Distributed AI Training Disrupts Industry Power Dynamics | JQJO

Flower AI and Vana have collaborated to create Collective-1, a 7-billion parameter large language model (LLM) trained using a distributed computing approach across numerous globally dispersed computers. This innovative method, unlike traditional data center-based training, leverages both public and private datasets, including data from X, Reddit, and Telegram. The distributed training approach, pioneered by Flower AI, allows for scalability beyond Collective-1's size, with plans for larger models underway. This disruptive technology could democratize AI development, empowering smaller companies and countries with limited resources to create advanced AI models, potentially altering the industry's power dynamics.