The nine ＂treasures of the town hall＂ of the World Artificial Intelligence Conference were announced, and they will meet the audience from now on.

The 2023 World Artificial Intelligence Conference opened in Shanghai World Expo Center and Expo Exhibition Hall this morning (July 6), and the nine "treasures of the town hall" that received wide attention also unveiled their mystery and met the audience.

The selection of treasures of the town hall aims to select innovative exhibits with outstanding contributions and leading technologies from exhibitors from the perspective of professionals, encourage innovative achievements in the field of artificial intelligence around the world, guide and motivate artificial intelligence enterprises to pursue excellent quality, and continuously improve the core competitiveness of technologies, products and enterprises.

The conference selected Ant Jian AI Security Detection Platform 2.0 from Ant Technology Group Co., Ltd., Amazon Bedrock from Amazon Tech Services (Beijing) Co., Ltd., Wenxin Yige from Baidu, Graphcore® C600 IPU PCIe Card from Weiwei (Beijing) Technology Co., Ltd., and Shengteng AI’ Big Model Super Factory’ from Huawei Technologies Co., Ltd., The AI Drug R&D Automation Solution of Jingtai Technology, the Yunxiang Intelligent Computing Cluster of Shanghai Suiyuan Technology Co., Ltd., the Q-Truck of Shanghai Xijing Technology Co., Ltd. and the XMusic of Tencent’s Multimedia Lab are the treasures of the nine towns and halls.

What are their characteristics? What new changes will it bring to the industry? Let’s watch together!

Ant identification AI security detection platform 2.0

Exhibitor: Ant Technology Group Co., Ltd.

Booth: Expo Exhibition Hall H2-D101

The pictures in this article are all provided by Shanghai Economic and Information Committee.

Ant Jian AI security detection platform 2.0, which integrates more than a thousand patents, is independently developed to ensure the security, reliability and transparency of AI, and is committed to providing a safe and credible one-stop solution for AI.

In terms of safety, the self-developed induced generation technology is applied to the safety detection of AIGC, which helps the output of AIGC meet the safety standards and social ethics. In terms of reliability, the platform builds a set of practical attack standards, which provides comprehensive and multi-dimensional robustness evaluation services for the AI system, which is helpful for the improvement and optimization of the AI system. In terms of transparency, it integrates logical reasoning, causal inference and other technologies to provide 7 dimensions and 20 evaluation indicators covering completeness and accuracy, and makes a comprehensive and objective quantitative analysis of the interpretation of AI system.

The solution of Ant Jian 2.0 has been applied to 10+ scenarios such as finance, insurance, security, government affairs, intelligent customer service, etc., providing all-round security for the AI system.

Amazon Bedrock

Exhibitor: Amazon Tech Service (Beijing) Co., Ltd.

Booth: Expo Exhibition Hall H1-B109

Amazon Bedrock is a fully hosted generative AI service launched by Amazon Cloud Technology, which provides basic models from leading AI startups and Amazon self-developed through API to help customers build and expand generative AI applications conveniently and safely.

Customers can choose from a variety of basic models to find the model that best suits their business scenarios, and at the same time, with the help of Amazon Bedrock’s serverless experience, they can quickly get started and use it directly, or fine-tune it with their own data on the premise of ensuring data security and privacy, and easily integrate and deploy it into applications through familiar Amazon cloud services and product tools without managing any infrastructure.

Wen Xin yi ge

Exhibitor: Baidu

Booth: Expo Exhibition Hall H2-B101

Wenxinyige is an AI art and creative auxiliary platform launched by Baidu. It is a product and technical tool released by Baidu in the aspect of "AI drawing" relying on the continuous technological innovation of the flying paddle and Wenxin model. It is positioned for people with design needs and creativity, and intelligently generates diversified AI creative pictures based on Wenxin big model to assist creative design and break the creative bottleneck.

Wenxin style can not only inspire visual content creators such as painters and designers, but also provide high-quality and efficient illustrations for text content creators such as media and authors. Based on the large model of Wenxin, Wenshengtu technology behind Wenxinyige is a fully self-developed native Chinese Wenshengtu system, which can provide a variety of AI mapping services, including image editing, skeleton and line drawing recognition, and small sample training.

Graphcore® C600 IPU processor PCIe card

Exhibitor: Weiwei (Beijing) Technology Co., Ltd.

Booth: Expo Exhibition Hall H2-B826

The PCIe card of Graphcore C600 IPU processor focuses on reasoning and training, which can support various mainstream AI applications and has unique advantages in search and recommendation. It provides low latency and high throughput without losing accuracy, helps AI developers to solve the pain point of "both accuracy and speed are difficult", and provides a new path for AI applications to unlock the powerful computing power of IPU, so as to meet the strong demand of customers and machine intelligence practitioners for easy-to-use, efficient and better TCO reasoning products. C600 not only provides powerful computing power, ease of use and flexibility, but also achieves low time delay and low power consumption. When running a typical workload, the cooling design power consumption is only 185 watts, which can greatly reduce the operating expenses of the data center for the operation and maintenance personnel.

Bow IPU is the world’s first 3D Wafer-on-Wafer processor launched by Graphcore. Each chip can provide AI computing power of up to 1.4 petaFLOPS and significant power efficiency improvement for artificial intelligence applications. On the basis of the classic Wafer-on-Wafer IPU of Graphcore, C600 adds FP8 for AI with low precision and mixed precision.

Shengteng AI "Big Model Super Factory"

Exhibitor: Huawei Technologies Co., Ltd.

Booth: Expo Exhibition Hall H1-A701

Shengteng AI "Big Model Super Factory" enables the development and application innovation of big models from the whole process of data-model preparation, computing power preparation-model training and model deployment online-integration.

At present, more than 20 domestic large models including Pengcheng Pangu, the first Chinese NLP large model with 200 billion parameters in the industry, Zidong Taichu, the first multi-modal large model in the industry, and Huawei Yunpangu series have been hatched based on Shengteng AI. At the same time, Shengteng AI has also adapted and supported dozens of mainstream open source models such as ChatGLM, LLaMA, GPT-3 and BLOOM.

In the landing of the big model industry, Shengteng AI joined hands with upstream and downstream partners in the industrial chain to promote the landing of big model technology and exert greater value. At present, multi-modal artificial intelligence industrial consortium, intelligent remote sensing open source ecological consortium and intelligent fluid mechanics industrial consortium have been established. In the future, it is planned to set up consortia in the fields of intelligent breeding and AI biopharmaceuticals to promote the aggregation and development of large model industries.

AI Drug R&D Automation Solution

Exhibitor: Jingtai Technology

Booth: Expo Exhibition Hall H1-B130

Jingtai Technology’s AI drug R&D automation solution realizes the high automation and intelligence of the AI drug R&D experiment process: it not only breaks through the key speed-limiting bottleneck of chemical synthesis in the DMTA cycle of drug R&D, but also improves the efficiency and success rate of drug R&D experiments, and further solves the problem of missing standardized process data in the AI drug R&D experiment, and can record standardized and traceable process data and positive and negative result data in the whole process for the AI model to learn and iterate. Better design reaction steps, form mutual positive feedback of AI, automation experiment and data, and realize the positive cycle of automation for AI and AI for science.

Jingtai Technology has completely developed the operation scheduling system from scratch, and cooperated with scientific research institutions at home and abroad to solve a series of problems in the field of laboratory automation, including difficult process assembly line, ultra-high flexibility, continuous non-standard, complex temperature and humidity, multi-information source acquisition and response, and ultra-high precision requirements. At present, a commercial automation laboratory station cluster has been developed, and continuous high-speed iterative research and development has been carried out to provide more scenarios and higher efficiency ratio.

Yunxiang intelligent computing cluster

Exhibitor: Shanghai Suiyuan Technology Co., Ltd.

Booth: Expo Exhibition Hall H2-C115

Yunxiang intelligent computing cluster is a high-performance artificial intelligence acceleration cluster for large-scale intelligent computing centers, which aims to meet the demand for ultra-large-scale computing power in the AIGC era. The cluster is equipped with an energy-efficient AI computing chip, Guisi, which won the first prize of the "2022 Wu Wenjun Artificial Intelligence Special Award" chip project, and can support the cluster of over kilocalories and break through the E-class computing power; It is the first liquid cooling scheme based on OAM open computing standard in the industry, and the PUE can be reduced to below 1.1.

At present, based on this cluster product, the cluster has been trained on a kilocalorie scale in Zhijiang, a national key laboratory, providing more than 100P of advanced AI computing power, so as to efficiently support a number of AIGC applications, including the application of melting media, text generation PPT, cross-modal image generation, and the development and frontier exploration of various AI4S scientific computing applications, such as computational genes, computational pharmaceuticals and computational materials.

Yunxiang intelligent computing cluster can be widely used in smart city computing bases and research institutes to develop computing clusters, and meet the standardized AI computing needs of other enterprises and institutions.

Intelligent power-changing unmanned commercial vehicle Q-Truck

Exhibitor: Shanghai Xijing Technology Co., Ltd.

Booth: Expo Exhibition Hall H1-C121

Q-Truck, the world’s first intelligent power-changing unmanned commercial vehicle, is equipped with intelligent energy service PowerOnair, which can complete unmanned battery recharge in 5 minutes, and is suitable for many logistics scenarios such as land, sea, air, iron and factories. Q-Truck also participated in the movie Wandering Earth 2 as the only driverless truck.

Q-Truck, as a driverless brand developed and designed by China and made in China, has independent intellectual property rights, which helps users to realize the mixed operation of unmanned driving and manual driving for the first time in the world, and realize the normal operation of unmanned truck collection without security officers for the first time. Its products have been commercialized in Thailand, United Arab Emirates, Britain, Malaysia, Mexico and other countries, marking the successful landing of China AI unmanned business plan and providing a brand-new sample for intelligent innovation and green development of global smart logistics.

Tencent multimedia lab XMusic

Exhibitor: Tencent

Booth: Expo Exhibition Hall H2-B108

Tencent Multimedia Lab, based on AIGC technology, recently launched XMusic generative universal composition framework, which supports multi-modal content such as video, pictures, text, labels and humming as input prompts to generate high-quality music with controllable mood, style and rhythm.

Based on the self-developed multi-modal and sequential modeling technology, XMusic can parse the cue words into the symbolic music element space, and use this as a control condition to guide the model to generate high-quality music with beautiful and coherent music, obvious melody fluctuation and alternating rhythm. Whether it’s a carefully shot short video, a meaningful photo, or a line of words, users can use XMusic to create their own music that fits the artistic conception. At the same time, the commercial-grade music generation ability of XMusic has broad application prospects in many fields, such as video soundtrack, interactive entertainment, auxiliary creation, music education and so on.