Embracing Specialized Large Models: An Interpretation of Zhou Hongyi's Keynote Speech at the World Internet Conference in Wuzhen

Embracing Specialized Large Models: An Interpretation of Zhou Hongyi's Keynote Speech at the World Internet Conference in WuzhenThe 2024 World Internet Conference Wuzhen Summit was held on November 20th in Wuzhen, Zhejiang Province. 360 Group founder Zhou Hongyi delivered a keynote speech titled "Embracing Specialized Large Models, Leading the Development of the New Industrial Revolution" at the plenary session

Embracing Specialized Large Models: An Interpretation of Zhou Hongyi's Keynote Speech at the World Internet Conference in Wuzhen

The 2024 World Internet Conference Wuzhen Summit was held on November 20th in Wuzhen, Zhejiang Province. 360 Group founder Zhou Hongyi delivered a keynote speech titled "Embracing Specialized Large Models, Leading the Development of the New Industrial Revolution" at the plenary session. In his speech, Zhou delved into the profound impact of large model technology on the future of society and industrial development, offering unique insights into the development path of China's large model industry. He argued that the emergence of large models will propel human society into an intelligent era, acting as the driving engine of a new round of industrial revolution. However, he also pointed out that the Chinese large model industry faces numerous challenges during its development, requiring it to leverage its strengths and avoid weaknesses, choosing a development path suited to its own circumstances.

Zhou emphasized that large models will sweep through and reshape all industries in the future, becoming the core driving force of the next industrial revolution. He believes that, thanks to strong government support and collective industry efforts, China's large model industry has already joined the global ranks of leaders. However, he frankly acknowledged that limitations still exist in computing power chips and data. These limitations prompted him to consider the optimal development strategy for China's large model industry, leading to his suggestion of "leveraging strengths, avoiding weaknesses, and choosing a development path suited to our own circumstances."

So, what exactly is this "development path suited to our own circumstances"? Zhou believes that the future development of large models will follow two distinct paths: one is pursuing scale, aiming to create powerful general-purpose large models like "atomic bombs," exploring the limits of human cognition, and venturing into the vast expanse of space; the other is focusing on vertical fields, creating specialized large models like exquisite "tea eggs," achieving high efficiency within specific domains.

Zhou expressed skepticism about OpenAI's attempt to build an omniscient and omnipotent super general-purpose large model. He argued that, on the one hand, the data resources available for training large models are nearly exhausted, and the Scaling Law has encountered a significant bottleneck; on the other hand, developing a universally applicable general-purpose large model presents insurmountable logical difficulties. Therefore, he favors the development of specialized large models, which not only avoids fierce competition in computing power and data but also better aligns with the actual situation of China's large model industry. He believes that focusing on training data in specific domains can effectively improve model efficiency and accuracy, reducing dependence on massive amounts of data.

To more clearly illustrate the application prospects of "specialized large models," Zhou presented six application scenarios:

1. Technological Equality: Fully integrating personal needs, bridging the digital divide, and ensuring everyone can equally enjoy the convenience brought by artificial intelligence. This requires large model technology to be more widely applied and presented to ordinary users in a more understandable and operable way.

2. Universal Intelligence: Large models will become standard equipment for various intelligent devices, from large intelligent connected vehicles to small personal computers and mobile phones. All devices will benefit from the intelligent upgrades brought by large models. This will drive the booming development of the intelligent hardware industry and profoundly change people's lifestyles.

3. Empowering Traditional Industries with "Digital Transformation and Intelligent Upgrading": Using large model technology to promote the digital transformation and intelligent upgrading of traditional industries, improving production efficiency and product quality. This requires training a large number of specialized models to solve professional problems in various industries and achieve collaborative work among multiple models.

4. Empowering Emerging Industries: Large model technology will provide strong support for future emerging industries such as autonomous driving, embodied intelligence, and biopharmaceuticals, accelerating their innovation and development. This will create new industrial models and development opportunities.

5. AI for Science: Utilizing artificial intelligence technology to change the methods of basic scientific research and accelerate the progress of scientific discovery. This will drive a paradigm shift in scientific research and lead to new scientific breakthroughs.

6. Addressing New Challenges in Large Model Security: The development of large model technology also presents new security challenges, requiring strengthened security protection measures to prevent and address potential risks.

Zhou also introduced 360's practices and achievements in large model applications. 360's fully developed AI search and AI office products are helping hundreds of millions of users improve their work efficiency; in line with the trend of universal intelligence, 360 has released its first AI children's watch equipped with a large model; and 360 has achieved significant results through more than 100 PoC projects, widely deploying enterprise-level large model solutions in government offices, public security, and industrial production scenarios. Particularly noteworthy is 360's proposed "model-building-with-models" concept, offering a feasible solution to tackling new challenges in AI security. This demonstrates that 360 not only focuses on the application of large model technology but also attaches great importance to large model security.

Zhou Hongyi's speech not only provided an in-depth analysis of the development trends of large model technology but also pointed out the development direction for China's large model industry. He emphasized that China's large model industry should fully leverage its advantages, avoid direct competition with international giants in computing power and data, and focus on developing specialized large models to secure a place in the global large model industry competition and contribute more to China's economic and social development. His "tea egg" analogy vividly illustrates the development strategy of specialized large models, both practical and in line with China's national conditions. This not only provides valuable experience for the development of China's large model industry but also offers a new perspective on the development of the global large model industry. Zhou Hongyi's speech undoubtedly injected new vitality into the rapidly developing large model industry and provided important directional guidance for its future development. His views and suggestions deserve careful consideration and reference from industry professionals. The future development of China's large model industry will depend on its ability to seize opportunities, choose the right path, ultimately achieve sustainable development, and contribute Chinese wisdom to the global artificial intelligence industry.

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:[email protected])