Bootstrapping Neural Electronics from Lunar Resources for In-Situ Artificial Intelligence Applications

Artificial intelligence and robotics are leveraging technologies for lunar exploration. However, future lunar surface exploration will require exploitation of in-situ resources to reduce (and ultimately eliminate) the costs imposed by the transport of materiel from Earth. Solid-state manufacturing of electronics assets from lunar resources to eliminate its supply from Earth is impractical. We propose the in-situ manufacture of vacuum tube-based computational electronics which requires only a handful of materials that are available on the Moon. To offset the problem of exponential growth in physical footprint in CPU-based electronics, we propose the implementation of analogue neural network hardware which has Turing machine capabilities. We suggest that the artificial intelligence requirements for lunar industrialisation ecology can be demonstrated in principle by analogue neural networks controlling a small rover. We pay particular attention to online learning circuitry as the key to adaptability of analogue neural networks.