В США провели саммит по достижению превосходства в принятии решений с помощью ИИ

"The history of failure in war can almost be summed up in two words: too late." General Douglas MacArthur's warning underscored the urgency and importance of the first Decision Superiority Summit, hosted this summer by Lawrence Livermore National Laboratory (LLNL), where Department of War (DOW) decision makers, scientists, engineers and analysts explored how computer modeling and simulation, artificial intelligence, machine learning and high-performance computing can enable faster, better-informed decisions in the increasingly competitive technological arena of near-peer conflict. "The goal of this summit was to engage stakeholders and subject-matter experts from across the national security spectrum and kick off an ongoing conversation about meeting the urgent need for AI-enabled military operations," said Michael Schneider, associate program lead for LLNL's Decision Superiority Laboratory. "We want to understand mission needs, gaps and challenges and start to envision what it will take to integrate existing and advanced capabilities to guarantee U.S. strategic advantage." At the summit, decision superiority was defined as using data and computational resources to create strategic leverage in a highly competitive and rapidly evolving technological landscape. Speakers also described it as the ability to improve decision making across domains by combining decision science, multi-agent optimization and data engineering to provide warfighters with tools that can support real-time decision making in complex conflict situations. "Decision superiority means developing the tools that will allow the U.S. military to make faster, better decisions in complex competition, crisis and conflict scenarios in near-real-time," said Schneider. The summit opened with a welcome from Pat Falcone, LLNL's deputy director for science and technology, followed by sessions on mission needs, data infrastructure and AI-enabled operational support. A recurring theme at the summit was the limits and unwieldy demands of current large language models, data availability, the need for compressed data training time and the potential for learning systems that can adapt to deliver strategic advantage as conditions change. Presenters emphasized the need to move beyond static tools and toward capabilities that can quickly and accurately respond to new information in real time. The summit made clear that decision superiority depends not just on algorithms, but on resilient infrastructure, agile logistics, robust data management and institutional, human-in-the-loop know-how. A highlight of the summit was George Chapline's presentation on quantum mechanics and Bayesian filters. Chapline, a distinguished theoretical physicist at LLNL, connected quantum theory, Bayesian statistical probability and mammalian neurology to highlight the summit's broad technical scope and approach to decision superiority. Summit attendees also toured El Capitan, the world's second most powerful supercomputer and a critical tool in enabling large-scale analysis and rapid modeling for real-world defense applications. The summit concluded with a focus on future work to advance decision support capabilities for DOW and national security missions. "We're at an exciting time for technology solutions for decision support as the data, infrastructure and algorithmic capabilities can come together to create the kinds of advantages that were only imagined for decades. There is a very strong demand signal from the national security community, and I'm proud of LLNL's leadership in 'creating the future' to rapidly meet those demands," said Schneider.