Yigit Taha OZTURK 
Near Protocol announced plans to create the world’s largest open-source artificial intelligence model at the Redacted conference in Bangkok. This model will be 3.5 times larger than Meta’s current open-source Llama model, with 1.4 trillion parameters.
The project will be developed through the contributions of thousands of participants on the new Near AI Research hub. From today, November 10, participants can join the training of a smaller 500 million parameter model. The project will grow across seven models, with only the best contributors advancing to work on progressively more complex and larger models.
The model’s privacy and security will be ensured through encrypted Trusted Execution Environments (TEEs). Contributors will be rewarded, and constant updates will be encouraged as the technology progresses. Near Protocol co-founder Illia Polosukhin mentioned that the expensive training and computing costs will be funded through token sales, raising about $160 million.
Polosukhin explained that token holders would be repaid from the inferences made when the model is used. This creates a business model that can be reinvested into future models.
With contributions from Illia Polosukhin, one of the authors of the groundbreaking transformer research behind ChatGPT, and Alex Skidanov, who has OpenAI experience, the Near team is well-equipped for the task. However, Skidanov mentioned that training such a large model would require “tens of thousands of GPUs,” which is not ideal for a centralized system. A decentralized network of compute may be the solution, but it would require new technology that does not yet exist.
Near Protocol‘s ambitious open-source AI project could revolutionize both the crypto and AI industries if successful.
You can also freely share your thoughts and comments about the topic in the comment section. Additionally, don’t forget to follow us on our Telegram, YouTube, and Twitter channels for the latest news and updates.