31, the Tongyi Qianwen APP was officially opened to the public for the first time, available in major mobile app markets, allowing all users to experience the latest model capabilities of Tongyi.
"The title of my speech today, 'Expanding the Boundaries of Intelligence, Ushering in a New Era of Cloud Computing,' was chosen by Tongyi Qianwen," Zhou Jingren said at the beginning.
In April this year, Alibaba Cloud officially launched the large language model Tongyi Qianwen, entering the battle of large models. Zhou Jingren stated that after six months, the newly upgraded Tongyi Qianwen 2.0 version has achieved a significant leap in performance.
According to reports, in 10 authoritative evaluations, Tongyi Qianwen 2.0's overall performance has surpassed GPT-3.5 and Llama2, further narrowing the gap with GPT-4. Compared to version 1.0, version 2.0 shows significant improvements in complex instruction understanding, literary creation, general mathematics, knowledge retention, and hallucination resistance.
As of October, Alibaba Cloud has established deep collaborations with over 60 leading industry partners to promote the application of Tongyi Qianwen in various sectors including office work, cultural tourism, power, government affairs, healthcare, transportation, manufacturing, finance, and software development.
Zhou Jingren revealed that Alibaba Cloud plans to open-source the Tongyi Qianwen 72B version soon, enabling developers across industries to innovate based on the open-source model. Previously, Alibaba Cloud has already open-sourced the 7B and 14B versions, with cumulative downloads exceeding 1 million.
Not just Tongyi Qianwen, the entire Tongyi model family has undergone comprehensive upgrades.
To date, Alibaba Cloud's Tongyi large model system is primarily divided into two categories: base models and application models.
Base models are represented by the large language model Tongyi Qianwen and the painting creation model Tongyi Wanxiang. Additionally, to facilitate the integration of large models across industries, Alibaba Cloud has developed 8 industry-specific application models based on Tongyi.
At the conference, three new members of the Tongyi AI family made their debut: the personalized character creation platform Tongyi Xingchen, the intelligent investment research assistant Tongyi Dianjin, and the AI reading assistant Tongyi Zhiwen. The intelligent coding assistant Tongyi Lingma has been widely adopted within Alibaba Cloud, while the work and study AI assistant Tongyi Tingwu has already amassed over 1 million users.
In short, the eight "siblings" of the Tongyi family now cover nearly every aspect of work, study, and daily life. Some can assist with meetings, copywriting, and coding, while others help users quickly digest text, audio, and video content. There are even AI companions that can create personalized characters for chatting or virtual relationships based on user preferences...
Large models are driving a new wave of innovation across industries, aiming to revolutionize production and service methods. However, both training and inference for these models demand significantly higher computing power.
"Currently, half of China's large model companies operate on Alibaba Cloud, with 2.8 million AI developers active on the ModelScope community and over 100 million model downloads. Fostering China's AI ecosystem is Alibaba Cloud's top priority," stated Zhou Jingren at the event.
As cloud computing and large model technologies continue to converge, this integration may present new opportunities for ordinary entrepreneurs, especially those without technical backgrounds.
This year has seen emerging innovations like the viral 9.9 yuan AI portrait app "Miao Ya Camera," the children's story generator "Tongyu Story" that creates picture books in under a minute, and the content creation tool "Caiyun Xiaomeng" for video creators—all examples of this third wave of AI-driven innovation.
As early as the 1960s, American computer scientist John McCarthy first proposed the term "artificial intelligence" and envisioned computing power being provided as a public service, much like water and electricity.
In 2023, trends revealed at the Cloud Computing Conference and the practical applications of large models across various industries demonstrate that cloud computing, as the foundation of AI's explosive growth, has become the utility McCarthy imagined. It is further penetrating all sectors, bringing unexpected new applications.
In this third wave of technological innovation, the integration of these two fields relates to Alibaba Cloud's ambitions and the future vision of industrial transformation, making it a development worth anticipating.