[AI-as-OS] Artificial Intelligence as Operating System

Artificial intelligence (AI) methods have been penetrating new areas of application for years. AI systems are increasingly able to take on complex tasks with increased autonomy - for example through agentic approaches that operate an AI agent on a computer to carry out any task automatically or even autonomously, sometimes without explicit human instructions, but based on identified intentions (see, among other things, the open source software OpenClaw, which can take care of calendar maintenance or file management, for example). The extent to which AI algorithms and models are now being integrated into everyday applications is also reflected in the fact that end devices are increasingly operable without any visible applications and can only be operated via voice control, which is fundamentally rethinking the way we interact with them (such as Deutsche Telekom's AI Phone). The central control units of such end devices - the operating systems themselves - are also increasingly being expanded to include new AI functionalities in order to continuously increase system efficiency. This includes initial theoretical and partly practical approaches to using the latest AI technologies for file system or storage management. These developments mark a fundamental paradigm shift: AI is no longer seen as just an add-on, but increasingly as a native architectural component of the operating system itself. The AI-as-OS program wants to investigate how artificial intelligence is no longer understood as a separately encapsulated functionality that can only be executed by the operating system, but as the operating system itself - Artificial Intelligence as Operating System. Behind this is the vision of an AI that not only supports individual applications, but also acts as a central control system of the entire computer ecosystem and manages system resources (including CPU, memory, system security) in a data-based, adaptive, and context-aware manner. The aim of the program is to make fundamental contributions to the development of AI-based operating systems as the control unit of a computer system. By comprehensively integrating AI technologies into deep operating system layers, such a system is intended to manage system resources, enable context-aware operational decisions and increase system security through adaptive threat detection and defense. Since the underlying technology is largely still in a speculative stage, the spectrum of development work ranges from the general architecture of such an operating system to the integration of AI-based core mechanisms to new considerations on the security and stability requirements for such an operating system, depending on the target platform (e.g. desktop, data center, edge device) and application environment. The program's research focuses include: • AI-centered operating system architectures: Concepts for the basic structure of an operating system in which AI is not seen as an additional function, but as an integral part of the architecture. • AI-native core mechanisms: Integration of AI-based processes into core operating system functions such as resource management, process control and memory organization. • Control, security and stability: Requirements for traceability, controllability and robustness of an AI-supported operating system as well as approaches for adaptive threat detection and defense. From today's perspective, it remains unclear whether and to what extent the vision of an AI-native operating system can be technically realized. This is precisely where the program gains knowledge: AI-as-OS works on a still largely speculative future field at the interface of AI research and system software, the results of which can set the fundamental course for the operating systems of tomorrow.