Talk about the grand vision model

Big visual models are getting smarter

2711P-RDT15CB Wang Yaonan: A large language model is a neural network model that uses the knowledge in books, language and text as data for training, and can reason and answer according to the knowledge it has learned. It is a data-driven artificial intelligence.

The data of the large vision model is mainly from a variety of images, including visual data generated by humans and nature. For example, the medical large vision model is to input human organs and lesion images as visual data into the large model for training, which can understand CT photos like a doctor, and can achieve the effect of reasoning the patient’s condition after taking photos.

The current large model of vision is not like the human brain as we think, and the gap is still very big. With the increase of learning data and the adjustment of model parameters, the model will be larger and larger, and the knowledge will be more and more, and its 2711P-RDT15CB intelligence level will be higher and higher and more intelligent.

We need to improve the level of computing power and speed up the computing speed to build the model faster. Design better model architectures, including more interpretability, security and control; More efficient learning algorithms also need to be researched.

In fact, the visual grand model is not just a few years ago, it is a step by step development. In the 1980s, with the development of artificial intelligence, people began to study neural networks, it was already there. It’s just that, in recent years, computing power and algorithmic capabilities have improved, and people can try to build large models, resulting in large language models and large vision models. In the past, when we didn’t have enough computing power, the models we made weren’t that big.

Talk about 2024 outlook

2711P-RDT15CB Hope the business from Guangzhou growth to the country and the world

Wang Yaonan: Hunan University’s National Engineering Research Center for Robot Vision Perception and Control Technology has settled in Guangzhou Zengcheng, located in the Guangdong-Hong Kong-Macao Greater Bay Area, in 2022, and established the Huda 2711P-RDT15CB Guangdong-Hong Kong-Macao Greater Bay Area Innovation Research Institute (Guangzhou Zengcheng).

2711P-RDT15CB The research institute is mainly engaged in the research and application of machine intelligent vision, including special operation robots in intelligent manufacturing, medical and pharmaceutical fields, as well as general large vision models. For example, these studies applied to the manufacturing industry can replace a lot of manual work to complete product quality testing, especially in the 3C and high-end parts industries. At present, the Institute has developed hardware and software systems, the main function is to enable Guangdong enterprises to digitize, intelligent transformation, promote the development of manufacturing industry.

In addition, the Institute also develops intelligent machine vision and control systems, mainly for the production of high-end intelligent equipment, such as industrial Internet software, which has a lot of algorithms.

Guangdong is the forefront of reform and opening up and the main economic battlefield, the industrial chain, supply chain is complete, there are many manufacturing enterprises, we came to Guangdong layout is out of market demand. In addition, many research and development teams of our research center have landed many projects in Guangdong to transform scientific and technological achievements.

In the first half of 2023, we did encounter some challenges, mainly on the supply 2711P-RDT15CB side of components. In 2024, I believe these problems will be solved, and the industrial chain, supply chain and research and development capacity of Guangdong’s AI industry will be improved. In this big domestic market, new tracks will open up. In the dynamic atmosphere, everyone takes the initiative, takes responsibility and innovates, and challenges can be solved.

I have full confidence in the development of Guangdong’s AI industry. Over the past 30 years, we have made great efforts in scientific and technological self-reliance and accumulated many scientific achievements. Guangdong is the largest province in our country.

The field of artificial intelligence and robotics that we are engaged in has caught up with a good opportunity, and the explosive growth in the past two years has brought many new markets and driven new industrial tracks. In the future, there will be more and more intelligent terminals, which will also drive the development of the production-oriented facilities and equipment industry. In 2024, I hope our research institute in Zengcheng, Guangzhou will be able to expand its business to the whole country and the world.