Monday , May 17 2021

Artificial intelligence "throws the threshold" to promote innovation Experts: Be cautious about security risks



A group of open artificial intelligence platforms greatly reduces the technical threshold, enabling artificial intelligence to access SMEs, it is no longer a "luxury" that can only be owned by a financially strong person. Experts believe that the open platform is a new trend in the development of artificial intelligence technology, which will promote the development of artificial intelligence at the top, but at the same time will inevitably bring more security risks and must be prevented.

Artificial intelligence "lower threshold" to promote innovation

The Beypoys and Drone Bush "Apollo" minibuses recently conducted a public check in Beijing Haiyan Park. This unmanned vehicle is equipped with a Baidu Apollo system, which can be based on high-precision maps and intelligent sensors. Optimal planning.

More than just Apollo, in more than a year since the opening of the Baidu Apollo Pilot Platform, there are 133 ecological partners in the world and more than 11,000 developers.

"Leading giants in the downstream and downstream of the automotive industry, including Daimler, Ford, Volvo, Intel, Microsoft, BYD, BAIC, Chery and others, joined Apollo's ecology. The development of artificial intelligence technology, which Baidu has accumulated over the years, has accelerated the innovation of entire ecology through open source methods, "said Zhang Yakin, president of Baidu.

Some time ago, the 5th World Internet Conference took place in Wuzhen, Zhejiang, Baidu Apollo Autopilot Open Platform and the platform for a fully-fledged Amazon Machine Learning Platform in the direction of China's new intelligence platform Zhou Yi, supET Industrial Internet Platform, Jingdong Intelligent Supply Chain At the conference, a number of artificial intelligence platforms were published, such as the service platform and the Xiaomi Smart Home Open Platform platform, as an important component of the 15 leading scientific and technological achievements in the etotyl online.

As the Chinese director of the World Council for Scientific and Technological Advancements for the Internet, Academician He Hedgeen of the Chinese Academy of Sciences talked a lot about the results of the artificial intelligence platform. "The results of the publication were voted by 47 world-renowned Internet industry experts who represented the innovative wisdom in the Internet field and the future development trend of the Internet." The platform has become the leading technological advancement "main event." He Hezhen believes that the results similar to the platform will create environmentally friendly Suitable for other small and medium enterprises to use the platform and technology for further development.

On the open platform Baidu Autopilot, developers can easily use related technologies without even learning. The new Zhou Yi artificial intelligence platform in China could reduce the developmental cycle of artificial intelligence by three to five times, and reduce the cost of computing power development by five to ten times. The Amazon Machine Learning A complete managed platform, trained programmers can complete projects that require 10 to 20 doctoral experts to complete in 9 months, in just three to four weeks.

"Almost all artificial intelligence devices today are targeted at specific application scenarios, vertical integration, low industrial synergy, development costs often reach several hundred million yuan, and the cycle will last for several years. Our industry urgently needs to have open adaptability and interoperability Labor Intelligent Platform for Hardware and Software to Reduce the Cost of Development, "said Wu Xiong An, CEO and CEO of Anmou China.

"Everyone knows the importance of artificial intelligence, but with companies like Google, Tencent, Baidu, there are not many companies that can really develop advanced artificial intelligence technology. So we open up the technological capabilities accumulated in the past." Chairman and CEO Li Yanhong said.

At the time of wisdom, "fast" is not just "everyday," but can be calculated for hours and minutes. "Wang Endong, an academician at the Chinese Academy of Engineering and chief scientist at Inspur Group, said the speed of innovation is the development of artificial intelligence. Basic competitiveness.

More security risks are worthy of caution

Many industry experts and insiders have said that the development of artificial intelligence has experienced very low levels, and the current cycle of development is accelerating. There is consensus in the industry and it should be fully acknowledged that the security problems of the artificial intelligence will develop. "Dynamic Evolution".

Artificial intelligence "throwing thresholds" will create more hackers, more viruses and shorter attack cycles. Hehe said that in the area of ​​network security, the popularity of artificial intelligence will help identify network security risks, and also reduce the threshold of hackers, allow more people to learn hackers, more and more viral variants , and the cycle will become more and more brief.

"With the advancement of artificial intelligence, the characteristics of artificial intelligence will be intensified with a" double-edged sword, "Hehe said.

Zhou Hongyi, president and CEO of the 360 ​​Group, said that cyber attacks today were intelligent, automated and extremely fast. "Monitoring network security often finds some network attacks. After the discovery and analysis of the survey, the solution is not implemented. The network attack has passed and new attacks have begun."

Sensors, training data and open source software use artificial intelligence have security risks. The White Paper on Artificial Intelligence, published by the Chinese Institute of Information and Communications, recently showed that the research and development of products and applications for artificial intelligence is mainly based on the framework for learning artificial intelligence and components published by tech giants like Google, Microsoft , Amazon, Facebook, Baidu, etc., and these open source and Component boxes do not have a stringent security management and certification test, while vulnerabilities and backlinks are vulnerable to malicious exploitation .

"There is a possibility of" data poisoning ", trying to mix unusual data in the training data, and the trained system of artificial intelligence will make mistakes. If the software is written by someone, there must be loopholes, such as 1000 lines of code There will be 4 holes in it. "Zhou Hongyi said that due to the inevitable holes, the current network attack and defense is very uneven. "Strictly speaking, traditional intrusion detection, firewall, anti-virus software technologies have become powerless in defending such attacks."

"At present, the ecology of artificial intelligence is not fully established, especially in a relatively fragmented state in all aspects of the industry, which means hackers can use it." Jiang Wei, Senior Researcher and Artificial Intelligence Expert, Blue Shield, said artificial intelligence Security risks include denial of service, disclosure of information, system hijacking, etc., which can lead to missed judgments, misconceptions, systemic accidents, kidnappings, or making smart devices in zombie attack tools.

You need to build a risk management and management system

"Artificial intelligence has multidisciplinary and very complex features. We need to plan and constantly advance and focus on strengthening originality." Experts like He Hejun said that while they continue to increase the research and development of artificial intelligence and emphasize the potential of applications for artificial intelligence, it is necessary to override the challenge of artificial intelligence and to "strengthen and avoid weaknesses" of technology and law to prevent risks.

Above all, it is necessary to break through the key core technologies of artificial intelligence. Wang Endong believes that domestic enterprises for artificial intelligence and scientific research institutions must have a clear understanding. We still have obvious gaps at the basic level, and we must strengthen the basic research and contribution. He Hezhen said that the application of artificial intelligence has just begun, but is still in its infancy, and at the moment there is no way to explain the cause-effect relationship. More hopes are being put on the advancement of artificial intelligence.

Second, data silos should be opened to improve security protection. "Network security is the whole process of" leaving the whole body. "The big problem we face is that every company" individually fights, "the data are not open and there is no global or national vision, and the process of real cyber attacks can not be seen. "Zhou Hongjun said.

Third, the establishment and improvement of laws, regulations and institutional systems that guarantee the sound development of artificial intelligence. "Science and technological development have far exceeded the existing social regulations and ethical norms, and it is extremely urgent to establish a new network order." Li Cheng, director of the John Thornton Center of the Brookings Institution for American Think Tanks, believes that in order to make the cyber space harmonious and orderly, Society, think tanks and enterprises must establish a unified network law and rules.

Shen Xiangyang, vice president of Microsoft Global, said that Microsoft proposed "six principles" for "fairness, tolerance, transparency, accountability, security and security, privacy and confidentiality" to guide the development of artificial intelligence. Managing and directing the development of "responsible" products for artificial intelligence.

In addition, we must strengthen international exchanges to explore management response strategies. Bruce McConnor, global vice president of the East and West Institute of the United States, suggested that all parties in the world should prevent the risk of building a secure wall by building official and unofficial rules for dealing with cyber space. (Reporter Zhu Khan)


Source link