The research proposes the concept of neural computers, exploring emerging machine forms that unify computation, memory, and I/O.

ME News message, April 10 (UTC+8). Recently, a study proposed a cutting-edge concept called “Neural Computer” (NCs), a new machine form that unifies computation, memory, and input/output into an acquired runtime state. Its long-term goal is to achieve a “Fully Neural Computer” (CNC) with stable execution, explicit reprogramming, and persistent capability reuse. As an initial exploration, the study instantiated the neural computer as a video model, generating sequences of screen frames in command-line interface and graphical user interface setups based on instructions, pixels, and user actions. The results show that the acquired runtime can obtain early interface primitives, especially input/output alignment and short-term control, but issues such as regular reuse, controlled updates, and symbol stability still need to be addressed. The article lays out a development roadmap around these challenges and notes that if they can be overcome, a fully neural computer may establish a new computing paradigm beyond today’s intelligent agents, world models, and traditional computers. (Source: InFoQ)