Lately, I've been thinking a lot about AGI and where it might take us, but honestly, that feels like a puzzle we won't solve anytime soon. Instead, let's take a different approach: Imagine it's 2074, and we're looking back on the 2020s. What would we say were the major milestones in AI and computing? What applications and research grabbed our attention? What challenges were we struggling with? And most importantly, how are the things we’re doing today laying the groundwork for future breakthroughs?

Let's drawing some inspiration from how past decades shaped the present.

A Look Back in Time: Milestones in Computing

To understand the significance of the 2020s, let’s briefly review how computing evolved over the decades:

• 1950s: The birth of practical computing, as hardware dominated research and development. Early machines like the UNIVAC I were primarily number crunchers, and Fortran became the first high-level language. During this period, computing was in its infancy—similar to how AI models might be viewed today.
• 1960s: Computers became more accessible with the creation of mainframes (like IBM System/360) and the development of early networks (ARPANET). Artificial intelligence emerged, though it was mostly theoretical and hardware-driven.
• 1970s: Microprocessors and personal computers such as the Intel 4004 and Apple I shifted computing to individuals. Operating systems like UNIX and early programming languages, such as C, provided new abstractions over hardware.
• 1980s: The personal computer revolution continued, and the rise of user-friendly interfaces (IBM PC, Apple Macintosh) and object-oriented programming (C++) transformed software development. Graphical user interfaces made computers accessible to non-experts.
• 1990s: The internet revolution exploded. We saw the birth of the World Wide Web, the release of Windows 95, and the rise of open-source software like Linux. Java and Python programming languages simplified coding. The internet connected everyone, much like how we expect future AI networks to connect agents.
• 2000s: Mobile computing and cloud infrastructure became pivotal. With the rise of smartphones and services like AWS, computing went global. Data exploded, setting the stage for AI’s next leap.
• 2010s: Artificial intelligence began to shine with advancements in machine learning and deep learning. AlphaGo’s victory was a symbolic moment for AI’s potential. The groundwork for LLMs (Large Language Models) was laid as data and compute power converged.

2020s: The Era of AI as a General-Purpose Intelligence Platform

If we think about AI models like LLMs and AGI as general-purpose computing platforms—similar to hardware processors in the 1950s—are we in the early days of AI development? Just as the first computers changed the world, today's AI models are poised to do the same, only much faster. Given the speed of today’s information exchange, AI advancements will likely far outpace the historical development of computing.

From Hardware (Models) to Software (AI Applications)

Much like the 1970s and 1980s, when attention shifted from hardware development to the software that runs on it, we’re now seeing AI focus shifting from building bigger and better models to deploying practical AI applications. In the same way that early programming languages and operating systems abstracted hardware complexities, we’re starting to see platforms emerge that allow developers to apply AI without needing deep knowledge of model architecture.

• Future question: Could this shift allow AI to become as commonplace and versatile as personal computers were in the 1980s? Will every business, organization, and individual rely on general-purpose AI to handle everyday tasks?

Personal AI: The Next Revolution

Just as personal computers became mainstream in the 1980s, we’re on the brink of AI becoming personalized. Personal AIs will evolve beyond basic assistants like today’s Siri or Alexa into fully customizable entities that manage multiple aspects of daily life—task management, communication, learning, and even decision-making. This shift could parallel the revolution that brought PCs from niche tools to household essentials.

• Future question: Will personal AIs become so powerful and integrated into daily life that they become indispensable? Could they manage tasks, offer advice, and anticipate needs in ways we can barely imagine today?

Multi-Tasking AI Systems: An AI Operating System

The creation of operating systems in the 1960s and 1970s made it possible to run multiple tasks on the same hardware. Now, AI is evolving in a similar way. We’re designing frameworks that allow a single AI model to handle diverse tasks efficiently, abstracting away the complexity of managing different models for different applications. These AI systems could become as essential as operating systems were in early computing, handling everything from personal tasks to complex organizational workflows.

• Future question: Could AI evolve into a new kind of OS, where different AI agents or functions are abstracted and managed by a higher system? Will AI systems provide seamless multitasking and resource allocation similar to modern operating systems?

High-Level Abstractions (Prompt Engineering and AI Languages)

In the same way that early programming languages like COBOL and C abstracted away the complexity of writing machine-level code, we’re seeing a rise in high-level prompting techniques. These allow developers to communicate with AI systems without needing to understand their underlying architectures. As more sophisticated abstractions are built, interacting with AI might become as straightforward as programming was in the 1990s.

• Future question: Will AI languages develop into specialized, high-level tools that make working with AI as easy as writing code? Could prompting and AI programming become core skills for developers in the future?

AI-Managed Knowledge Bases: The New Database

As AI begins to manage and curate vast repositories of information, we might see the next generation of knowledge management systems emerging. In the same way that databases revolutionized data storage in the 1970s, AI-managed knowledge bases could make it easier to query, organize, and extract insights from vast datasets without manual intervention.

• Future question: Could AI become the backbone of future databases, managing and retrieving information autonomously? What happens when AI learns to optimize, clean, and generate knowledge without human input?

Multi-Agent AI Networks: AI’s Version of the Internet

Just as ARPANET connected early computers, multi-agent systems could connect AI models to work collaboratively on complex tasks. Imagine a network of AIs communicating, sharing tasks, and solving problems together. The potential for AI to coordinate across networks could be as revolutionary as the birth of the internet in the 1990s.

• Future question: Will AI networks evolve to the point where AIs interact seamlessly, creating a web of collaborative systems? Could multi-agent AI networks usher in a new level of problem-solving and innovation?

Complexity in AI Applications: The Next Challenge for system

Much like the growing complexity of software in the 1980s and 1990s demanded new programming practices, AI’s increasing complexity presents a similar challenge. How do we build systems that can scale and manage increasingly intricate AI applications? Will the tools we use to build, debug, and scale these systems keep up with AI’s growing demands?

• Future question: As AI applications become more complex, will we see new abstractions, tools, and frameworks emerge to manage them? How will we deal with the scaling challenges that come with such sophisticated AI systems?

2020s: Shaping the Future of AI and Computing

The 2020s will likely be remembered as a decade where AI transitioned from theoretical advancements to real-world, impactful applications—just as earlier decades saw computing evolve from experimental hardware to a global infrastructure. Our current focus on AI applications, multi-agent systems, personal AI, and AI-managed knowledge systems is setting the stage for future breakthroughs.

As we push the boundaries of AI and its potential, it’s clear that our choices today will profoundly shape the next fifty years. While we don’t have all the answers, asking the right questions helps guide us toward a future where AI isn’t just a tool—it’s a platform for innovation across every industry and aspect of daily life.

So, where do you think we’re headed?

For more interesting AI experiments and insights, please visit my AI experiment and throughts website https://yunwei37.github.io/My-AI-experiment/ and github repo: https://github.com/yunwei37/My-AI-experiment