Instruere: Mine the next AI revolution

[thedanand]

Project Name: Instruere
Project Track: Web3
Team Name: Supedevs
Team Member(s): 3,(thedanand, afzal_1112, sahil_0110)
DevPost Project Link: (Instruere | Devpost)

Project Goal:

• Merging Mining with Real-World AI Progress: We’ve designed a protocol where mining directly fuels advancements in AI, creating value in real world leveraging blockchain.

• Leveraging Untapped Resources: We’re harnessing TPU-enabled smartphones and laptops to tap into underutilized devices for decentralized and efficient AI training, onboarding untouched audience to web3.

• Inclusive and Fair Mining Rewards: Our approach ensures that everyone participating in AI training through mining is rewarded, making the process accessible and beneficial for all, regardless of the device they use.

• Bringing Blockchain to Everyone: Our user-friendly extension empower newcomers to contribute to AI training through blockchain mining, making it easy for more people to engage with Web3 in a simple and intuitive way.

Project Value:

• AI Advancement through Blockchain: Connects mining activities directly to real-world AI advancements, ensuring contributions are both useful and relevant.

• Resource Optimization: Utilizes untapped resources like TPU-enabled smartphones and laptops for decentralized AI training, maximizing the efficiency of existing devices.

• Inclusive Mining Rewards: Guarantees that every participant earns rewards for their contributions to AI training, fostering a sense of community.

• Federated Machine Learning: Distributes training tasks based on device specifications, ensuring efficient resource use and minimizing the strain on any single device.

• User-Friendly Interfaces: Our simple extension makes it easy for everyday users to engage in mining, driving broader adoption of blockchain and AI technologies.

• Scalable AI Solutions: Empowers a large network of users to contribute, meeting the increasing demand for AI resources and enhancing scalability in model training.

• Dynamic Adaptation to Device Capabilities: Adjusts training scripts and rewards according to each device’s unique capabilities, optimizing performance and efficiency.

Project Info:Tome

Project Website: [ ]

Project Test Instructions: Please describe and provide step-by-step instructions on how the judges and the community can test out your project

Vision:

To create a decentralized blockchain protocol that harnesses the collective computational power of users’ devices for training AI models while offering rewarding experiences for all participants.

Problem Statement:

While traditional blockchain mining has made remarkable advancements, there’s a fantastic opportunity to further enhance scalability and accessibility. With the increasing demand for AI training resources, we envision a solution that taps into the potential of underutilized devices, such as TPU-enabled smartphones and laptops, to meet this growing need.

Key Features:

1. Decentralized Mining:

• Miners contribute by executing training scripts for machine learning models on their local devices, making efficient use of their computational resources while actively participating in the ecosystem.

2. Script Uploading:

• Small developers can effortlessly upload their Python scripts for training AI models to the Instruere platform, empowering them to contribute their creativity and expertise.

3. Repository ID Validation:

• We utilize the Hugging Face API to validate whether contributors have successfully trained a model. By establishing a reliable proof of work we ensure that only genuine contributions are recognized, enhancing the overall effectiveness of our vision.

4. Block Addition:

• Upon successful validation of a repository ID, the Tron program adds a new block to the blockchain, celebrating the contributions of each participant and promoting a sense of achievement.

5. Mining Rewards:

• Miners are generously rewarded with fungible tokens for their active participation in executing training scripts and validating repository IDs, creating a positive feedback loop that encourages further engagement.

6. Federated Machine Learning:

• Our innovative approach distributes the training load based on device specifications, ensuring that each participant can contribute effectively while optimizing overall performance.

7. User-Friendly Applications:

• The platform features a simple extension that make it easy for users, regardless of their technical background, to upload scripts and join the mining process.

Project Details:

Technical Architecture:

1. Deploying training script

• Deployment Interface: AI engineers upload Python scripts directly through deployment interface.
• IPFS Storage: Python scripts and checkpoints are securely stored on IPFS for decentralized access.

2. Client-Side Mining Interface

• Script Retrieval: Miners initiate the mining process by fetching Python scripts from IPFS upon pressing the “mine” button.
• Local Execution: The scripts are executed locally, utilizing miners’ device resources for AI model fine-tuning.

3. Federated Machine Learning

• Device-Specific Training: A federated learning approach allows TPU-enabled smartphones to distribute training tasks effectively.
• Checkpoint Storage and Retrieval: Checkpoints are saved on IPFS and can be loaded by other miners for continuous training.

4. Model Verification and Block Addition

• Fine-Tuned Model Submission: After training the fine-tuned AI models get pushed to Hugging Face, with the repository ID pre-defined by AI Engineer.
• Verification: Nodes can verify contributions using Hugging Face’s custom API before triggering block addition through smart contracts.

5. Reward Distribution

• Reward Mechanism: A smart contract initiates reward distribution based on miners’ contributions to the fine-tuning process.
• Incentive Structure: Rewards are allocated dynamically, reflecting participation and resource contributions.

Smart Contract links: [ TRON-IDE ]

Project Milestones:

• Phase 1: Prototype Development (0-3 Months) – Build the decentralized AI training framework with TPU, GPU integration and smart contracts.

• Phase 2: Beta Testing (3-6 Months) – Onboard users, conduct beta tests, and refine features based on feedback.

• Phase 3: Public Launch (6-12 Months) – Launch the full platform, integrate cross-chain functionality, and establish partnerships.

• Phase 4: Scaling & Ecosystem Growth (12-18 Months) – Expand user base, support more AI models, and grow the ecosystem with new features.

COMPLETED

• User Onboarding and Environment Setup: Users can sign up, connect their wallets, and upload machine learning scripts linked to their Hugging Face repository IDs. Miners can register their devices and set up local environments.
• Backend Development: The MineClient is being integrated into the backend to facilitate local script execution by miners.
• Frontend Enhancements: A simple extension for the frontend is being developed to enhance user interactions with uploaded scripts and miner functionalities.
• Validation: Custom Hugging face API to verify the POW.
• Reward Mechanism: Currently working on the reward system to ensure miners receive fungible tokens upon successful validations.

[Prince-Onscolo]

information here ae not enough? hope you update it soon.

[manfred_jr]

Welcome to Hackathon Season 7, I quite understand the rush that’s quite similar to panic selling but there are incomplete parts of your offering that needs completion, please fix up so as to equally encourage active participation from forum members.

[thedanand]

thanks for the review good sir,
I am still working on the presentation aspect
will update the information in some hours

[thedanand]

lend me some hours, will keep on updating information

[Smart1]

@thedanand

• Merging Mining with Real-World AI Progress: We’ve designed a protocol where mining directly fuels advancements in AI, creating value in real world leveraging blockchain.
• Leveraging Untapped Resources: We’re harnessing TPU-enabled smartphones and laptops to tap into underutilized devices for decentralized and efficient AI training, onboarding untouched audience to web3.

1. How does the mining process specifically contribute to advancements in AI, and what metrics will you use to measure this impact?

2. What strategies will you implement to encourage users to utilize their TPU-enabled devices for decentralized AI training, and how will you ensure the security and privacy of their data?

3. How does your protocol plan to onboard and educate users unfamiliar with web3 technologies, particularly those using underutilized devices?

[Chukseucharia]

Welcome to the hackathon season 7, for this project to succeed, it require users to contribute their device resources which may be difficult especially for people with less blockchain knowledge.

My question: How will you market the platform to attract users who may find it difficult to contribute their device resources?

[Shangra1]

1. How does empowering a large network of users to contribute facilitate scalability in AI model training?

2. What specific mechanisms are in place to ensure the quality and reliability of contributions from users in scalable AI solutions?

3. In what ways can a scalable AI solution address the growing demand for AI resources across different industries?

[thedanand]

• The interface is really simple, it consists of an extension with a mine button. Users have nothing to do apart from installing the extension and pressing mine button.
• Any user who doesn’t have much knowledge of web3 jargons can still contribute and we can motivate them with rewards and incentives of their mining.
• I’ll foster a supportive community using social media where people can share their experiences and engage with each other. Events like mining challenges and leaderboards will make participation fun and rewarding.

[thedanand]

1. How does empowering a large network of users to contribute facilitate scalability in AI model training?

By enabling a large group of people to contribute their device power, we create a decentralized system for AI model training. Instead of relying on just a few big servers, tasks are spread across thousands of devices—laptops, smartphones, and even TPU-enabled phones. This means:

• More contributors equal more computing power, allowing AI models to be trained faster.
• No bottlenecks because the workload is shared, making the process smoother and more efficient.
• As more users join, the system grows naturally, creating a scalable solution that can handle bigger AI projects without needing expensive infrastructure.

2. What specific mechanisms are in place to ensure the quality and reliability of contributions from users in scalable AI solutions?

• API-Based Validation: Using APIs like Hugging Face, we check if users have successfully completed their AI training tasks and uploaded their models as proof of work.
• Device-Specific Assignments: We match tasks to the right devices based on their capacity, ensuring users only get tasks their devices can handle.
• Federated Learning: This keeps data secure by training models on individual devices without sharing sensitive information.
• Reward Systems: We offer token incentives for high-quality contributions, and a reputation system ensures users consistently deliver reliable results.

These systems work together to create a trusted, high-quality network of contributors, which makes decentralized AI training both scalable and effective.

3. In what ways can a scalable AI solution address the growing demand for AI resources across different industries?

• Tapping into Global Resources: Industries can access a worldwide pool of computational power, without needing expensive servers, to meet their AI needs on demand.
• Faster Training Times: The decentralized nature allows for quicker model training and real-time deployment, which is crucial for industries looking to innovate and adapt.
• Cross-Industry Flexibility: From healthcare to finance, scalable AI can be tailored to fit a wide range of needs—whether it’s predictive modeling, automation, or data analysis.
• Cost Savings: By decentralizing AI training, businesses can save on infrastructure costs and make AI accessible, even for smaller companies that might not have the resources for large-scale computing power.

[thedanand]

1. How does the mining process specifically contribute to advancements in AI, and what metrics will you use to measure this impact?

• Whenever users mine, a python script is fetched from ipfs which is uploaded by AI engineers who don’t have access to high quality GPUs thus contributing to advancements in AI by giving a platform to open source developers who lack resources to do so.
• The script runs locally on the miners system as pow, when the model is trained a new block is added to record the contribution and verified by other nodes thus incentivizing the miner for their resources.
• The python script consist of the model deployment repo of hugging face, using our custom verification API the nodes can verify that the finetuned or trained model has been deployed on huggingface.
• Surge in users on our protocol will be the biggest and the most important metric to measure this impact.

2. What strategies will you implement to encourage users to utilise their TPU-enabled devices for decentralised AI training, and how will you ensure the security and privacy of their data?

We want to make contributing to AI training with TPU-enabled devices an exciting and rewarding experience for everyone:

• The interface is really simple, it consists of an extension with a mine button. Users have nothing to do apart from installing the extension and pressing mine button.
• Any user who doesn’t have much knowledge of web3 jargons can still contribute and we can motivate them with rewards and incentives of their mining.
• I’ll foster a supportive community using social media where people can share their experiences and engage with each other. Events like mining challenges and leaderboards will make participation fun and rewarding.
• Whenever users complete the mining process and their block is added and verified they will be rewarded for allocating their TPU and GPU-enabled devices

Security and Privacy Measures:

• Federated Learning: We use federated learning, so all training happens directly on users’ devices, ensuring that no sensitive data is shared. Only the results of their work are sent back, protecting their privacy.

3. How does your protocol plan to onboard and educate users unfamiliar with web3 technologies, particularly those using underutilized devices?

• the second answer summarizes this question above.

[Anderson]

@thedanand

While traditional blockchain mining has made remarkable advancements, there’s a fantastic opportunity to further enhance scalability and accessibility. With the increasing demand for AI training resources, we envision a solution that taps into the potential of underutilized devices, such as TPU-enabled smartphones and laptops, to meet this growing need.

I really want to try mining with your project when i am able to test ?

[thedanand]

I am currently working on the reward function of the smart contract, as soon as I am done with it I will integrate and share the link to test.

[Nweke-nature1.com]

Welcome to Grand hackathon S7
Apart from devices can boost his/her mining rewards with a subscription strategy?
What kind of token is being rewarded to miners? Or Are you planning to launch a token?

[Prince-Onscolo]

seen good job, i will go through and give you my feedback

[ines_valerie]

Welcome to the Hackathon of season 7, goodluck on your journey, thank you

[kishan1729]

Welcome to hackathon season 7. I like your idea but I have one question: How does Instruere ensure that the AI models being trained are verified as genuine contributions, and how does the platform reward miners for their participation in training these models?

[Chukseucharia]

Ok, in what way will you use the resources of your users. Sell to third party or?