Bittensor Achieves Breakthrough in Decentralized AI Training for 100B+ Parameter Models, $14B ReadyAI

Sep 21, 2024, 04:58 AM

Bittensor has achieved a significant breakthrough in decentralized AI training, allowing for the training of models with over 100 billion parameters. This decentralized training is facilitated by collaborations with MacrocosmosAI, Manifold Labs, and const_reborn. The platform also boasts incentivized training over a wide web with 1000x model compression. Bittensor's advancements promise 100x faster training speeds, with potential future improvements reaching up to 850x, aided by ALADDIN & BISTRO. The decentralized approach is seen as a potential competitor to centralized AI systems, with ReadyAI, a Bittensor subnet, valued at $14 billion. All models on Bittensor are required to be fully open source, including weights and data. The $TAO token is integral to these developments.