A new era of 'big model + small model' opens, AI basic software defines the future of AI

Description: An AI company’s response to the new era of AI

On May 30, the "Hangzhou General Artificial Intelligence Forum" hosted by China Academy of Information and Communications Technology, Zhejiang Provincial Department of Economy and Information Technology, Hangzhou Municipal People's Government, and China Artificial Intelligence Industry Development Alliance was grandly held. Sun Xudong, member of the Party Leadership Group of Hangzhou Municipal People's Government and deputy mayor, Li Yongwei, chief engineer of the Zhejiang Provincial Department of Economics and Information Technology, Wang Xiaoli, deputy secretary of the Party Committee of the China Academy of Information and Communications Technology, Pan Yunhe, academician of the Chinese Academy of Engineering and chairman of the China Artificial Intelligence Industry Development Alliance, and hundreds of other government executives Leaders, representatives of outstanding AI companies, experts and scholars from top universities, and mainstream media gathered at the conference to discuss the opportunities and challenges of large model application. Fang Lei, chairman of Jiuzhang Yunji DataCanvas Company, was invited to attend the conference and delivered a keynote speech on "AI Basic Software Catalyzes the Rapid Development of Large Model Applications".

Fang Lei, Chairman of Jiuzhang Yunji DataCanvas Company

AI basic software is the key, and a new era of “big and small” has begun

The applications of large-scale models released recently are emerging one after another, triggering new thinking about changes in the industry, and may even reform the field of artificial intelligence. Fang Lei proposed that in the development process of the AI ​​industry with Chinese characteristics, central enterprise cloud, computing power construction and large models constitute the "troika" of industry development. Under the influence of the three parties' mutual causation, mutual promotion and mutual superposition, my country's AI The industry will usher in tremendous development.

Fang Lei pointed out that new infrastructure will bring golden opportunities. The construction of central enterprise clouds, especially operator clouds, is growing rapidly, and its size is already comparable to that of leading cloud vendors. At the same time, the emerging central enterprise clouds will also undertake part of the new infrastructure construction of computing power. With the effect of this growth rate, the cooperation model of software and hardware manufacturers in the ecosystem will change, and the landscape of cloud computing will also change.

In terms of computing power, not only the massive hardware investment in computing power itself will bring abundant computing power, but there is also huge room for end-to-end computing power optimization, and the price of computing power will no longer be a gap. When technology crosses the singularity, the importance of AI basic software becomes even more prominent. The cost efficiency of large-scale model applications will be determined by the performance of the AI ​​infrastructure software that is the core of the model ecosystem.

Although large-scale models currently perform well, there are still high technical and cost thresholds for actually applying them to business scenarios in different industries. Fang Lei pointed out that the current era of "size" is ushering in, not only the integration and use of large models and small models, but also the miniaturization of large models, or miniaturization and fine-tuning based on large models, is also a trend. This kind of The method can solve a large number of problems at a low cost.

"Big and small are a relative change." Currently, the parameter standards of large models are not uniform. Compared with the parameter level, the effect of the model and whether it can support rapid iteration is more important for users' practical applications. Users can fine-tune and iterate customized small models quickly and cost-effectively based on a large white-box model, so that they can efficiently implement large model applications in rich scenarios. This once again highlights the importance of the basic AI software tool chain.

Multimodality defines a new AI future, from “software” to “thought-ware”

The popularity of multi-modal large model technology is not only prompting the restructuring of the industry. Fang Lei believes that "multi-modality" will open up a new future for AI. From then on, the relationship between models and data will become more subtle and revolutionary. sex, software may be redefined.

Multimodality includes not only pictures and text, but also the fusion of sensor data and structured data. Multimodality allows us to operate the world by linking AI software with sensors and the physical world. This is the first time that AI has allowed us to see the prototype of the human brain. This is as historically significant as the large model.

Secondly, multi-modality will achieve the unification of data, which is a sign of the advent of the vectorization era. Just like no enterprise has linked its knowledge base and data warehouse in the past, it is naturally considered an impossible situation. However, the vectorization of current documents has been realized by large models. All data can be unified on a multi-modal basis in the form of vector data, which means that data is truly "unified" for the first time. This revolution, brought about by large-scale inverse modeling, will exponentially improve our ability to leverage data.

Fang Lei predicts that future technology will bring about changes and redefine the software industry. Traditional software is a tool for process automation and forward operation, but under the influence of multi-modal large models, future software will be reversely flexibly assembled into various forms to serve users according to people's ideas. This is exactly the same. This is the transformation direction of AI software proposed by Jiuzhang Yunji DataCanvas from "software" to "thought-ware" that helps all walks of life.

As an AI basic software supplier, Jiuzhang Yunji DataCanvas Company has always adhered to the positioning of "allowing customers to have independent AI capabilities". In the AI ​​era influenced by multi-modal large models, this positioning remains firm. In the future, the development of the AI ​​industry will accelerate towards a diversified path where a hundred flowers bloom. Jiuzhang Yunji DataCanvas will rely on a series of independently innovative open, automatic, cloud-native A-based software tools to explore the future of the vector world side by side with the entire industry.

The above is the detailed content of A new era of 'big model + small model' opens, AI basic software defines the future of AI. For more information, please follow other related articles on the PHP Chinese website!