Grass (GRASS) — Decentralized AI Data Collection

Advanced11/4/2024, 6:54:01 AM
Grass is a DePIN project built on the Solana network that leverages unused internet bandwidth to gather information from public networks. This information is then used to train large language models (LLMs) and establish a transparent data marketplace that rewards all participants. The protocol utilizes the bandwidth of users' devices to search for necessary information, process the collected data, and record its provenance history on the blockchain using zero-knowledge proofs (ZKPs).

What is Grass?

Introduction to the Grass Project

Grass is a DePIN project built on the Solana network that leverages unused internet bandwidth to gather information from public networks. This information is then used to train large language models (LLMs) and establish a transparent data marketplace that rewards all participants. The protocol utilizes the bandwidth of users’ devices to search for necessary information, process the collected data, and record its provenance history on the blockchain using zero-knowledge proofs (ZKPs).
LLMs, or large language models, are trained on billions of words and phrases from the internet to understand how language works. The more data they have, the smarter they become. Grass provides a continuous stream of public network data, ensuring that AI models stay up-to-date and improve over time. Grass is the flagship product of Wynd Network, founded in 2023 by Andrej Radonjic and Chris Nguyen.

Grass Architecture

Grass’s creators describe it as a fairer data marketplace model compared to existing Web2 monopolies, which don’t share revenues from AI training content and control access to information. Essentially, Grass has created a network of thousands of user devices that gather information from websites and deliver it to interested clients, mainly for training AI models. The network’s input end is the client, which sends Grass requests to obtain data from specific sources. The output end is the web server containing the requested information. The protocol directs the client’s requests to a specific node that contacts the server, scrapes the data, encrypts it, and sends it back.

Key participants in this system include:

  • Nodes: User devices with the Grass client installed, providing unused internet bandwidth to fetch information from network servers and pass it to routers.
  • Routers: Special coordination nodes that track the status of connected nodes. Routers direct requests to specific nodes and forward received responses to validators.
  • Validators: Validators verify client requests and pass them to routers, as well as encrypt and sign data received from routers. Additionally, they evaluate nodes’ responses based on data integrity, timeliness, and other criteria.

Problems Grass Addresses

Problem 1

Grass serves as a data layer for AI infrastructure, collecting, cleaning, and structuring the information needed to train AI models. By making data accessible, Grass removes disparities between large labs and small AI developers. For instance, Reddit provides free API access to Google through exclusive agreements, but third-party users cannot access it. Twitter, Meta, Medium, and other Web2 platforms have also restricted data scraping. Grass replaces large data centers, which are easy to detect and block, with a decentralized network of user devices. This allows data to be collected through tens of thousands of small channels with residential IPs, bypassing data access restrictions. The protocol only requests publicly available information.

Problem 2

Another issue is the “poisoning” or intentional distortion of data arrays by data sources or providers. This is a common strategy used in “data wars” to counter data scraping. It involves deliberately distorting data accessible through open APIs, including introducing “noise,” to make industrial-scale collection and further use challenging. For example, after launching new AI models like Gemini, media reports have highlighted biased responses towards certain racial or social groups. This is a direct result of training on incorrect information. Given the vast amount of data, manual verification or tracking of changes made during structuring is impractical. Grass addresses this issue through blockchain and zero-knowledge proofs, enabling the verification of information sources, confirming which node responded to the request, and where the information originated. Therefore, any independent AI developer can request information from network servers via Grass at a relatively low cost or purchase cleaned and structured databases for model training.

Grass’s Funding Background

September 14, 2024: Grass completed a Series A funding round led by Hack VC, with participation from Polychain Capital, Delphi Digital, Lattice, and Brevan Howard Digital. The amount raised was undisclosed.
December 18, 2023: Grass completed a $3.5 million seed round led by Polychain Capital and Tribe Capital.
Latest funding round valuation: $1 billion

Grass Tokenomics

$GRASS: Incentive Token

$GRASS will serve as the primary incentive mechanism for the protocol, allowing holders to participate in the Grass network in the following ways:

  • Transactions and Buybacks: $GRASS will be used to support network scraping transactions, dataset purchases, and LCR usage.
  • Staking and Rewards: $GRASS can be staked on routers to facilitate network traffic, and rewards will be given to contributors who enhance network security.
  • Network Governance: $GRASS holders can participate in the development of the Grass network, including proposing and voting on network improvements, coordinating partnerships, and determining incentives for all stakeholders.

Token Distribution

The total supply of $GRASS tokens will remain fixed at 1,000,000,000 tokens.

  • First Season Airdrop Rewards: 10%
  • Future Incentives: 17%
  • Router Rewards: 3%
  • Ecosystem Development: 22.8%
  • Early Contributors: 22%
  • Investors: 25.2%

Gate.io Has Launched Spot and Future Trading for GRASS. Check Out the Latest Prices, Charts, and Data of GRASS/USDT Spot and GRASSUSDT Perp!

How to Participate in the Second Season Airdrop

Grass’s first season airdropped 100 million tokens, and the second season airdrop of 170 million tokens has now begun. There are three mining rates: mobile at 1x, advanced nodes at 1.25x, and desktop at 2x.

  1. Step 1: Register at https://app.getgrass.io/register
  2. Step 2: Verify your email after registration, then download the node for your device
  3. Step 3: Connect your Solana wallet
  4. Step 4: Start mining
* The information is not intended to be and does not constitute financial advice or any other recommendation of any sort offered or endorsed by Gate.io.
* This article may not be reproduced, transmitted or copied without referencing Gate.io. Contravention is an infringement of Copyright Act and may be subject to legal action.

Grass (GRASS) — Decentralized AI Data Collection

Advanced11/4/2024, 6:54:01 AM
Grass is a DePIN project built on the Solana network that leverages unused internet bandwidth to gather information from public networks. This information is then used to train large language models (LLMs) and establish a transparent data marketplace that rewards all participants. The protocol utilizes the bandwidth of users' devices to search for necessary information, process the collected data, and record its provenance history on the blockchain using zero-knowledge proofs (ZKPs).

What is Grass?

Introduction to the Grass Project

Grass is a DePIN project built on the Solana network that leverages unused internet bandwidth to gather information from public networks. This information is then used to train large language models (LLMs) and establish a transparent data marketplace that rewards all participants. The protocol utilizes the bandwidth of users’ devices to search for necessary information, process the collected data, and record its provenance history on the blockchain using zero-knowledge proofs (ZKPs).
LLMs, or large language models, are trained on billions of words and phrases from the internet to understand how language works. The more data they have, the smarter they become. Grass provides a continuous stream of public network data, ensuring that AI models stay up-to-date and improve over time. Grass is the flagship product of Wynd Network, founded in 2023 by Andrej Radonjic and Chris Nguyen.

Grass Architecture

Grass’s creators describe it as a fairer data marketplace model compared to existing Web2 monopolies, which don’t share revenues from AI training content and control access to information. Essentially, Grass has created a network of thousands of user devices that gather information from websites and deliver it to interested clients, mainly for training AI models. The network’s input end is the client, which sends Grass requests to obtain data from specific sources. The output end is the web server containing the requested information. The protocol directs the client’s requests to a specific node that contacts the server, scrapes the data, encrypts it, and sends it back.

Key participants in this system include:

  • Nodes: User devices with the Grass client installed, providing unused internet bandwidth to fetch information from network servers and pass it to routers.
  • Routers: Special coordination nodes that track the status of connected nodes. Routers direct requests to specific nodes and forward received responses to validators.
  • Validators: Validators verify client requests and pass them to routers, as well as encrypt and sign data received from routers. Additionally, they evaluate nodes’ responses based on data integrity, timeliness, and other criteria.

Problems Grass Addresses

Problem 1

Grass serves as a data layer for AI infrastructure, collecting, cleaning, and structuring the information needed to train AI models. By making data accessible, Grass removes disparities between large labs and small AI developers. For instance, Reddit provides free API access to Google through exclusive agreements, but third-party users cannot access it. Twitter, Meta, Medium, and other Web2 platforms have also restricted data scraping. Grass replaces large data centers, which are easy to detect and block, with a decentralized network of user devices. This allows data to be collected through tens of thousands of small channels with residential IPs, bypassing data access restrictions. The protocol only requests publicly available information.

Problem 2

Another issue is the “poisoning” or intentional distortion of data arrays by data sources or providers. This is a common strategy used in “data wars” to counter data scraping. It involves deliberately distorting data accessible through open APIs, including introducing “noise,” to make industrial-scale collection and further use challenging. For example, after launching new AI models like Gemini, media reports have highlighted biased responses towards certain racial or social groups. This is a direct result of training on incorrect information. Given the vast amount of data, manual verification or tracking of changes made during structuring is impractical. Grass addresses this issue through blockchain and zero-knowledge proofs, enabling the verification of information sources, confirming which node responded to the request, and where the information originated. Therefore, any independent AI developer can request information from network servers via Grass at a relatively low cost or purchase cleaned and structured databases for model training.

Grass’s Funding Background

September 14, 2024: Grass completed a Series A funding round led by Hack VC, with participation from Polychain Capital, Delphi Digital, Lattice, and Brevan Howard Digital. The amount raised was undisclosed.
December 18, 2023: Grass completed a $3.5 million seed round led by Polychain Capital and Tribe Capital.
Latest funding round valuation: $1 billion

Grass Tokenomics

$GRASS: Incentive Token

$GRASS will serve as the primary incentive mechanism for the protocol, allowing holders to participate in the Grass network in the following ways:

  • Transactions and Buybacks: $GRASS will be used to support network scraping transactions, dataset purchases, and LCR usage.
  • Staking and Rewards: $GRASS can be staked on routers to facilitate network traffic, and rewards will be given to contributors who enhance network security.
  • Network Governance: $GRASS holders can participate in the development of the Grass network, including proposing and voting on network improvements, coordinating partnerships, and determining incentives for all stakeholders.

Token Distribution

The total supply of $GRASS tokens will remain fixed at 1,000,000,000 tokens.

  • First Season Airdrop Rewards: 10%
  • Future Incentives: 17%
  • Router Rewards: 3%
  • Ecosystem Development: 22.8%
  • Early Contributors: 22%
  • Investors: 25.2%

Gate.io Has Launched Spot and Future Trading for GRASS. Check Out the Latest Prices, Charts, and Data of GRASS/USDT Spot and GRASSUSDT Perp!

How to Participate in the Second Season Airdrop

Grass’s first season airdropped 100 million tokens, and the second season airdrop of 170 million tokens has now begun. There are three mining rates: mobile at 1x, advanced nodes at 1.25x, and desktop at 2x.

  1. Step 1: Register at https://app.getgrass.io/register
  2. Step 2: Verify your email after registration, then download the node for your device
  3. Step 3: Connect your Solana wallet
  4. Step 4: Start mining
* The information is not intended to be and does not constitute financial advice or any other recommendation of any sort offered or endorsed by Gate.io.
* This article may not be reproduced, transmitted or copied without referencing Gate.io. Contravention is an infringement of Copyright Act and may be subject to legal action.
Start Now
Sign up and get a
$100
Voucher!