Baidu released a number of the latest AI technology achievements: virtual digital human, super realistic digital human live broadcast platform, autonomous vehicle

Published on 2 Weeks ago   263 views   0 Comments  

Finishing | Liu Yan

Source:AI front

Original title: autonomous vehicle without steering wheel is coming! The era of "one digital person per capita" is also coming? Baidu's AI is new in the world

On July 21, at the 2022 Baidu World Conference, baidu released a number of Baidu's latest AI technology achievements, including the autonomous vehicle Apollo RT6 without steering wheel, the sigaga virtual digital human, the world's first large model in the aerospace field


The era of "one digital person per capita" has come


Virtual digital human can be said to be the "protagonist" of this Baidu World Conference.

"Xijiajia" and "C" of Du Xiaoxiao station

"Sigaga", a new member of Baidu AI digital people family, is the AI planner, deputy host and opening guest of this conference, and will preside on the same stage with SA Beining. She and another digital person, Du Xiaoxiao, appeared in many links of the whole conference. "Sigaga" also has many "super abilities" such as painting, poetry writing, music composition, multilingual speaking, real-time live broadcasting and so on.ย ย 

He Junjie, senior vice president of Baidu group and general manager of Baidu mobile ecosystem group (MEG), said, "the era of one digital person per capita has come."

He Junjie showed the latest progress of Baidu AI digital human "Du Xiaoxiao". He believes that Du Xiaoxiao plays the role of "Ai ambassador", connecting technology to the last mile of users.

"Du Xiaoxiao" is an important evolution of Baidu's search ability. At present, "Du Xiaoxiao" has been launched on Baidu app. Users can directly search "hello" on Baidu app to call Du Xiaoxiao.

It is reported that in Baidu app, Du Xiaoxiao can connect all smart life services: in the face of personalized questions from users, Du Xiaoxiao quickly matches professional respondents through Baidu "ask and ask" to provide one-to-one real person consulting services. At present, the "ask and ask" has covered more than 40000 respondents in 19 fields such as law, emotion, psychology and government affairs.

The understanding and generation ability of digital people Du Xiaoxiao and higaga created by Baidu AI can make them automatically generate creative works. At the meeting, Gong Jun digital man and Du Xiaoxiao performed the aigc single "every minute, every second, every day". Du Xiaoxiao can complete 40 college entrance examination compositions in 40 seconds and an art work in dozens of seconds. He can also create talk shows, live broadcast and bring goods, etc. with 24-hour uninterrupted emotional interaction.ย 

Robin Lee, the founder, chairman and CEO of Baidu, shared his thoughts on the field of aigc. "In the next decade, aigc will subvert the existing content production mode, and can generate AI original content at a 'tenth of the cost' and a hundred times the production speed.

Robin Lee judged that aigc will go through three development stages: the first stage is the "assistant stage", which is used to assist human beings in content production; The second stage is the "cooperation stage", in which aigc appears in the form of virtual human with virtual reality coexisting, forming a situation of man-machine symbiosis; The third stage is the "original stage", in which aigc will complete content creation independently.


Super realistic digital human live broadcast platform release


"The newly released digital people live broadcasting platform allows AI digital people to do live broadcasting by themselves. We interact with users all year round to help those rural uncles and small and micro enterprises in mountainous areas who are not good at e-commerce to improve their lives." At the meeting, AI digital man sigaga introduced the Xiling digital man live broadcast platform of Baidu AI Cloud. The platform mainly focuses on e-commerce live broadcast, brand marketing, interactive entertainment and other fields, so as to achieve 24-hour pure AI live broadcast of super realistic digital people.ย 

At present, the live broadcasting industry is hot. However, behind the explosion of live broadcast, the weak stability of live anchors, the high cost of live broadcast hardware and scenes, the high operating threshold, and the inability to effectively use idle traffic have also become constraints to its development.

Liu Wei, CEO of chuangjiang technology, said, "taking the team's construction of a live broadcast room for the sale of electric toothbrushes as an example, only the hardware cost is at least 70000 yuan, not including the venue and anchor.". Baidu AI Cloud Xiling realizes the rapid generation of content and reduces the operation cost through the AI driven way, without the need for live anchors and central control personnel.

There are only three core reasons for judging the popularity of a live anchor: first, the anchor is highly expressive, and the appearance and style of dressing are memorable; Second, have personality, be able to chat, talk with the audience and reply wisely; Third, keep changing, infiltrating into every time gap of public life. To make digital people in the live broadcast room infinitely close to real people, the core test is the expressive and interactive power of digital people.

In terms of expressiveness, the accuracy of digital people of most products in the current market is low, and "paper people" can easily make the audience "play". Baidu AI Cloud has been researching and developing the production of super realistic digital human for three years. Through the technology of digital human natural action engine, the image of AI digital human produced by the digital human live broadcast platform is friendly, and the action is smooth and natural. Users can feel the "true feelings" of AI digital human.

At the level of interaction, the essence of anchor is content. Relying on low-level full stack AI technologies such as vision, NLP, voice interaction and the Plato 10 billion parameter dialogue model, Baidu AI Cloud's digital human live broadcast platform can make digital people laugh and reply wittily with audiences like real people. At the same time, intelligent creation supported by AI big model will also help digital people have "original ability" in the future.

To achieve "one digital person per capita", standardized products will become the key to reducing the operating threshold.

Xiling digital human live broadcast platform, as a SaaS product, can realize "plug and play". Without a complete operation team and expensive live broadcast hardware equipment, one person can complete the whole process, making digital human live broadcast as simple as writing documents and PPT. At the same time, the digital people live broadcast platform has also opened up the whole service process, providing partners with live broadcast agent operation hosting, zhongzhiren training and other services.ย ย 

Liu Qian, deputy general manager of Baidu AI Cloud AI and application products, said, "the launch of the digital human live broadcast platform will enable everyone to have a more stable, ultra realistic and intelligent digital anchor with a low threshold, and truly liberate enterprise productivity".

The development of hyper realistic digital human also poses new challenges to visual technology. Ding Errui, director of Baidu's visual technology department and augmented reality technology department, said that Baidu is about to release "photo guided image generation technology". Taking a few or even one photos with a mobile camera can produce a realistic, beautiful and driveable hyper realistic digital portrait of everyone, which will improve the efficiency of 90% compared with the traditional production process.

It is reported that at present, dozens of digital people created by Baidu AI Cloud Xiling have landed in all walks of life. CCTV net small C, CCTV news AI sign language anchor, and higaga have participated in large-scale live broadcasts.


Baidu's sixth generation mass-produced unmanned car Apollo RT6 released


At the meeting, baidu released the sixth generation mass-produced unmanned car - Apollo RT6.

Based on the breakthrough of automatic driving technology, Apollo RT6 not only has the driverless ability of complex urban roads, but also costs only 250000 yuan.

Robin Lee said: Baidu reduced the cost of autonomous vehicle to one tenth of the industry, equivalent to the price of an ordinary new energy vehicle. The cost of driverless cars has fallen sharply, so that we can deploy tens of thousands of such cars all over the country. It will be half cheaper to take a taxi in the future than it is now.

According to the plan, Apollo RT6 will be the first to be put into use in radish express in 2023, and tens of thousands of vehicles can be deployed across the country in the future. Its mass production will accelerate the large-scale deployment of unmanned vehicles.

Apollo RT6 is a mass-produced vehicle independently developed and designed by Baidu for future travel. The whole vehicle is deeply designed for passenger needs and driverless travel scenarios.

It is reported that the appearance of Apollo RT6 adopts the space shuttle body and Galaxy waist line, highlighting the sense of technology and security; The innovative keel skylight highly integrates the roof sensor and skylight structure, innovating the shape of the self driving refitted vehicle; In order to further improve the convenience of travel, the vehicle is also equipped with Smart Interactive lights with recognition and prompt functions, intelligent electric side sliding doors and other functions.

In terms of intelligent cockpit, Apollo RT6 has created a versatile intelligent space. Apollo RT6 supports two modes: steering wheel and no steering wheel. The front row can be equipped with seats, vending machines, desks, game consoles, etc. according to different travel scenarios to meet the diverse needs of passengers for office and entertainment. In addition, the 1050mm large space in the rear of Apollo RT6, independent seats and the backward moving design of intelligent interactive system realize the size of A-class car, the space of B-class car and the experience of C-class car.

Apollo RT6 comes from the "Apollo Galaxy" architecture platform developed by Baidu. As the first model of this platform, Apollo RT6 realizes 100% vehicle specification level and full redundancy system of the whole vehicle, which is two to three orders of magnitude higher than the reliability of modified vehicles on the market, so as to ensure the safety of passengers. Apollo RT6 hardware has seven redundant systems, such as architecture redundancy, computing unit redundancy, braking system redundancy, etc. if any single component or system fails, the standby redundant system can instantly complete the filling; In terms of software, it is equipped with a fault diagnosis and risk degradation system integrating the whole vehicle and auto drive system.

Lizhenyu, senior vice president of Baidu group and general manager of intelligent driving business group, said that Apollo RT6 is both a car and a good driver. Its self driving technology is equivalent to a driver with 20 years of driving experience. Apollo RT6 is equipped with Baidu's latest generation of driverless system, which has stronger L4 level automatic driving ability than the previous generation of models, and can cope with all kinds of complex roads and scenes in the city. Apollo RT6 is equipped with a vehicle specification level high computing power redundant dual computing unit, and the computing power can reach 1200tops. The 38 sensors in the whole vehicle are deeply fused to realize the full coverage of far, middle and near triple detection capabilities, with more accurate perception.

In the past nine years, baidu Apollo has launched six generations of unmanned cars. Compared with the previous five generation models, Apollo RT6 realizes the deep integration of AI technology and vehicle engineering, and has the advantages of high safety, high quality and low cost.


AI helps realize "green light freedom"


Baidu believes that the future urban intelligent transportation should not only have "smart cars", but also "smart roads".

Vehicle road coordination can effectively solve important problems such as safety, congestion and carbon emissions. Robin Lee believes that "intelligent transportation based on vehicle road coordination can improve traffic efficiency by 15%-30% and promote the absolute growth of GDP by 2.4%-4.8% per year; with the in-depth practice of intelligent transportation, China's first tier cities will no longer need 'purchase restriction' and 'traffic restriction' within four years. Intelligent Transportation system is expected to reduce traffic safety accidents by 90%

According to the latest data disclosed at the conference, Baidu's intelligent transportation scheme has been practiced and verified in more than 50 cities across the country, including Guangzhou and Baoding, among which AI information control technology is the most advantageous.

At this meeting, the achievements of intelligent transportation displayed by Baidu are more focused on infiltrating AI into people's real life. Through three cases, baidu shows how intelligent transportation construction can help urban managers make accurate decisions and bring "green light freedom" to ordinary people.

There are 17million truck drivers in China. In the scene of truck logistics transportation, the highway closure caused by fog, rainstorm and other extreme weather often delays the delivery time, and there are also safety problems. Baidu intelligent expressway system applied on Beijing xiong'an expressway has broken through the technical problem of quasi all-weather traffic. In fog and other adverse weather conditions, drivers can receive reminders of driving information such as weather, vehicle distance, speed and front collision sent by intelligent terminals, so as to ensure the smooth and safe passage of drivers. At the same time, with the improvement of fusion perception and AI algorithm, the deployment distance of the equipment was successfully extended to 1 km, and the cost was reduced by 30%.

In Yongchuan District of Chongqing, AI can help traffic police "command" traffic. With the participation of AI, signal lights can be automatically timed, and applications such as river closure control and dynamic green wave can continuously optimize urban traffic. The front-line traffic police don't need to stand in the scorching sun to command the traffic, but can grasp the congestion macroscopically in the command Hall of the smart city. The data shows that after the application of AI, the handling efficiency of traffic police increased by 35%.ย ย 

In a super large city like Beijing, traffic congestion occurs frequently, especially on the way to and from work. The construction of intelligent intersection in Beijing Yizhuang has solved this pain point. Li Zhenyu, senior vice president of Baidu group and general manager of intelligent driving business group, described the effect of intelligent intersection as "counting cars by lights". He said that the construction of intelligent intersection makes citizens no longer rely on luck to travel with green lights all the way. For example, under the action of 332 smart intersections, one trip can reduce the commuting time of user Yang Xiaofang by 20 minutes.


The world's first large model in the aerospace field was released


Baidu CTO Wang Haifeng and Liu Jizhong, deputy commander-in-chief of the China lunar exploration project and director of the National Space Administration lunar exploration and space engineering center, released the world's first large model in the space field - "aerospace Baidu Wenxin model".

Aerospace - Baidu Wenxin integrates learning from data and knowledge in the aerospace field, can intelligently collect, analyze and understand Aerospace data, and help make technological breakthroughs in deep space intelligent perception, planning and control.

Liu Jizhong said, "the application of the 'Aerospace Baidu Wenxin big model' can encourage researchers to devote more energy to innovation and creation. In the future, the two sides will give full play to their respective advantages and carry out comprehensive and in-depth cooperation in deep space exploration intelligent technology research and development, application platform development and project implementation, space science and technology innovation, science popularization and personnel training."

Wang Haifeng also shared the practical application of the open source open platform for in-depth learning of the propeller industry to help herdsmen, farmers and power plant workers use AI. For example, in the pasture, the ewe delivery prediction and early warning system developed based on the propeller can reduce the workload of manual nursing in the process of ewe delivery and reduce the rate of dystocia; Agronomic experts use flying oars to ensure that intelligent agricultural machines such as driverless tractors operate accurately according to the predetermined trajectory; In the power plant, the intelligent fault prediction system of power plant equipment based on the propeller can intelligently diagnose equipment faults and conduct efficient maintenance.

Data shows that at present, the propeller has gathered 4.77 million developers. In terms of industrial knowledge enhancement Wenxin big model, Wenxin family has released more than 20 big models, including Pengcheng Baidu Wenxin, State Grid Baidu Wenxin and SPD Baidu Wenxin jointly released.

It is worth mentioning that Baidu used AI to restore the fragments of Fuchun Mountain Residence, one of China's top ten handed down famous paintings, at the conference. Based on the literary heart model, it "complements" the fragments of the famous painting "Fuchun Mountain Residence", which is handed down from generation to generation. After 300 years, the mountains and rivers are merged into one, and a poem is inscribed, "one peak, one shape, hundreds of trees welcome, and Tianshui meets the cross-strait situation". Now, search "Fuchun Mountain Residence map" on Baidu app, and everyone can participate in the "virtual completion" of "Fuchun Mountain Residence map" and create their own "Fuchun Mountain Residence map".

New report card of Baidu AI Cloud: release Kaiwu, Kyushu

Shen Shuang, executive vice president of Baidu group and head of Baidu AI Cloud business group, shared Baidu's technological achievements in enterprise intelligence upgrading.

Digital transformation and upgrading has become a scientific and technological trend that cannot be ignored at present and in the future. Among them, cloud computing is one of the important infrastructures to promote digital transformation. Baidu AI Cloud focuses on its ability to "integrate cloud and intelligence". First, it cuts into the core scenarios of key industries, accumulates industry experience, and then deposits the general needs of different industries into general AI products, so as to build them into standardized products, reduce the threshold of AI use, and help enterprises from digital driving to intelligent traction.

In helping traditional industries to carry out intelligent transformation, baidu showed application cases in power, energy, water conservancy, manufacturing, agriculture and other industries. Application practice and data show that AI fan inspection can replace power workers to rush to the most dangerous place on the front line, and improve the inspection efficiency by up to 10 times; In the intelligent reconstruction project of Shijiazhuang urban community heating system, through intelligent management and scheduling, the whole Shijiazhuang urban heating network can save 20% energy

In order to help the intelligent transformation of traditional industries and reduce the threshold of AI use. Shen Shuo released a new version of Kaiwu 2.0 industrial Internet platform and Jiuzhou District brain on the spot. At present, Kaiwu 2.0 has accumulated more than 200 industrial solutions, precipitated 38000 industrial models, covering 9 major fields such as quality control, safety production, energy conservation and emission reduction, and manufacturing, providing digital services for enterprises in the industrial field to go to the cloud, use data, and endow intelligence.

In addition to Kaiwu 2.0, the conference also released the Baidu AI Cloud Jiuzhou district and county brain, which uses digital technology to promote grass-roots urban and rural governance and incubate digital applications of local industries, culture and ecological scenes.


AI intelligent hardware: building a "future home" in a small way


In terms of AI intelligent hardware, Jing Kun, CEO of Xiaodu technology, released three new products focusing on health on the spot - Xiaodu Tiantian intelligent fitness mirror M30, Xiaodu voice intelligent alarm clock and Xiaodu large screen eye protection learning machine P20.

It is reported that Xiaodu Tiantian intelligent fitness mirror M30 can improve the boring and difficult pain points of fitness. It has 17 kinds of fitness courses, including somatosensory games, double games, and multi person online fitness mode, which can meet the fitness needs of people of all ages and enhance the fun and interaction of sports. Xiaodu Tiantian intelligent fitness mirror also has the functions of "Ai matchmaker projection technology" and intelligent action guidance. In addition to fitness functions, Xiaodu Tiantian intelligent fitness mirror M30 can also listen to songs, karaoke, cast screens, brush short videos, and broadcast weather and news.ย ย 

Xiaodu voice intelligent alarm clock is jointly built by Xiaodu and many experts in the field of sleep, such as the center for excellence and innovation in brain science and intelligent technology of the Chinese Academy of Sciences. It can recommend suitable white noise and light music for users based on professional sleep tests, and create exclusive sleep assistance solutions. Small voice intelligent alarm clock can also monitor sleep quality and environment during sleep, and generate sleep quality reports. Xiaodu voice intelligent alarm clock also has infrared remote control function, which can control new and old household appliances at home in one sentence, creating a sleepy environment.

The small and large screen eye protection learning machine P20 is equipped with a 15.6-inch AI eye protection screen, which has 20 eye protection functions, such as anti blue light, anti glare, no stroboscopic, and adaptive adjustment of AI screen light. At the same time, based on the ability of AI camera, it can monitor the child's sitting posture in real time. When it finds the wrong sitting posture, it will actively send a voice reminder to correct it. In addition, it covers the whole subject content including preschool education and the whole stage of compulsory education, and is equipped with AI precision learning system to help students diagnose difficulties and weaknesses in learning.ย 

Jing Kun also announced at the scene that in the future, 100 full house intelligent experience stores will be added offline to accelerate the implementation of the small "future home".


Generic placeholder image
Promote your coin to 10k unique users daily
contact us PM Twitter
130 views   0 Comments   3 Days ago
51 views   0 Comments   3 Days ago
106 views   0 Comments   6 Days ago
211 views   0 Comments   1 Weeks ago
183 views   0 Comments   1 Weeks ago
133 views   0 Comments   1 Weeks ago
164 views   0 Comments   1 Weeks ago