Why is this sentence the theme of the 2023 Bund Conference
Walking in the technology exhibition area of the Bund Conference, green elements can be seen everywhere, including photos of ant forests, simulations of green computing links, and even a patch of farmland. Sitting at the Bund Conference forum, topics related to safety, privacy, and ethics are also hot topics.
These details actually point to the theme of this year's Bund Conference - "Technology, Creating a Sustainable Future".
The setting of this theme is of great practical significance. The contradiction between traditional industrial civilization and the supply of natural resources and the carrying capacity of the ecological environment has not yet been resolved, and the continuous development of technology seems to have begun to bring certain hidden dangers to the world. How to promote resource conservation and environmental protection through technology, how to make AI safe for our use, how to bridge the digital divide and bring a sustainable future, has been widely discussed at this conference.
Jointly Facing Global Issues
According to the Intergovernmental Panel on Climate Change (IPCC) of the United Nations, the annual climate investment required between 2020 and 2030 will be 3 to 6 times higher than existing investments. Meanwhile, the global biodiversity funding gap is expected to reach $600 to $825 billion by 2030. By 2050, there may be more.
"It is in this context that financial technology can provide a promising and even essential solution," said Rajan Mubarak, President of the World Conservation Union, at the Bund Conference. Financial technology not only helps fill financing gaps, but is crucial for efforts to mitigate and adapt to climate change and reverse biodiversity loss. By utilizing blockchain technology, financial technology can also promote transparency and security in carbon emissions and biodiversity credit trading markets, as well as traceability of sustainable products.
In fact, at the Bund Conference, relevant technologies have already been showcased - the green energy industry collaboration network jointly built by Ant Chain and its partners integrates blockchain and Internet of Things technology, connects information from various links such as power supply, power grid, load, and energy storage, and helps promote trust, cooperation, and value exchange between upstream and downstream industries.
Technology can promote sustainability, but people are beginning to realize that technology itself is also creating some problems.
Currently, humanity is in an era of data explosion. According to statistics, 90% of data in human history was generated in the past few years, and 50% was generated in just two years.
But every data call brings a certain amount of carbon emissions. Simply put, this is because there are many computers in the data center, and each operation generates electricity consumption, resulting in carbon emissions. According to Google's calculations, it takes 14.8 days to train the OpenAI GPT-3 model on 1000 Nvidia V100 GPUs. With a PUE of 1.1 in the data center, the total energy consumption reaches 1287 MWh, while the well-known ChatGPT needs to run on approximately 10000 GPUs.
Will the surge in energy consumption caused by the demand for computing power become an obstacle to sustainable development? People see green computing as a way to solve problems.
Building a green and low-carbon computing power industry chain requires close cooperation from upstream, midstream, and downstream of the computing power industry, including data center cloud vendors, data center hardware equipment vendors, and data center users to jointly build end-to-end green computing power, in order to achieve greater overall energy conservation and emission reduction effects.
At this year's Bund Conference, we can already see the possibility of ecological formation. From top-level design to different roles in the industry chain, various industries have shown a common determination and confidence in green computing.
At the policy level, He Feng, Deputy Director of Shanghai Communications Administration, mentioned that Shanghai is currently carrying out green computing related work around the "Computing Power Pujiang Action Plan" and suggested accelerating the research and development of key green computing technologies to improve the utilization rate of data center resources; Accelerate the standardization and definition of data centers; Deepen industrial collaboration and jointly build a new ecosystem of green computing.
At the hardware level, the industry has achieved significant results through measures such as architecture upgrades and reducing data center energy consumption. Duan Xiaodong, Vice President of China Mobile Research Institute, introduced that China Mobile has created a new type of intelligent computing center from five core aspects: "new Internet, new computing efficiency, new storage, new platform, and new energy conservation". "Facing the explosion of intelligent computing power, it is necessary to build a computing power network through open, collaborative, inclusive, and win-win methods, and promote the construction of a better green and energy-saving computing power foundation in the future intelligent computing era.".
According to the calculation of the Central Environmental Joint Certification Center, during the 2022 Double 11 period, Ant Group will save 1.538 million kilowatt hours of electricity through "green computing" technology, which can provide 150 million hours of use for a regular energy-saving lamp; Reducing emissions by 947 tons of carbon dioxide equivalent is approximately equivalent to the annual carbon sequestration of 79000 trees. Ant Group's Chief Technology Officer and President of the Platform Technology Business Group, He Zhengyu, stated that Ant Group has formed four core capabilities: low-carbon computing power selection, global resource scheduling optimization, workload optimization, and green monitoring and evaluation, which will provide a "measure" for green computing in the era of intelligent computing.
Security research is urgent and cannot be delayed
AI godfather Jeffrey Hinton once said in a media interview, "AI will be smart enough to threaten humanity, but I don't have a solution."
As the most representative technology of contemporary times, the hidden concern of AI has a long history. As early as 1942, Asimov proposed the "Three Laws of Robots" in his science fiction novel, which stipulated that robots must not harm humans or sit idly by while humans are harmed; Unless violating the first law, robots must obey human commands; Unless violating the first or second law, robots must protect themselves. At that time, there was only one robot in the world that could only be used for display.
By 2023, the security risks that once existed in science fiction novels seem to have come to mind. "AI security risks have shifted from future challenges to immediate threats that require joint efforts from government, industry, academia, and research," said Academician Li Yuhang, Chairman of the Cloud Security Alliance for Greater China.
In fact, just as the application scope of AI is diverse, security risks are not limited to privacy, but exist at key nodes throughout the entire lifecycle of AI, from data to algorithms to model applications.
"The challenges we face are very complex and multifaceted, and research on AI security is urgent." Wang Weiqiang, General Manager of the Machine Intelligence Department and Chief Scientist of the Security Laboratory of Ant Group, gave an example. These challenges include data poisoning, data leakage, model vulnerabilities caused by biased data, algorithm bias, and other issues. They also include the illusion of large models and deep forgery and malicious exploitation.
"When designing a car, the first thing to consider is the brakes. If the brakes are not working properly, even at high speeds, it can be dangerous." said Li Renhan, a professor at Shanghai Jiao Tong University. For artificial intelligence, the rule is the brake pads. Due to the anthropomorphism of artificial intelligence, the requirements for artificial intelligence also need to be similar to those for humans in order to coexist harmoniously.
Cheng Xueqi, Deputy Director of the China Institute of Science and Technology, also stated that in the future, the security and ethical issues of algorithms and artificial intelligence need to be promoted simultaneously through technological governance and social governance. "In terms of technological governance, it is automation and normalization, while social governance is standardization and systematization."
Such progress is underway. At the Bund Conference, the world's authoritative international industry organization "Cloud Security Alliance" in Greater China announced the establishment of an "AI Security Working Group", with more than 30 companies including China Telecom, Ant Group, Huawei, Baidu, Volcano Engine, Xi'an University of Electronic Science and Technology, and the National Financial Evaluation Center as the first wholesale units. The organization is committed to jointly solving the security challenges brought about by the rapid development of AI technology.
The AI Security Working Group will focus on the prominent security risk issues of artificial intelligence, and output white papers, industry knowledge graphs, group standards, system frameworks, solutions, etc. in the fields of endogenous security, derivative security, and assisted security of artificial intelligence, providing a clear and systematic AI security research framework for the entire industry.
In addition, the Bund Conference also released the industry's first platform ecological data security standard and launched a big model ethics co construction plan
More efforts are underway. As Li Yuhang said, development and security are the twin wheels of advancing the digital economy. Research and development should not be abandoned for the sake of choking, nor should security be ignored.