About Bloke

Bittensor, Gensyn, and Together are Bloke's closest competitors in the Web3 ML landscape. Gensyn and Together's inability to orchestrate a MoE model limits their ability to compound AI. Bittensor is a Parachain on Polkadot and therefore cannot independently implement its consensus mechanism. Bloke's composable ML architecture, also known as AI legos, forms the foundation of AI compounding.

If AI compounding becomes a reality and Bloke becomes the most powerful ML model in the world.

Opportunity

AI is essentially an advanced productive force, whose rapid development depends on three core elements: data, algorithms, and arithmetic power; blockchain is more of a production relationship that facilitates changes in these three elements through changes in incentives, coordination, and forms of organization. We see a narrative direction in which these two fields are intertwined:

1. Blockchain + Arithmetic: The implementation of AI models requires powerful computing power, which the big players or some of the compute resource providers have on hand, but one can still consider a long-tail market, where dispersed compute resources (personal graphics cards/devices), etc. contribute arithmetic and are thus incentivized by cryptocurrencies. Cryptocurrencies that benefit from this narrative are for example RNDR(Render Network) and other projects that can contribute arithmetic.

2. Blockchain + Algorithms: Unlike the first two, which are "resource-intensive", algorithms themselves are technology-intensive and are the secret formula and barrier to continuous iteration for AI companies, and it's hard to "create" a better algorithm from 0 through cryptocurrency incentives; the logic of contributing, coordinating, and incentivizing doesn't work for algorithm creation.

However, you can use incentives to 'sift' through existing algorithms to create a better algorithm without everyone using the same thing. Similar to how the Prophecy Machine project encourages competition through incentives to pick out better data sources. Neither directly contributing data nor arithmetic, through the blockchain network and incentives to schedule and screen different algorithms, thus allowing the AI field to form a free competition, knowledge sharing algorithm (model) market.

There are some problems with the current ecological status quo in the AI field. the players in the AI track are currently isolated in each of their algorithms and models. Due to commercial competition, you can't let two algorithms learn from each other to make progress together; this also means that from the AI supply side, the competition is zero-sum: if one AI wins the market, the others will go out.

Assuming that Model A is fluent in English and Model B is fluent in writing code when a user needs to ask the AI to explain the code with English comments, the combined output of the two algorithms will be the most effective, but this is not possible in the current environment; Bloke's goal is to allow the algorithms and models of different AIs to collaborate, learn, and combine to form a more powerful model that can better serve developers and users.

In the field of DeFi, financial components such as stablecoin, lending, liquidity mining, etc. are all open-source and unlicensed, and the demand side can combine them arbitrarily, just like Lego blocks, to form new products and services.

The same is true in the AI algorithm field, where AI algorithm models that specialize in image processing, word processing, or audio processing can be combined to serve different tasks and form AI Legos. For Bloke itself, it will neither compute by itself nor provide data to do machine learning on the chain, but mobilize all other AI models under the chain to work together.

About Bloke

Bloke was founded in 2021 to make AI more accessible and efficient. Bloke can expand its AI functions faster and more efficiently than isolated models by putting together AI Lego blocks. One potential solution is the use of blockchain technology to incentivize and coordinate a global network of machine learning nodes to collaborate on specific problems. By adding incremental resources to the network, overall intelligence can be increased by building on the work of previous researchers and models.

The goal of getting different AIs to collaborate is a big one, but how can it be achieved? Bloke's answer is to build a blockchain network that coordinates and operates by mining incentives. Bloke uses a high-performance public chain and smart contracts on it to specifically handle the collaboration of the AI models and has its token $BAI for incentivizing it.

To learn more about how Bloke works, continue reading.

Bloke aims to scale its AI capabilities quickly and efficiently using this approach. As a layer1 blockchain that functions as an on-chain oracle, Bloke connects and orchestrates off-chain ML nodes to create a decentralized mixture-of-experts (MoE) network. This network combines multiple models, optimized for different capabilities, to create a stronger overall model.

The supply side consists of two layers: AI (Miners) and blockchain (Validators).

Developers can use Validators to build applications and access use-case-specific AI capabilities from the network.

There are at least three questions that need to be answered to figure out how Bloke works: First, what are the roles in the chain? Second, what are these actors doing? What are the connections between them? Third, what behaviors do the tokens incentivize in these roles?

Roles

Users need better AI models, validators are responsible for filtering better AI models according to different uses, miners provide their own AI models, and nominators choose to support different validators. Users input their needs, validators route the needs to miners in the Bloke network; miners output the answers, validators then evaluate the quality of the answers, and finally return to the users.

User: The end user of the AI model provided by Bloke. It can be individuals or developers seeking AI models for their applications. Sequencer: Sorting transactions and creating blocks for the Bloke network.

Miners: Understood as providers of various AI algorithms and models around the world that host AI models and make them available to the Bloke network; different types of models make up different subnets, such as models that specialize in pictures or sounds.

Validators: Evaluate the quality and effectiveness of AI models, rank AI models based on performance for specific tasks, and help consumers find the best solutions.

Nominator: Delegate tokens to a specific validator to show support, or you can switch to a different validator to delegate.

Network of ML nodes

Bloke is an open AI supply and demand chain: the miners provide different models, the validators evaluate different models, and users use the results provided by the best models. The users need better AI models and input their requirements. The validators route them to miners in the Bloke network, responsible for filtering better AI models for different purposes; The miners output answers, and provide their own AI models; And the validators evaluate the quality of the answers and ultimately return them to the users. The nominators choose to support different validators.

Token incentives

$BAI is the utility token of the Bloke network. It incentivizes supply-side participation and serves as a demand-side payment. Additionally, it can be delegated (staked) to specific validators, who can bond to specific miners, making the token a revenue share conduit.

For validators: the more accurate and consistent the screening and evaluation of the AI model, the greater the reward. To become a validator, you need to stake a certain number of $BAI tokens.

For miners: provide their models in response to user demand, and receive BAI tokens based on their contributions.

For nominators: entrust their $BAI to the validators, similar to the liquidity staking reward

For users: pay $BAI tokens to start a task, which is equivalent to spending.

Ideally, different AI models in this network would collaborate and the probability is that different models will perform differently for different tasks; since these tasks are visible to the chain of checkable network nodes, the models do learn from each other to adapt differently to the task.