Near Protocol Announces Plans For World’s Largest Open-source AI Model

This initiative, unveiled by NEAR Co-Founder and Foundation CEO Illia Polosukhinat at the Redacted conference in Bangkok, Thailand, aims to surpass Meta’s Llama model by a factor of 3.5, setting a new benchmark in AI capabilities.

The Vision Behind the 1.4 Trillion Parameter Model

Near Protocol’s ambitious project is not just about size; it’s about fostering a collaborative AI ecosystem. The development process will be crowdsourced, inviting thousands of contributors to participate through the newly established Near AI Research hub. Starting with a 500 million parameter model, the project will progressively scale up, with contributors advancing based on their performance and contributions. This tiered approach ensures that only the most effective participants work on the increasingly complex models, promoting a meritocratic development environment.

Funding and Monetization Strategies

Training a model of this magnitude is a costly endeavor, with estimates around $160 million. Near Protocol plans to fund this through token sales, offering a unique investment opportunity within the crypto community. Token holders are expected to receive returns from the model’s usage, creating a sustainable financial loop that supports ongoing development and innovation. This strategy not only secures the necessary capital but also aligns the interests of investors with the project’s success.

Leveraging Expertise in AI and Blockchain

Near Protocol’s leadership brings a wealth of experience to this initiative. Co-founder Illia Polosukhin was a contributor to the seminal transformer research paper “Attention Is All You Need” that laid the groundwork for models like ChatGPT. His deep understanding of AI architectures is complemented by co-founder Alex Skidanov’s tenure at OpenAI during the development of ChatGPT. Their combined expertise positions Near Protocol uniquely at the intersection of AI and blockchain technologies

Source: Wikipedia.

Addressing Technical Challenges

Developing a model of this scale presents significant technical hurdles, particularly in terms of computational resources. Training such a large model traditionally requires centralized clusters of tens of thousands of GPUs, which can be both logistically challenging and costly. Near Protocol is exploring decentralized training methods, inspired by emerging research from organizations like DeepMind, to distribute the computational load across a network. This approach could democratize access to AI development and reduce reliance on centralized infrastructure.

Decentralization and Privacy Considerations

A core tenet of Near Protocol’s philosophy is decentralization, which extends to this AI initiative. By utilizing encrypted Trusted Execution Environments, the project aims to preserve user privacy while ensuring secure and verifiable contributions. This method not only protects sensitive data but also incentivizes continuous updates and improvements to the model, fostering a dynamic and responsive AI ecosystem.

Implications for the AI and Blockchain Sectors

Near Protocol’s endeavor represents a significant convergence of AI and blockchain technologies. By creating an open-source model of this scale, the project could democratize access to advanced AI capabilities, enabling a broader range of applications and innovations. Moreover, the integration of blockchain for funding and governance introduces a novel paradigm for AI development, potentially setting a precedent for future projects in the tech industry.

Near Protocol’s plan to build a 1.4 trillion parameter open-source AI model is a testament to the potential of collaborative innovation at the intersection of AI and blockchain. By leveraging decentralized development, innovative funding mechanisms, and a commitment to privacy, this initiative could pave the way for more inclusive and transparent AI advancements. As the project progresses, it will be a focal point for both the AI and blockchain communities, exemplifying the transformative possibilities when these technologies converge.