埃默里大学和佐治亚理工学院的研究人员开发了名为CapuchinAI的人工智能系统，用于在自然环境中自动开展灵长类动物认知研究[1]。该系统集成了面部识别和实时触屏测试功能，在哥斯达黎加塔博加森林保护区的现场测试中，以97%的准确度识别了卷尾猴[1]。试点阶段共有16只野生卷尾猴与该系统互动[1]，它们成功学会通过触屏操作获取食物奖励[1]。
该系统采用Raspberry Pi微型电脑运行，单次续航时间为8小时[1]。这一研究为在野生环境中以可扩展的方法研究灵长类动物的认知能力提供了新的可能[1]。论文已发表在《美国灵长类学杂志》上[1]，由费德里科·桑切斯·瓦尔加斯担任第一作者，马塞拉·贝尼特斯担任高级作者[1]。该项目获得了埃默里大学AI.Humanity项目种子基金的资助[1]。
Researchers from Emory University and Georgia Institute of Technology have developed CapuchinAI, an artificial intelligence system designed to conduct automated cognitive research on primates in their natural habitat [1]. The system integrates facial recognition and real-time touchscreen testing capabilities [1].
Field testing at Taboga Forest Reserve in Costa Rica demonstrated the system's effectiveness, with facial recognition achieving 97% accuracy in identifying capuchin monkeys [1]. During the pilot phase, 16 wild capuchins successfully learned to interact with the touchscreen interface to obtain food rewards [1]. The system operates on a Raspberry Pi microcomputer with an eight-hour battery life [1].
The research, led by first author Federico Sánchez Vargas with Marcela Benítez as senior author, was published in the American Journal of Primatology [1]. The project received seed funding from Emory University's AI.Humanity initiative [1].