The Intelligent Computing Industry Alliance of Lingang New Area was established. SenseTime’s large devices enable large AI models to be “new every day”

The "Lingang New Area Intelligent Computing Conference" with the theme of "AI leads the era, computing power drives the future" was held on June 2. At the meeting, the Intelligent Computing Industry Alliance of the New Area was formally established. As the computing power provider of the alliance, SenseTime will work with the upstream and downstream enterprises of the intelligent computing industry, as well as universities and research institutes to carry out resource sharing, technical exchanges and project cooperation to promote the new area. Intelligent computing industry applications empower economic development.

SenseTime has laid out the artificial intelligence track in advance, built SenseCore, a large-scale AGI infrastructure device, and built the "SenseTime Daily New SenseNova" large-scale model system on this basis. While promoting its own AGI development strategy, it also provides The industry provides large model algorithm services, training and inference optimization, and data services.

Xu Li, Chairman and CEO of SenseTime, demonstrated the high-precision effect of large-scale 3D scene reconstruction of the real scene in Lingang based on SenseTime's large device. With the combination of large computing power and neural rendering technology, it can be completed in just a few hours Complete high-precision rendering work that previously took weeks or even months to complete manually. He said: "In the era of large AI models, the amount of model parameters will increase at an exponential rate, and the amount of data will also increase massively with the introduction of multi-modality, which will lead to a sharp increase in the demand for computing power."

Enabled by large devices, SenseTime has achieved rapid development in the field of large models. The "Scholar 2.5" multi-modal large model, which was open sourced in March this year, takes the lead in more than 20 authoritative data sets in the three mainstream visual tasks of detection, segmentation and classification. This provides efficient and accurate solutions for general scenario tasks such as autonomous driving and robots. Perception and understanding ability support. Global medium-range weather forecast is one of the most critical prediction tasks in the basic science of artificial intelligence, especially in meteorological and climate forecasting. The large global medium-term weather forecast AI model "Fengwu" launched in April this year achieved effective forecasting of core atmospheric variables at high resolution for more than 10 days for the first time, and surpassed the GraphCast model in 80% of the evaluation indicators. By modeling high-resolution global atmospheric data, "Fengwu" can simulate extreme weather, such as typhoons, to accurately predict typhoon trajectories. UniAD, the industry's first end-to-end autonomous driving solution with integrated perception and decision-making built by SenseTime, surpasses the SOTA method in a number of key data sets and indicators, improving lane line prediction accuracy by 30% and predicting motion displacement. The error has been reduced by nearly 40%, and the planning error has been reduced by nearly 30%. The SenseEarth 3.0 remote sensing large model launched by SenseTime last month not only has the most comprehensive interpretation categories in the industry, but also has achieved technological breakthroughs in many indicators such as interpretation efficiency, generalization ability, and interpretation accuracy.

Not only that, SenseTime has also recently jointly proposed the generalist AI agent GITM, which can survive, explore and create like humans in the best-selling game "Minecraft" that closely simulates the real world, and all in the main world. In the technical challenge, 100% task coverage was achieved, and in the standard task "Obtain Diamonds", it increased by 47.5% compared with the current best result. In the field of large language models, the Chinese language large model application platform "SenseChat" of SenseTime was released in just over a month after its release in early April. Many aspects such as multi-turn dialogue and long text understanding capabilities have been significantly improved, and it ranks among the industry leaders in multiple authoritative test data sets and evaluation lists. SenseTime AI Code Assistant has also achieved a 619% increase in reasoning efficiency through rapid iteration, which means it can support 6 times the number of users with the same computing power resources, greatly reducing usage costs.

Xu Li said: "In the AGI era, the ability of a model can be measured by computing power. We use SenseCore, a large device of SenseTime, to build the infrastructure of the AGI era and update it daily in terms of model iteration speed and problem-solving capabilities. , and continue to unlock more possibilities of AGI. SenseTime looks forward to establishing in-depth cooperation with more peers to discuss new models in the industry and explore new directions in the future." As of May this year, SenseTime's large devices have served more than 40 cores in total. Customers, including more than 10 large model customers, cover cutting-edge fields such as intelligent driving, biopharmaceuticals, chip design, smart business, university scientific research, etc., and have achieved large model delivery in more than 20 implementation scenarios.

Yang Fan, co-founder of SenseTime and president of the large device business group, believes: "Large models are not simply violent aesthetics, but behind them are a large number of software engineering system problems. SenseTime's large devices have accumulated a large amount of expert knowledge and tools for a long time. , with end-to-end product application solution capabilities. The AI ​​industry is about to usher in a more prosperous era of navigation, and SenseTime's large devices will achieve better services through larger-scale, high-efficiency, and low-cost AI infrastructure. Support an industrial ecology where a hundred flowers can bloom in the future."

At this conference, SenseTime also jointly released the "New AI Intelligent Computing Infrastructure White Paper" with the China Academy of Information and Communications Technology, which will analyze the development characteristics and construction needs of new AI infrastructure from multiple levels to adapt to production A major shift in paradigm supports the development of AI into the 2.0 stage. SenseTime also discussed with expert representatives from integrated circuits, artificial intelligence, biomedicine, high-end equipment manufacturing and other industries and scientific research circles how to efficiently transform intelligent computing power into industrial value to empower the core competitive industries of the Lingang New Area In order to focus on seizing new opportunities for industrial development, we will help build Lingang into a new highland for artificial intelligence. SenseTime will work closely with upstream and downstream companies to jointly develop flexible and scalable artificial intelligence infrastructure and promote a prosperous and open AGI industry ecosystem.

The above is the detailed content of The Intelligent Computing Industry Alliance of Lingang New Area was established. SenseTime’s large devices enable large AI models to be “new every day”. For more information, please follow other related articles on the PHP Chinese website!