These Startups Are Building Advanced AI Models Without Data Centers

image via WIRED

April 30, 2025, 4:00 PM

• •Flower AI and Vana collaborated to create Collective-1, a new language model using a distributed training approach.
• •Collective-1 leverages a network of computers and private data sources, potentially disrupting the traditional AI model-building methods.
• •The distributed approach could allow smaller companies and universities to build advanced AI by pooling resources, potentially reshaping the AI industry.
• •Vana's software enables users to share personal data, giving them more control over its use in AI training.

Researchers have developed Collective-1, a new large language model (LLM) utilizing a distributed approach to AI training, challenging the conventional method of building AI models. This approach, pioneered by Flower AI and Vana, leverages distributed computing resources and both public and private data sources, including user-contributed data from platforms like X and Reddit. This innovation allows for a more flexible and potentially more accessible method for AI development, opening opportunities for smaller entities to participate in building advanced AI. The distributed approach to training could also unlock new types of data and give users greater control over their data, which will allow AI to expand to fields such as healthcare and finance.

Read original article