Operating Systems with Brains
The concept of what an “operating system” can be is undergoing a radical transformation. We’re on the edge of seeing AI not just as a tool or assistant but as the very backbone of our computing environments. Let’s explore how the near-future AI’s are going to to redefine the traditional operating system.
[We are here not talking about Operating Systems in the sense of the system managing your computer, but the system managing your work and life]
Advancements in AI Agency
The current capabilities of LLMs such as GPT-4, Claude-3 and Gemini 1.5 are already quite impressive—they can reason, plan, and even interact with external tools via APIs. However, the full potential of AI agency remains largely untapped. To achieve a more autonomous state, significant enhancements are necessary:
- Refined Reasoning and Goal-oriented Processing: Future iterations like GPT-5 (or whatever provider) need to dramatically improve how they reason through complex problems and pursue goals more iteratively and persistently.
- Robust Memory Enhancements: As the complexity and duration of tasks increase, so does the need for advanced memory capabilities. Maintaining context over longer interactions without losing focus is crucial for the next wave of AI systems. Gemini 1.5 Pro does this well with 1 Mio tokens but that is not the end.
- Expanded Tool Integration: Broadening the range of accessible tools, APIs and systems will empower AI to manage and execute tasks across diverse environments, making it a more integral part of the technological infrastructure.
A Small Stepping Stone: OpenAI Assistants
OpenAI Assistants have laid the groundwork for what could eventually be fully autonomous AI systems, but it is just one single step while LangChain, AutoGen, CrewAI and other AI Agent Frameworks are today a few steps ahead. These AI Agent Systems already demonstrate the ability to handle tasks across varied environments, suggesting a roadmap towards AI systems that are not just reactive but proactive, capable of real-time adaptation, optimization and working autonomous.
Beyond Assistance: Proactive AI Engagement
The evolution from static models to dynamic, interactive systems hints at a future where AI could directly interact with both digital and physical environments more independently. This may transition AI from a background technology to a central operating system capable of managing complex infrastructures, potentially revolutionizing how we interact with our digital and physical surroundings.
OpenAI’s Vision: A Broader Scope
The trajectory of OpenAI’s developments indicates a broader vision beyond simple LLMs. The company’s ambition to integrate GPTs into a variety of applications and its disruptive impact on companies that rely on basic AI applications underscore its commitment to leading AI innovation. Their mission is to make AI a part of our daily life, not just providing Large Language Models. [And this is just my personal outside view of things, but backed by Andrej Karpathy’s X post on LLM OS]
This shift not only promises enhancements in how AI interacts with the world but also redefines the entire ecosystem of technology, positioning AI at the heart of digital transformation. As we move closer to realizing these advancements, the anticipation of AI’s capabilities continues to grow, suggesting a future where AI is not just part of the system—it is the system.