AI is transitioning away from traditional, corporate centralised frameworks and towards more scalable, equitable decentralised infrastructure. On-chain, Web 3 AI is outperforming the open-source Web 2.0 approach to AI and demonstrating more efficient resource utilisation and more performant model hosting and training.

Io.net and FLock’s partnership demonstrates the transformative potential of a decentralised AI stack to redirect governance and value accrual away from the minority of private corporations and into the hands of the public.

Together, the two Web 3.0 AI leaders form a complementary duo: FLock provides the collaborative model training and fine-tuning platform, while io.net provides decentralised compute, enabling the public to access these models.

The Need For Decentralised AI

Centralisation hinders AI innovation, presenting limited computing power and high costs that create significant barriers to public participation.

Controversies in the AI industry, including Gemini AI’s data bias scandal and the proliferation of deepfakes demonstrate a lack of incentives for contributors and a system that shuts out transparent, public guidance in favour of private interests.

The AI industry has seen continued favouring of corporate interests, from OpenAI’s abandonment of transparency pledges to Getty Images’ lawsuit against Stability AI. These examples of corporate biases severely affect model access, transparency, and performance, illustrating the impact of misaligned values between technology developers and users.

Open source Web 2.0 AI also faces challenges with monetisation. Research projects are often abandoned prior to publication due to a lack of capital or incentive to productise models.

On-chain AI Stack

By decentralising the AI stack and hosting models on independently operated nodes, the industry can remove centralised interests and incentivise the community of developers in a transparent way.

With community governance and contribution, an on-chain AI stack can produce a plethora of useful models created by, and for, communities. Users can earn rewards for contributing data, models, or compute.

In a decentralised stack, AI models are trained across a community-owned and -operated network. Federated learning powers the platform, where models learn from decentralised data sources without moving the data itself. This allows for collaboration without the problem of centralised data collection and potential misuse.

Building on Web 2.0’s open source approach, models are free to use, and the addition of Web 3.0 crypto economic incentives drives momentum in open source, composable development.

Decentralised hosting leverages idle computing power to make compute more accessible. The price of renting an A100 for a month can go up to $3,500, whereas io.net offers compute at 90% lower cost than traditional cloud providers.

The main benefit of decentralised hosting is the far greater access to computing power at a fraction of the price of comparable centralised services.

The io.net and FLock.io partnership

FLock.io and io.net joined forces to combine the benefits of decentralised training and hosting. In this partnership, models trained on FLock.io are provided as a service on io.net.

Key advantages are streamlined model creation, accelerated scalability, and lower usage costs, while providing a system that is composable and traceable via on-chain inference.

Access to compute can be considered as our generation’s ‘digital oil’. io.net is monetising distributed hosting and solving the problem of shortages and high costs by building an ecosystem fueled by a currency of compute. Io.net is on a mission to build the world’’s largest AI compute DePIN (decentralised physical infrastructure network).

What’s next? FLock training and validation clients will run on io.net’s compute. FLock plans to deploy additional LLMs to io.net for inference. For example, several gaming, AI companions and Web 3.0 agents have used FLock for fine-tuning, and can be added to io.net’s decentralised AI compute network.