Oncomorphic computing: a bio-derived paradigm for decentralized adaptation across software and hardware

Abstract

Contemporary artificial intelligence excels in stable and resource-rich settings, but many of the environments in which autonomy and adaptive restructuring are most needed are defined by noise, damage, scarcity, and unpredictability. Spacecraft, distributed sensors, medical implants, digital cell twins, and robotic swarms must often continue functioning despite partial hardware failure and network fragmentation. Here, we present oncomorphic computing as a paradigm for designing resilient machine intelligence across software and hardware. We abstract selected survival principles seen in cancer biology, particularly decentralized behavior, phenotypic plasticity, heterogeneous specialization, resource recruitment, migration of function, and tissue-level adaptation, and repurpose them bidirectionally to complement standard approaches: using cancer dynamics in computing and engineering, and, conversely, bringing biological decentralization to the center of digital cancer research. In this way, we shift the approach from a highly organized and predictable ideal of machine intelligence to computational systems defined as groups of agents continuously negotiating access to memory, power and processing bandwidth through independent restructuring under scarcity. We discuss how this paradigm could address current computing bottlenecks and outline its implications for future hardware applications.