Comment
Author: Admin | 2025-04-28
To demonstrate to prospective users the real-time capabilities of their networks. Akash is training their own foundational model and has already launched chatbot and image generation offerings that create outputs using Akash GPUs. Similarly, io.net has developed a stable diffusion model and is rolling out new network functionalities which better mimic the performance and scale of traditional GPU datacenters. Decentralized Machine Learning Training In addition to generalized compute platforms that can service AI needs, a set of specialized AI GPU providers focused on machine learning model training are also emerging. Gensyn, for example, is “coordinating electricity and hardware to build collective intelligence” with the view that, “If someone wants to train something, and someone is willing to train it, then that training should be allowed to happen.”The protocol has four primary actors: submitters, solvers, verifiers, and whistleblowers. Submitters submit tasks to the network with training requests. These tasks include the training objective, the model to be trained, and training data. As part of the submission process, submitters pay a fee up-front for the estimated compute required from the solver.Once submitted, tasks are assigned to solvers who conduct the actual training of the models. Solvers then submit completed tasks to verifiers who are responsible for checking the training to ensure it was done correctly. Whistleblowers are responsible for ensuring that verifiers behave honestly. To incentivize whistleblowers to participate in the network, Gensyn plans to periodically provide purposefully incorrect proofs that reward whistleblowers for catching them. Beyond providing compute for AI-related workloads, Gensyn’s key value proposition is its verification system, which is still in development. Verification is necessary to ensure that external computations by GPU providers are performed correctly (i.e., to ensure that a user’s model is trained the way they want it to be). Gensyn tackles this problem with a unique approach, leveraging novel verification methods called,“Probabilistic proof-of-learning, Graph-based pinpoint protocol, and Truebit-style incentive games.” This is an optimistic solving mode that allows a verifier to confirm that a solver has correctly run a model without having to completely rerun it themselves, which is a costly and inefficient process. In addition to its innovative verification method, Gensyn also claims to be cost effective relative to centralized alternatives and crypto competitors - providing ML training at up to 80% cheaper than AWS while outcompeting similar projects like Truebit in testing. Whether these initial results can be replicated at scale across a decentralized network remains to be seen. Gensyn wants to harness excess compute from providers like small data centers, retail users, and in the future even smaller mobile devices like cell phones. However, as the Gensyn team itself has admitted, relying on heterogenous compute providers introduces several new challenges. For centralized providers
Add Comment