Google Deepmind introduced the open and globally distributed Dipaco training system, which resembles the work of torrents for neural networks.
Development of DeepMind is a response to the recent Microsoft and Openai statement about the launch of a project worth $ 100 billion to create the next generation supercomputer and the data processing center. The goal is to develop innovative superims of models. The initiative promises to become a significant event in AI, strengthening the positions of Microsoft and Openai as leading players in the market.
Dipaco (Distributed Path Composition) offers the innovative approach to neural networks, using computing power in the world in the world in the world will allow To scaling AI models to sizes, inaccessible to traditional centralized systems.
Distributed Dipaco
system
Dipaco is based on two key algorithms that together create a powerful tool for developing super -widespread AI models:
- diloco, which provides parallelization of data processing processes;
- dipaco, which is responsible for paralleling the processes of teaching models.
The success of Dipaco can call into question not only the value of multi -billion dollar investments of Microsoft and Openai, but also radically change the direction of development of AI. The project has the potential to democratize access to the training of super -minded AI models, thereby destroying the monopoly of current leaders in the market and contributing to rapid progress in the field of AI
In addition, Dipaco can significantly complicate the task for governments, especially for the United States and China, which strive to regulate the development of AI through control over large data centers. The distributed Dipaco structure makes such control much more complex and less effective.