Charting the course towards Artificial General Intelligence that is truly open, accessible, and beneficial for everyone – prioritizing transparency and shared progress over closed systems.
{/* Content managed by JS */}
(Note: Progress percentage is illustrative, based on key historical and anticipated milestones towards the goal.)
Free Open AGI represents the goal of creating Artificial General Intelligence – capable of human-level understanding and learning across tasks – with the crucial distinctions that it must be freely available and fully open-source. This openness fosters trust, safety, and broad innovation.
Truly free means no barriers to access - no paywalls, no usage restrictions, no required registrations. Anyone with an internet connection can use it without cost.
Completely open means all model weights, architecture details, and training methodologies are publicly available. No black boxes - transparency enables trust and collaborative improvement.
Artificial General Intelligence represents systems that can understand or learn any intellectual task that a human being can, rather than being limited to narrow domains.
Alan Turing introduces the concept of a test for machine intelligence in his paper "Computing Machinery and Intelligence," laying the philosophical foundation for AI.
Yann LeCun successfully applies backpropagation to train convolutional neural networks for recognizing handwritten digits, proving neural networks' practical potential.
The release of the first version of the Creative Commons licenses enables easier sharing of research and code, accelerating collaborative AI development.
AlexNet's dramatic win in the ImageNet competition demonstrates the power of deep learning with GPUs, sparking the modern AI revolution.
Google open-sources TensorFlow, making powerful machine learning tools accessible to everyone and standardizing neural network development.
The "Attention Is All You Need" paper introduces transformers, which become the foundation for nearly all subsequent major AI advances, both open and closed.
OpenAI's GPT-3 demonstrates remarkable capabilities of large language models, increasing public awareness and indirectly fueling the drive for open alternatives, though initially kept proprietary.
The release of ChatGPT makes AI assistants mainstream, proving massive public demand for AGI-like systems and further highlighting the need for open, accessible versions.
Meta's restricted release and subsequent leak of Llama weights supercharges the open large language model movement, enabling widespread research and fine-tuning.
Qwen becomes the first major open model family to surpass GPT-4o on key benchmarks (like LiveBench), proving open weights can achieve state-of-the-art performance.
DeepSeek releases R1, matching OpenAI-o1 capabilities at 1/50th the cost - the DeepSeek moment for open AGI.
The culmination of these efforts: a true Artificial General Intelligence that is freely available to all humanity, with fully open weights, architecture, and unrestricted access.
This project, charting the road to Free Open AGI, is currently driven by a solo developer and advocate passionate about open access to transformative technology. While direct contribution pathways are evolving, you can follow progress, explore related resources on this site, and champion the cause for open AI.