总部位于加利福尼亚州的初创公司 Beacon AI 周二宣布，其人工智能驱动的飞行员助手已在三架正在执行任务的美国空军 KC-46A“佩加萨斯”加油机上完成了五次飞行。
California-based startup Beacon AI announced Tuesday that its artificial intelligence-powered pilot assistant completed five flights aboard three operational US Air Force KC-46A Pegasus refueling aircraft.
该公司在一份声明中说，这套名为 Murdock 的便携系统在八月和九月的三周内，跨越三个机场累计记录了 20 个飞行小时，测试始于佛罗里达州埃格林空军基地的“翡翠旗”军事演习期间。
The portable system, known as Murdock, logged 20 flight hours across three airfields in three weeks in August and September, beginning during the Emerald Flag military exercises at Eglin Air Force Base in Florida, it said in a statement.
测试是在空军空中机动司令部通过 AFWERX 战术资金增加计划签订的合同下进行的，此前该公司已与美国特种作战司令部达成协议。
The tests were conducted under a contract with the Air Force's Air Mobility Command through the AFWERX Tactical Funding Increase program, following an earlier agreement with US Special Operations Command.
Murdock 从飞机航空电子设备收集数据，为机组人员提供实时辅助和作业信息，且不修改关键飞行系统，也不接管飞机控制权。
Murdock collects data from aircraft avionics to provide flight crews with real-time assistance and operational information without modifying flight-critical systems or taking control of the aircraft.
Beacon AI 将此次测试描述为将 AI 驱动辅助部署到实战军用飞机（而非专用测试平台）上迈出的重要一步。
Beacon AI described the tests as a significant step toward deploying AI-powered assistance aboard operational military aircraft rather than dedicated test platforms.
联合创始人兼首席执行官 Matt Cox 表示：“将这类系统引入驾驶舱的正确方式是循序渐进：交到真实机组在真实任务中使用，在其已有优势基础上构建，并随着作战经验和证据的支持扩展其能力。” 公司称，该系统旨在辅助飞行员，同时将所有作战决策和权力留给机组人员。
"The right way to bring a system like this into the cockpit is to do so progressively: put it in the hands of real crews on real missions, build on what it already does well, and expand its capabilities as operational experience and evidence support it," said co-founder and CEO Matt Cox The company said the system is designed to support pilots while leaving all operational decisions and authority with flight crews.
Beacon AI 计划利用实战测试中收集的数据和反馈，将 Murdock 的能力扩展到飞行全生命周期，涵盖排班、飞行作业到飞后复盘等环节。
Beacon AI plans to expand Murdock's capabilities in the flight lifecycle, from scheduling and flight operations to post-flight reviews, using data and feedback gathered during operational testing.