21 novembre 2018 | International, C4ISR
To maintain tech edge, US seeks export controls on AI
By: Kelsey D. Atherton In just two words, the phrase “artificial intelligence” captures a deep techno-utopian promise, the notion that through craftsmanship humans can create learning and thinking machines outside the processes of organic life. AI is typically the realm of technologists and science fiction writers. Now it is also in the world of export controls prohibitions and restrictions on technologies as overseen by the Department of Commerce. In a proposed rule announced Nov. 19, the Bureau of Industry and Security wants to set out guidelines establishing “criteria for identifying emerging technologies that are essential to U.S. national security.” The stated goals of such controls are tied to both security and protectionism for existing American industry, especially the science, technology, engineering and manufacturing sectors. The proposed rules encompass 14 technologies, covering brain-computer interfaces to advanced surveillance technology. Nestled in that list of technologies is “artificial intelligence (AI) and machine learning technology,” which is further broken into 11 related tools. Here is a list of all the kinds of AI that the new rules seek to put under Commerce export controls: Neural networks and deep learning (e.g., brain modelling, time series prediction, classification) Evolution and genetic computation (e.g., genetic algorithms, genetic programming) Reinforcement learning Computer vision (e.g., object recognition, image understanding) Expert systems (e.g., decision support systems, teaching systems) Speech and audio processing (e.g., speech recognition and production) Natural language processing ( e.g., machine translation) Planning (e.g., scheduling, game playing) Audio and video manipulation technologies (e.g., voice cloning, deepfakes) AI cloud technologies AI chipsets Several of these are as much mathematical concepts, or processes, as they are distinct, controllable technologies. Others, like AI cloud technologies, suggest always-online servers, which by the very nature of the internet, are difficult to control within borders. Tackling an entire technological field, especially one with as low a barrier to entry as coding, is a tricky proposition, even in the instances where the technology is clearly defined. Why might the White House go through all this trouble? “These revisions could compose an important element of a strategy of targeted countermeasures against the near-term threat posed by China's tactics for tech transfer and the long-term challenge of China's emergence as a powerhouse in innovation,” said Elsa B. Kania, adjunct fellow at the Center for New American Security. “However, the revision of this traditional mechanism for today's challenges is inherently challenging, particularly when development is driven by commercial technologies.” Unlike, say, controlling the components and designs of missiles in the Cold War, many of the technologies covered under these proposed rules have both commercial and military applications. We need not look abroad to find this. Project Maven, the tool Google created to process images collected from drones, was built on top of an open-source library. Identifying objects in images is hardly a military-specific task. Should companies within the United States be restricted in how they create, sell and share those same tools with researchers and commercial companies outside American borders? “China's national strategy of military-civil fusion, which seeks to create and leverage synergies among defense, academic, and commercial technological developments in dual-use technologies, increases the ambiguity and uncertainty of tech transfer and collaboration,” Kania said. “That is, the boundaries between defense and commercial technologies can become quite blurred as a result of the nature of these technologies and the Chinese government's strategy for their integrated development.” Putting in place controls to hinder the free flow of AI between American companies and businesses abroad may mitigate that risk somewhat, but countries set on acquiring the tools can pursue research by other means, including technology transfers, espionage, theft through hacking, or even straightforward investment and acquisition. Staying ahead in artificial intelligence likely cannot be done through commerce restrictions alone. “The U.S. must recognize that such controls may slow and hinder China's advances in these emerging technologies, but China's emergence as a powerhouse and would-be superpower in such emerging technologies will remain a critical long-term challenge,” Kania said. “We must not only pursue such defensive countermeasures, but also undertake a more offensive approach to ensuring future American competitiveness through investing in our own innovation ecosystem.” https://www.c4isrnet.com/it-networks/2018/11/20/to-maintain-tech-edge-us-seeks-export-controls-on-ai