On November 21, during the “Sub-forum on Responsible Development and Application of Artificial Intelligence” at the Wuzhen Summit of the 2024 World Internet Conference, Li Junkui, Vice President of Ant Group, pointed out in his speech that a “traffic light” system should be established for the healthy development of AI: AI applications that pose serious impacts on human values and social stability must be defended against and resisted; those with potential risks require thorough and in-depth analysis and research to establish a sound ethical and safety framework; and for AI applications that benefit humans, a green light should be given.
Li Junkui compared AI to a high-speed train, “It brings unprecedented opportunities, such as promoting economic development, social progress, and supporting public welfare causes. However, just as a train needs tracks and signals to guide it, the development of AI cannot run wildly without constraints. We must set up traffic lights for it to ensure its healthy progress.”
The red light represents risk control. “We must be vigilant against the malicious use of AI technology, such as creating false information, launching cyberattacks, or exacerbating social inequality. Deepfake technology is an example; if not restricted, it could destroy social trust systems and cause chaos.” Li Junkui emphasized that we must oppose such applications that pose serious impacts on human values and social stability, making the red light an uncrossable boundary.
The yellow light signifies managing AI systems beyond technology, treating AI fields with potential risks with caution. “For instance, although the field of autonomous driving has vast prospects, it currently faces many complex ethical issues. For example, when an autonomous vehicle faces an unavoidable collision, should it prioritize protecting the passengers or pedestrians?” Li Junkui believes that in such situations, thorough and in-depth research and discussion are necessary to establish a sound ethical and safety framework, allowing progress only when risks are controllable.
Li Junkui introduced that when Ant Group established the AI Safety and Ethics Committee at the beginning of last year, it aimed to use “good governance” to address “good intelligence.” Any AI application from the group undergoes evaluations, including technological ethics, before its release, truly using systems to guide AI-driven businesses to align with “people-oriented, technology for good.”
The green light is for giving a pass signal to AI applications that benefit humans. “Safety is the means, and the ultimate goal is to promote the healthy development of AI, widely apply it in industries, facilitate people’s lives, and bridge the technological divide caused by AI.”
Li Junkui stated that the process of setting up “traffic lights” for AI development is also the process of building responsible AI. The establishment of responsible AI faces three long-term trends: first, from “accuracy” to “human touch”; besides ensuring AI’s reliability and accuracy, it should also genuinely better serve humans. Second, from “passive prevention and control of explicit risks” to “active prevention and control of implicit risks.” Third, from “dialogue interaction” to “effective collaboration.”
Li Junkui believes that building responsible AI requires joint efforts from various sectors. For governments, reasonable regulations and policies should be formulated, with clear red lines and regulatory mechanisms; for social organizations and research institutions, research on AI ethics and risks should be strengthened to provide scientific bases for decision-making; for technology companies, moral and ethical standards should be consciously followed, integrating social responsibility into technological research and development; for the public, they should actively pay attention to and participate in discussions on AI development, jointly building responsible AI. “In summary, setting up good traffic lights and establishing responsible AI is a systematic and open proposition that is both urgent and requires a long-term perspective.”