Near’s Goal: Building an AI Model Bigger Than Ever
Near Protocol announced its goal to create a groundbreaking open-source AI model with 1.4 trillion parameters—over three times the size of Meta’s Llama model. This ambitious project, shared at Near’s Redacted conference in Bangkok, aims to set a new standard in AI with the help of a global, crowdsourced community.
A Crowdsourced Approach to AI Development
The project will bring together thousands of contributors from around the world. They’ll start by training smaller models, and as the project scales, top contributors will move on to more advanced models. Participants will be rewarded, and Near will use encrypted technology to protect privacy and ensure fair compensation as the AI evolves.
Funding and Self-Sustaining Support for the Project
With a cost of around $160 million, Near plans to fund this massive project through token sales. Near’s co-founder, Illia Polosukhin, explained that tokenholders will earn from the model’s usage, creating a self-sustaining loop where profits can support future models and expand the project.
Tackling Technical Challenges for Decentralized AI
Building such a large model is no easy task and will require thousands of GPUs working together. Co-founder Alex Skidanov acknowledged that achieving this in a decentralized way is challenging, given current technology limitations. However, new research may help Near overcome these technical barriers and make decentralized AI a reality.
The Importance of Decentralized AI for Privacy and Freedom
Polosukhin highlighted that decentralizing AI is critical to prevent any one company from gaining control over technology. At the conference, privacy advocate Edward Snowden reinforced this, warning that a centralized AI could lead to mass surveillance. He argued that Web3 and decentralized AI are essential to protect freedom in the digital world.