SenseTime, a strategic partner of the 2023 World Artificial Intelligence Conference (WAIC), held the "Boundless Love · Daily Innovation" artificial intelligence forum, and launched a multi-faceted and comprehensive upgrade of the "SenseTime SenseNova" large-scale model system. A series of large-scale model product updates and landing results under this system. In addition, SenseTime also focused on introducing and demonstrating the application practice of its large-scale model technology with all parties in the industry since its official release, including the latest smart cockpit products and vehicle-road-cloud collaborative transportation system created by SenseTime, as well as financial, Landed applications in the production practice of industries such as medical care, e-commerce, mobile terminals, and industrial parks.
Xu Li, chairman and CEO of SenseTime, said in the product release session: "The breakthrough of large models has set off a new round of technological revolution in artificial intelligence, followed by explosive growth in industrial demand and new application scenarios. And application models are emerging rapidly. SenseTime hopes to continue to promote the leapfrog improvement of AI infrastructure capabilities through "big model + large device", not only to create a basic model with more powerful general capabilities, but also to further efficiently integrate professional knowledge in different vertical fields to build A professional large model that understands the industry better and has more expertise fundamentally reduces the downstream application cost and threshold of the large model, allowing the industrial value of the large model to bloom in thousands of industries."
It means that "the speed of model iteration and the ability to deal with problems can be updated every day", and SenseTime's large-scale model system is undergoing high-speed iteration under its AGI strategic layout of "large model + large device". As a natural language processing model with hundreds of billions of parameters, SenseChat version 2.0 breaks through the limitation of the input length of large language models, and launches model versions with different parameter levels, which can perfectly adapt to different terminals and scenarios such as mobile terminals and cloud application requirements and reduce deployment costs. The model parameters of SenseTime's self-developed large-scale model SenseMirage 3.0 have increased from 1 billion to 7 billion since it was first released in April this year, and can achieve professional photography-level picture details.
Not only that, SenseAvatar 2.0 digital human generation platform, compared with the 1.0 version, has improved the voice and mouth fluency by more than 30%, realized 4K high-definition video effects, and brought AIGC to generate images and digital human singing functions. In addition, the space reconstruction efficiency of SenseSpace 2.0 has been increased by 20%, and the rendering performance has been increased by 50%. The construction time of every 100 square kilometers scene can be completed in only 38 hours (supported by 1200 TFLOPS/second computing power); and SenseTime Gewu SenseThings 2.0 restores the texture and material of small objects to millimeter-level fineness, and breaks through the collection of highly reflective and specular objects.
New day after day, multi-modal empowerment industry upgrade
Relying on the rapid iteration of the "SenseNova" large-scale model system in the underlying technology field, SenseTime is actively empowering industrial upgrading through the multi-modal capability combination of large models, and has brought many new breakthroughs leading the industry.
In the financial field , SenseTime cooperates with customers such as banks, insurance companies, and securities companies, and uses digital humans to perform tasks such as intelligent customer service and intelligent marketing, and provides new functions such as investment research analysis and research report writing through access to large language model capabilities. Realize cost reduction and efficiency increase. In addition, after mounting the financial knowledge base, it can also output content questions and answers 100% based on the customer's product description, and realize timely information update.
In medical scenarios , SenseTime has created a large Chinese medical language model "big doctor" based on massive medical knowledge and clinical data, providing multi-scenario multi-round conversation capabilities such as guidance, consultation, health consultation, and auxiliary decision-making. In the future, it will soon support medical science. Multi-modal comprehensive analysis of images, texts, and structured data can continuously improve medical language understanding and reasoning capabilities, and continue to empower hospitals to improve diagnosis and treatment efficiency and patient services.
Combining the comprehensive capabilities of Discussion 2.0 and Miahua 3.0, SenseTime also brings a variety of intelligent interactive solutions to mobile terminal customers, including question-and-answer interaction for information acquisition, knowledge interaction for life scenes, and content interaction for language and image generation etc., relying on the lightweight version of SenseTime's large model, it can be easily deployed and run on mobile terminals. In addition, in the immersive sci-fi experience space of "The Three-Body Problem Beyond Gravity" created by SenseTime based on Liu Cixin's award-winning novel "The Three-Body Problem" , SenseTime uses the ability of large models to break through the boundaries of imagination, create and display extreme A futuristic sci-fi voyage.
Facing offline scenarios , SenseTime brings smart solutions such as long-tail fault identification and complex defect judgment to power grid inspection through large-scale model capabilities. Based on the spatial reconstruction of Qiongyu 2.0, SenseTime has created digital twins of real-scene spaces for the regional development of Mashan Town in Jinan, the China Vision Park in Hefei, and Ruijin Hospital in Shanghai to improve the efficiency of operation and management. In the jewelry industry , relying on Gewu 2.0 SenseTime to carry out jewelry re-engraving for jewelry brands, show the characteristics of product craftsmanship in detail, and improve customer shopping experience.
On the online short video and live broadcast platform , digital humans generated by SenseTime Ruying 2.0 are being widely used. SenseTime has also reached channel strategic cooperation with a number of leading companies to jointly build the "cloud + AIGC + short video live broadcast" ecology, which will serve the industry Bring more efficient, low-cost, convenient and easy-to-use AI video and marketing tools.
In the field of smart cars , industrial applications such as SenseTime's smart cockpit, smart driving, and vehicle-road collaboration have also broken through the boundaries of innovation with the blessing of large models. In the smart cockpit, SenseTime perceives user needs in an all-round way through multi-modal integration such as vision and hearing, records user habits and preferences through tagged data, and provides exclusive personalized services. At the same time, SenseTime also uses the powerful environment understanding, logical thinking and content generation capabilities of the large model to bring a "cabin brain" that understands users better, and a digital human that can support rapid customization of image and voice for anthropomorphic interaction, bringing A smart cockpit experience integrating safety, entertainment, education and efficiency.
Outside the car cabin, relying on the powerful capability of "large model + large device", SenseTime deploys device-cloud collaboration, unifies traffic entry, and supports privatized deployment and tens of millions of application requirements. In the recent CVPR 2023, SenseTime and the joint laboratory were also the first to propose UniAD, a general-purpose large-scale model for autonomous driving that integrates perception and decision-making, created a large-scale model architecture for automatic driving that targets global tasks, and won the best paper award. It proposes a new direction for the development of autonomous driving technology and industry. Based on this, SenseTime builds a vehicle-road-cloud collaborative traffic system, develops a roadside visual perception large model with the help of a multi-modal multi-task general-purpose large model, combines Qiongyu 2.0 and Passing Object 2.0 to build intelligent traffic twins and simulations, and uses the consultation 2.0 Perceptual reasoning and human-computer interaction capabilities promote the evolution of Cheluyun to large-scale conversational interaction.
Under the wave of new technologies emerging from intelligence, SenseTime has built the long-term competitiveness and innovation cornerstone of the AGI era with large computing power and large models. Innovation and the large-scale application of generative AI lay a new path for long-term development. Facing the future, the fundamental value of large models is to reconstruct the productivity model and bring about paradigm innovation for the implementation of artificial intelligence industry. SenseTime is committed to continuously leaping in the AGI era through day-to-day efficient technology research and development and scene empowerment Out of cognitive limitations, embrace change, take the initiative to innovate, and outsmart the future.