When artificial intelligence is undergoing new changes every week, the limitations of traditional deep learning algorithms in security scenarios are becoming more and more prominent, and how to make AI safer and more popular and landing in vertical scenarios has become a concern issue.
On May 31, the 360 Smart Life Visual Big Model and AI Hardware New Product Conference was held, with the theme of "Safety Protection, Smart Vision", launched the 360 Smart Brain Visual Large Model and a variety of new AI hardware products, and announced that 360 Smart Life officially entered the SMB market, realizing a great leap from home scenes to enterprise commercial use.
The combination of multimodal large models and the Internet of Things will become the next outlet
Zhou Hongyi, founder of 360 Group, said that the big language model will bring a new industrial revolution, and predicted that the combination of multimodal big model and the Internet of Things will become the next outlet, and the big model will lead intelligent hardware into the AIoT era.
Image source: 360 Smart Life Vision Big Model and AI Hardware New Product Launch
Sun Hao, general manager of 360 Visual Cloud Business Line, said in an interview that after integrating visual ability and text alignment into a language model, it can use the ability of large language model to understand, reason, and make decisions. Therefore, the big language model actually plays a brain role in the multimodal model, and only if the big language model is strong, the ability of the visual big model will be strong.
Image source: 360 Smart Life Vision Big Model and AI Hardware New Product Launch
According to Sun Hao, the 360 intelligent brain visual model has three core capabilities at this stage:
1. Open Object Detection (OVD), by learning the graphic and text data of the Internet volume, so that the model can learn more target categories in a comprehensive manner;
2. Image title generation is an advanced form of understanding pictures, which can meet the needs of users to obtain information quickly and accurately;
3. Visual question answering (VQA), based on the recognition, understanding, and reasoning of pictures, answers various questions asked using natural language.
Image source: 360 Smart Life Vision Big Model and AI Hardware New Product Launch
When talking about the relationship between the visual big model and the 360 intelligent brain, Sun Hao said: "The big language model is a very important part in the field of multimodality. For our visual big model, we are based on the Internet's graphic data training, but one of the core of its intelligence is our large language model that integrates 360 intelligent brains. Therefore, the relationship between us and the 360 intelligent brain large language model is logically parallel, but in terms of ability, it is actually a help and supplement, and in the future, we are all an important capability component of the 360 intelligent brain multimodal large model. ”
"To a certain extent, large models are originally based on massive data to train and learn. Why is the multimodal large model combined with the Internet of Things may be the next outlet, because it can further integrate and learn the data perceived by the Internet of Things, so that it can produce further analysis and decision-making on our production and operation, life and produce greater value. Sun Hao said in the interview.
Officially entered the SMB market
In specific landing applications, 360 smart life has been explored for many years. Last year, 360 launched an intelligent security SaaS platform with scenario-based AI as the core, that is, 360 Visual Cloud. At present, 360 core intelligent security products have been connected to 360 visual cloud, which can be seamlessly connected in the same scene to form a data visualization solution.
Sun Hao introduced that the combination of 360 intelligent brain visual large model and hardware should be a gradual process, such as 360 first integrates the visual capabilities of the camera, through the combination of the device-side algorithm and the cloud, so that it may have the visual landing ability of end-cloud collaboration. If the large model is further optimized and can run on the edge, it will gradually be placed on the edge. Next, the information of the cloud is extracted and analyzed, and the perception of the mode of voice can be gradually integrated, and it can be imagined that its future is a process of gradual penetration and gradual reshaping.
Image source: 360 Smart Life Vision Big Model and AI Hardware New Product Launch
At this conference, 360 also officially announced a series of new products: 360 outdoor ball machine 6 Pro, 360 dazzling outdoor gun ball all-in-one machine, 360 binocular splicing panoramic camera, 360 dazzling intelligent NVR, etc., to achieve full coverage of scene channels from indoor to outdoor, from household to commercial, from online to offline. With the empowerment of the 360 intelligent brain visual big model, a richer product matrix can better meet the multi-scenario needs of To C and small and medium-sized enterprises, and can achieve wider compatibility. At the same time, 360 also officially announced 360 Smart Life's entry into the SMB market.
Sun Hao said in the interview that the first scenario chosen by 360 security to enter the B side is the SMB market, because small and medium-sized enterprises have many special needs. "At present, the scene boundaries of many small and medium-sized enterprises are not so clear, and many users choose 360 products in their own business scenarios, because they value the ability of 360 products to be simple and easy to use in home scenarios. Our products are easy to install, simple to use, decentralized deployment, centralized management, anytime, anywhere to take out the equipment to see the situation, these can make the digital landing threshold of small and medium-sized enterprises extremely low, which is also the core reason why we chose to enter the SMB market. He said.
In the future, 360 Smart Life will continue to cultivate the frontier field of intelligent security, explore the application of AI large models in security scenarios, and empower the digital upgrading of small and medium-sized enterprises with inclusive AI.