Human nonverbal communication tools are very ambiguous and difficult to transfer to machines or artificial intelligence (AI). If the AI understands the mental state behind a user’s decision, it can learn more appropriate decisions even in unclear situations. We introduce the Brain–AI Closed-Loop System (BACLoS), a wireless interaction platform that enables human brain wave analysis and transfers results to AI to verify and enhance AI decision-making. We developed a wireless earbud-like electroencephalography (EEG) measurement device, combined with tattoo-like electrodes and connectors, which enables continuous recording of high-quality EEG signals, especially the error-related potential (ErrP). The sensor measures the ErrP signals, which reflects the human cognitive consequences of an unpredicted machine response. The AI corrects or reinforces decisions depending on the presence or absence of the ErrP signals, which is determined by deep learning classification of the received EEG data. We demonstrate the BACLoS for AI-based machines, including autonomous driving vehicles, maze solvers, and assistant interfaces.

人类的非语言交流工具非常模棱两可，难以转移到机器或人工智能 (AI) 上。如果人工智能了解用户决策背后的心理状态，即使在不清楚的情况下，它也可以学习更合适的决策。我们引入了大脑-人工智能闭环系统 (BACLoS)，这是一个无线交互平台，可以进行人类脑电波分析并将结果传输到人工智能以验证和增强人工智能决策。我们开发了一种无线耳塞式脑电图 (EEG) 测量设备，结合了类似纹身的电极和连接器，可以连续记录高质量的脑电图信号，尤其是误差相关电位 (ErrP)。传感器测量 ErrP 信号，这反映了不可预测的机器响应的人类认知后果。人工智能根据 ErrP 信号的存在与否来纠正或加强决策，这取决于接收到的 EEG 数据的深度学习分类。我们展示了用于基于 AI 的机器的 BACLoS，包括自动驾驶车辆、迷宫求解器和辅助界面。