Sometime last year (2023), when the ChatGPT buzz was at its peak, I tried to visit the OpenAI site to check it out. Unfortunately, I couldn't access it due to location restrictions. Frustrated, I decided to build my own chatbot. I figured I'd focus on local languages, starting with my mother tongue. This sparked my curiosity about how such platforms work. I began researching and soon realized I needed a large dataset to train a model. However, there's a scarcity of isiNdebele content online. Faced with this challenge and my constraints, I decided to take a different approach.
From the onset, I knew that, as impressive as current chatbots and AI tools are, they represent just the beginning of the AI race. These tools are currently at their worst and can only improve over time. The future direction is uncertain, and I believe no one truly knows where it’s headed. With my foot in the game, even on the smallest scale, I figured i could reimagine the future, create a vision of it and chart my own path, wherever it led... and who knows—I might just change the world.
When the time came to start coding, I decided to take a different approach: ignore what everyone else is doing, and build a model that uses reason and logic to draw conclusions from limited data, rather than relying on statistical probability. Within a few months, I had a working prototype. While some might have seen it as another attempt to replicate ChatGPT, my model was unique, custom-built from the ground up, deliberately avoiding existing pre-trained models.
The idea gained some initial attention, but its potential was limited due to the niche language and audience. I eventually moved on to other projects and started the "Building Africa’s Next Tech Unicorn" channel a few months ago. During my downtime, I shared insights and tutorials on tech and AI. As I delved deeper into research, I found myself exploring the concept of Artificial General Intelligence (AGI)- a type of AI that possesses the ability to understand, learn, and apply knowledge across a wide range of tasks, similar to human intelligence. Initially abstract, it soon hit me: while building my chatbot, I had inadvertently created a cognitive architecture.
In simple terms, a cognitive architecture is a blueprint or framework for designing systems that simulate human-like thought processes. It outlines the structures and mechanisms that an AI system uses to perceive, reason, learn, and interact with the world. Unlike traditional AI models that often rely on large datasets and statistical methods, cognitive architectures aim to replicate the cognitive functions of the human mind, allowing the system to think and reason more like a human. Mine is not the most advanced or sophisticated (it's definitely far from it), but understanding the mechanics helped crystallise my version of how the future of AI could look.
It’s been just over 2 weeks (17 days to be exact) since I started working on the new yet-to-be-named platform. I'm still on version 0.1, but the deeper I delve, the more the idea compounds into something extraordinary.
What I'm about to say may sound naive or even crazy, but I firmly believe AGI is possible within the next 2 years. It’s within reach. While I'm not claiming I'll be the first to achieve it—perhaps someone else will beat me to it—it's not as distant as you might think. The groundwork and essentials for making AGI a reality are already in place. I've only been working on this project for under a month, but I'm excited about the progress I've made so far, and more-so the prospects.
Over the next few weeks and months I'll be sharing more articles. These will include my definition of AGI as well as updates on my development process and progress. We live in exciting times, and I'm thrilled to be a part of it.
Comments