Source of this article: Times Finance Author: Xu Dan
Image source: People's Vision
On the occasion of the two sessions, Li Yanhong, member of the National Committee of the Chinese People's Political Consultative Conference and chairman of Baidu, submitted three proposals, focusing on the innovation of autonomous driving policies, promoting the popularization of intelligent transportation to alleviate traffic congestion, and helping carbon emission reduction and development of "green AI" themes. Overall, AI is still the key word in Robin Li's proposal.
Autonomous driving is one of the directions of Baidu's vigorous development in recent years, and at present, mainland autonomous driving has entered a critical period of landing, and it is now necessary to break through the policy bottlenecks that are incompatible with the development of technology and industry.
Intelligent transportation is the focus of Li Robinhong for seven consecutive years, in view of high carbon emissions, serious congestion and other traffic problems, Li Robinhong suggested that the vehicle be electrified and intelligently upgraded, the establishment of intelligent transportation to help carbon emission reduction benefits evaluation standards and personal carbon credit incentive system pilot.
Green AI is the first time it appears in Robin Li's proposal. The 14th Five-Year Plan is a critical period for the mainland to accelerate the promotion of green and low-carbon development, and green AI-related technologies will surely become the main contributor to carbon reduction in the future.
Accelerate the innovation of unmanned autonomous driving policies
In 2021, the state deploys a series of innovative policy work, such as the Ministry of Public Security launched the revision of the Road Traffic Safety Law, and the Ministry of Industry and Information Technology and the Ministry of Communications have also introduced policies around product access and application pilots. Local governments are more motivated and faster to explore the "China model" of autonomous driving.
For example, Beijing opened the country's first pilot project for the commercialization of autonomous driving travel services; Shenzhen tried to explore local legislation in the fields of intelligent networked vehicle access management and accident liability identification; and Guangzhou launched a pilot project of mixed driving of intelligent connected vehicles and human-driven vehicles. Benefiting from a good innovative policy environment, autonomous vehicles such as self-driving taxis, buses, and unmanned delivery vehicles have begun to enter the daily lives of ordinary people.
For example, Baidu has provided autonomous travel services to citizens in 8 cities, including Beijing, Shanghai, Guangzhou, and Shenzhen, and has received more than 500,000 passengers. Mainland autonomous driving has entered a critical period of landing, technology has run and even led the world, but the development of high-grade autonomous vehicles in mainland China still faces many problems, and it is necessary to further break through the policy bottlenecks that are incompatible with the development of technology and industry, and stimulate innovation capabilities in the field of automatic driving. In this regard, Robin Li put forward the following suggestions:
The first is to guide and support local governments to introduce policies, clearly support unmanned vehicles without safety personnel on the road, and create a pilot area for manned operation policies for fully unmanned autonomous vehicles.
The second is to accelerate the revision and implementation of the Road Traffic Safety Law, and lay a legal foundation for accelerating the large-scale commercial and unmanned use of autonomous vehicles from the national level.
The third is to build intelligent transportation infrastructure in advance, give full play to the remote control advantages of 5G, promote a significant improvement in traffic efficiency and safety through vehicle-road collaboration, drive the transformation and upgrading of the automobile industry to intelligence and networking, and let the truly unmanned intelligent networked vehicles drive on China's roads as soon as possible.
Promote the popularization of intelligent transportation and alleviate road congestion
Carbon dioxide emissions from the transportation industry account for about 9% of the national total, of which road traffic emissions account for more than 80%. Intelligent transportation provides a new and effective path for alleviating carbon emissions in transportation, and the electrification and intelligence of vehicles, the networking of roads, and the sharing of lines are the core focus points. According to the International Data Corporation (IDC), intelligent transportation technologies such as driverless and operation, intelligent information control, smart parking, and MaaS one-stop travel services contribute more than 40% to energy conservation and emission reduction.
Robin Li's advice is as follows:
The first is the electrification and intelligence of the car. Accelerating the popularization of intelligent and connected vehicles and promoting the large-scale application of high-level autonomous vehicles can make everyone travel safer and improve urban congestion. The second is the networking of roads. Through the development of vehicle-road coordination, intelligent information control, intelligent parking and other technologies, improve the efficiency of road transportation, reduce carbon emissions, and alleviate the pressure of urban purchase restrictions. It has been estimated that intelligent transportation based on vehicle-road collaboration can improve road traffic efficiency by 15%-30%. For example, in Baoding, Baidu has carried out regional traffic governance by deploying intelligent information control systems at more than 100 intersections. Vehicle travel time is shortened by an average of 20%, and the average carbon emission reduction per intersection is 138.6 tons/year. The third is the sharing of banks. Encourage people to use public transportation more through MaaS one-stop travel services and green travel personal carbon credit system.
In recent years, in the field of intelligent transportation, the state has accelerated the introduction of relevant policies, and local governments have also actively promoted the landing. However, there is still a lack of coordination in construction and a lack of unified standards for the assessment of carbon emission reduction benefits in some areas, which restricts the full play of intelligent transportation to help carbon emission reduction. It is therefore recommended that:
The first is to accelerate the promotion of the intelligent transportation operator model. Support and encourage local governments to play a leading role, coordinate the planning, construction and operation of intelligent transportation projects, and promote the transformation of one-time integrators to a sustainable operator model. Guide state-owned enterprises or platform companies in charge of the government to accelerate the construction of intelligent transportation operator functions or set up special companies, and support technology enterprises to empower intelligent transportation operators with technical and operational experience. Encourage local governments to appropriately deploy new intelligent transportation infrastructure in advance, strengthen the pilot area of the Internet of Vehicles and the coverage area of the "double intelligence pilot", and explore the application value through scale.
The second is to establish evaluation standards for the benefits of intelligent transportation to help carbon emission reduction. Unite industry associations, universities, scientific research institutions and leading enterprises to accelerate the study of carbon emission reduction benefit evaluation standards, formulate quantitative calculation rules for intelligent transportation to help carbon emission reduction, so that the originally difficult to evaluate intelligent transportation to help carbon emission reduction benefits become quantifiable, statistical and evaluatable, explore the carbon reduction effect evaluation of new entities such as autonomous driving operators, and determine its carbon reduction value.
The third is to carry out a pilot project of the individual carbon credit incentive system. Support local governments to introduce incentive policies, build a public carbon emissions-related data platform, explore a mechanism for linking individual green travel carbon credits with preferential policies for public services, and enhance the public's sense of green and low-carbon travel.
Accelerate the development of green AI and guide the low-carbon development of computing algorithms
Computing power algorithms are an important productive force in the era of digital economy. Led by the "3060 double carbon" goal, it is necessary to continuously improve energy efficiency and develop green AI that is more environmentally friendly. On the one hand, it is to develop green computing power. Data centers are an important carrier of computing power.
According to the Calculation of the Chinese Academy of Information and Communications Technology, in 2020, the average annual PUE of mainland data centers will be 1.56, which is still far from the goal set by the state to reduce the PUE of new large-scale data centers to less than 1.3 by the end of 2023. Local governments are strengthening energy-saving reviews of data centers, and many leading technology companies are using green electricity and using technology to optimize processes and reduce data center energy consumption.
Baidu Yangquan Center introduced artificial intelligence technology into the data center for the first time, based on Baidu's self-developed flying propeller deep learning framework, established a data center deep learning model, realized the AI tuning of the cold source part of the system, and the average annual PUE of the single data center with the highest energy efficiency was as low as 1.08.
Another aspect is the development of greener algorithms. At present, hyperscale pre-training models have become one of the important evolution directions of deep learning, but energy consumption has also increased. It is necessary to optimize the strategy and parameter settings, build a large model of green intensification, and improve the energy efficiency ratio of the infrastructure.
Technologies such as artificial intelligence have opened up new paths for the low-carbon transformation of the whole society. The contribution of green AI-related technology to carbon reduction will increase year by year, and green computing power and algorithms are applied to various industries such as industry, energy, construction, and finance, with great potential.
The "14th Five-Year Plan" is a critical period for the mainland to accelerate the promotion of green and low-carbon development, and the state has successively made a series of major deployments, such as the recent full launch of the "East Counting West" project. Green AI has great development potential and wide application, but it also faces the situation that the use of green electricity in data centers is still facing dual control assessment of energy consumption, the green algorithm measurement standard system has not yet been established, and the lack of a sound carbon emission statistics and accounting system is lacking. Robin Li suggested strengthening policy guidance and improving the standard system, as follows:
The first is to explore the dual control of carbon emissions in the data center. According to the Calculation of the National Information Center, data centers only use about 2% of the electricity consumption of the whole society, supporting the scale of the digital economy, which accounts for about 36.2% of the country's GDP. It is recommended that relevant departments support areas with conditions to increase policy innovation, and the amount of green electricity used in data centers is not included in the energy consumption assessment.
The second is to establish a green algorithm metric that takes into account both performance and energy consumption, and advocate leading artificial intelligence enterprises to build pre-trained large models with high energy efficiency and excellent performance, and open it to the industry.
The third is to accelerate the improvement of the carbon emissions statistical accounting system and open the issuance of national certified voluntary emission reductions. Accelerate the establishment of a methodology library and industry standards for scientific and technological emission reduction, and strengthen the technical supply of green AI in the field of reducing carbon emissions.