The advance of synthetic intelligence applied sciences reminiscent of giant language fashions (LLMs) and autonomous brokers is spreading pretend information that’s extra deadly and complicated. Voices are rising inside and out of doors the science and know-how group that Korea should additionally safe sovereign AI know-how to reply, alongside cross-national coverage cooperation.
Cha Meeyoung (pictured), director of the analysis group on the Max Planck Institute for Security and Privacy in Germany, who delivered the keynote speech at this 12 months’s Korean scientists and engineers convention, expressed the identical view.
Cha, who met with Korean media on the Korea Science and Technology Center in Gangnam-gu, Seoul, on Monday, mentioned, “Policy and science must respond together to prevent cutting-edge AI from manipulating public opinion or exerting malicious influence on social decision-making.”
Cha, a professor on the School of Electrical Engineering on the Korea Advanced Institute of Science and Technology (KAIST), was appointed in 2024 as the primary Korean analysis group chief at Germany’s Max Planck Institute, usually referred to as a “Nobel Prize academy,” in recognition of her work analyzing social issues reminiscent of pretend information throughout the COVID-19 interval, poverty, and environmental air pollution utilizing huge data-based AI. Now main the “Data Science for Humanity” analysis group, she additionally delivered a keynote speech on the identical subject on the occasion that day.
Cha recognized that as a result of just lately quickly superior AI has modified the pretend information ecosystem itself, technological and coverage approaches should additionally change. This is as a result of, in a state of affairs the place AI disguised as people creates 1000’s of faux accounts and operates in an organized method, the after-the-fact response methodology wherein human “fact-checkers” individually decide whether or not info is fake is now not efficient.
Cha mentioned, “In the past, fake news was generated through networks between ‘people,’ such as on social media, whereas now AI intervenes directly, making it difficult to track and analyze the spread the way we used to.” She additionally confused, “The algorithm through which biased or false information is injected while each individual converses with AI is also difficult to identify at the researcher level,” including, “Going forward, ‘mechanistic interpretability’ research that looks inside the AI’s black box will become important to solve this.”
Assessments additionally adopted that Korea’s regulatory degree is inadequate in comparison with the pace and affect of faux information unfold. As the modification to the Act on Promotion of Information and Communications Network Utilization, referred to as the “Act to Eradicate False and Manipulated Information,” took impact that day, Cha mentioned, “Punishment for spreading false information and defamation is still at a slap-on-the-wrist level,” and “more responsibility must be placed on all responsible stakeholders, including platforms.” In explicit, she believed that coverage regulation of world platforms and AI growth corporations should be strengthened. She mentioned, “There is a need to require them to provide at least minimum benchmarks (to public institutions and researchers) so that it is possible to understand how information is exchanged between AI and individuals.”
Cha emphasised the significance of sovereign AI on this course of. Cha mentioned, “To respond to the malicious use of AI, it is essential to provide researchers with access rights to cutting-edge AI models,” and “Taking a lesson from the recent incident in which Anthropic blocked access to ‘Mythos 5’ and ‘Fable 5,’ Korea must also speed up AI model development at the national level and secure core technologies such as sovereign AI.”
Cha then mentioned, “Compared to Europe, where innovation is difficult to produce culturally and administratively, I think Korea’s aggressive and fast-paced research culture is a particularly advantageous structure for developing sovereign AI.” On the opposite hand, she additionally famous that “a long-term and stable research support system is something Korea can learn from Europe.” She mentioned, “The greatest strength of the Max Planck Institute is long-horizon research funding. Because there is no worry about applying for a budget every year or having it cut, researchers deeply consider what research topics will be meaningful even more than a decade from now.”
Finally, she confused, “Cutting-edge AI, which is currently developing with a focus on profit models and short-term rewards, is deepening polarization,” and “the more this is the case, the more we must ask fundamental questions about how AI can contribute to the happiness of ‘all humanity’ rather than ‘a few.'” She suggested, “We must design research and policy on AI with long-term rewards in mind, ones whose results can emerge even for future generations.”