AI/GPU Roadmap Spotlight: AI Provider

What Is AI Provider?

AI is naturally the most exciting type of computation for GPUs. We are developing a Windows AI Provider that will allow sharing GPU resources to perform AI computations on them. This solution will only allow running AI computations and it won't be possible to rent a virtual machine with GPU, load an image and perform any calculations on it. 

The AI Windows Provider, a project that we are doing in cooperation with our partner GamerHash AI, will be one of the modules integrated with the GamerHash AI application.

Why Are We Doing It?

We want to enable the connection of GPU resources to facilitate AI computations. Through cooperation with GamerHash AI, we are able to reach the gaming community and use their GPU resources to provide infrastructure for the AI industry. 

Gamers are the largest segment of the market using consumer GPUs. Using these resources could allow unlimited scaling of AI computations on Golem.

For Whom It Is?

AI Provider is for Gamers with PCs equipped with GPUs, willing to earn money by renting their machine.

Users:

Based on AI Provider, we aim to provide value to:

• AI/ML standalone developers and small teams.
• AI/ML-based SaaS companies and startups - founders, CTOs, AI/ML Ops, Data Scientists and Data Engineers.

Technical Implementation

Users will experience a significant change compared to what they're accustomed to with regular Linux providers. Instead of dealing with command line scripts, our Windows GUI application offers a simpler way to manage your node. The underlying implementation will take care of yagna daemon and provider agent, presenting essential information in a simple and beautiful interface.

The upcoming Provider will introduce new ExeUnits. Recognizing that our standard virtual machine ExeUnit isn't the best choice for the Windows environment at this moment, we're developing specialized ExeUnits tailored to handle specific AI tasks.  AI developers rely on a variety of frameworks to achieve their objectives. Our goal is to offer the most commonly used frameworks among them.

Requestor won’t access the Provider’s machine directly; instead, their requests will be routed to the framework. While this limits the Requestor's flexibility due to lack of general computations, it enhances security for providers.

Benefits for Users

AI Provider users will earn money in the form of USDT. Settlements with users will be handled by GamerHash AI. Users will also become part of the AI revolution and will be able to contribute to the development of the AI industry as well as the ecosystem of the decentralized Golem Network platform itself.

For users of Golem Network (Requestors) who will use the computing power provided by AI Providers, the main benefit will be the low cost of AI infrastructure and the possibility of doing large scale AI calculations (e.g. AI model inference).

Connection with other AI/GPU Projects

AI Provider will provide more computing power than GPU Provider due to the potentially larger number of users who will install it. All planned services and applications will be able to use this power. The only exception will be the Rent GPU application, which allows renting a virtual machine with GPU. This service  will not be able to use AI Provider solutions.

Milestones and Next Steps

We intend to complete the Proof of Concept of the project in Q2, the MVP version in Q3 and then beta test it with users from the GamerHash community. Release is planned for Q4.