This research program aims to develop an initial approach to the design of bio-inspired neuromorphic memristor circuits with complex physiological goal-setting mechanisms, with the potential to initiate groundbreaking innovations within a heterogeneous computing system. Current leading algorithmic data centers, characterized by high demand and energy consumption, are unable to independently handle minor new or complex tasks in a fixed computing environment. To enable a neuro-memristive computer to independently identify and solve unpredictable problems—rather than strictly fixed and predefined tasks that fall outside the scope of a traditional computer system—it is necessary to incorporate fundamental neurobiological properties of goal-directed behavior into the architectural foundation of AI. Since modern digital, quantum, optical, and software computing has limited logic and an energy-consuming system, implementing adaptive-autonomous intelligence will require a different approach. Building a brain-like computer based on new molecular solid-state two-terminal nanodevices into neural computing units will require combining the experience and efforts of leading national organizations and laboratories in several countries. The new AI accelerator will have not only analog signal processing but also simpler functionality. The most important thing is that the AI ​​bioprocessor will not require large databases and fitting models to results as in digital machine algorithms, since the materials and physics of the new multifunctional elements will be hyper-adaptively self-reconfigurable.