Near Protocol has announced plans to build the world’s largest open-source AI model. The platform announced the initiative on the opening day of the Redacted conference in Bangkok, Thailand. The initiative, a 1.4 trillion parameter model, will be larger than Meta’s open source model, Llama. Near Protocol noted that the initiative will include crowdsourced research from multiple researchers and contributors to the new Near AI hub. Contributors are reportedly on board to train a smaller 500 million parameter model starting November 10.
Near Protocol has announced that the project will continue to grow and will include seven models. It also noted that only the best researchers and contributors will be involved in each phase and model. Near Protocol plans to use an encrypted, trusted execution environment to reward contributors while encouraging continuous updates throughout the process. Illia POLOSUKHIN, co-founder of Near Protocol, told an event in Bangkok that the company plans to finance expensive training through a token sale. He noted that the model would cost around $160 million, which is a large amount but does not mean it cannot be raised on the crypto market.
The company must invest as many resources as possible in the project to realize its dream. For example, many GPUs need to be collected on site. However, using a decentralized network for computing requires technology that does not currently exist. The required distributed training methodology also requires the use of a high-speed connection. Polosukhin said he has not yet interacted with projects like the Alliance for Artificial Superintelligence, but he would be happy if both projects followed the same path.
Image: freepik
Designed by Freepik