Truly Decentralized Large Language Model with Cardano SPOs

[GENERAL] Name and surname of main applicant

Justin Diamond

[GENERAL] Are you delivering this project as an individual or as an entity (whether formally incorporated or not)

Individual

[GENERAL] Co proposers and Additional applicants

Floriane Le Floch

https://www.linkedin.com/in/floriane-le-floch/

[GENERAL] Requested funds in ada

75000

[GENERAL] Please specify how many months you expect your project to last (from 2-12 months)

9

[GENERAL] Please indicate if your proposal has been auto-translated into English from another language.

No

[GENERAL] Summarize your solution to the problem (200-character limit including spaces)

We will create a decentralized LLM by distributing its parameters across Cardano stake pool operators, integrated with Nunet, for enhanced security and resilience.

[GENERAL] Does your project have any dependencies on other organizations, technical or otherwise?

Yes

[GENERAL] If YES, please describe what the dependency is and why you believe it is essential for your project’s delivery. If NO, please write “No dependencies.”

We will build with Nunet, a Cardano Computation network

[GENERAL] Will your project’s output/s be fully open source?

Yes

[GENERAL] Please provide here more information on the open source status of your project outputs

The code developed for integrating the Large Language Model (LLM) with Cardano stake pool operators (SPOs) and Nunet will be available on platforms like GitHub.

[SOLUTION] Please describe your proposed solution.

Our proposed solution involves repurposing a portion of Cardano stake pool operators (SPOs), or other decentralizied components of Cardano or their partnerchains like Midnight, to host components of a Large Language Model (LLM) and other machine learning algorithms. This approach decentralizes the model's computational and storage requirements across multiple nodes, significantly enhancing its resilience and security against potential attacks and failures.

We perceive the centralization of current LLMs as a critical vulnerability. Centralized models, primarily hosted on a limited number of GPUs, are susceptible to various risks, including hardware failures, targeted cyber-attacks, and bottlenecks in processing capabilities. By distributing the LLM across a decentralized network of SPOs, we mitigate these risks, ensuring a more robust and secure infrastructure.

Our approach is unique in leveraging the existing Cardano infrastructure, a blockchain known for its robustness and decentralization. By utilizing the SPOs, we not only enhance the functionality and utility of the Cardano ecosystem but also contribute to the broader field of AI by presenting a novel method of hosting and operating LLMs.

The primary beneficiaries of this project will be the Cardano community and the broader AI research field. For Cardano, this project demonstrates the versatility and capability of its network, potentially attracting more developers and users to the platform. In the field of AI, it introduces a new paradigm for hosting and operating LLMs, setting a precedent for future decentralized AI projects.

To demonstrate and prove our impact, we will conduct benchmark tests comparing the performance, security, and resilience of our decentralized LLM against traditional centralized models. We will also engage with both the Cardano and AI communities for feedback and improvement suggestions, ensuring that our solution aligns with the needs and expectations of its intended users. This project aligns with Cardano's ethos of decentralization and innovation, offering tangible benefits to its network and stakeholders.

[IMPACT] Please define the positive impact your project will have on the wider Cardano community.

1. Enhancing Network Utility: By integrating a Large Language Model (LLM) into the Cardano network, we significantly increase the practical use cases of the blockchain. This integration not only showcases Cardano's capability to support advanced technological solutions but also potentially attracts a broader user base interested in decentralized AI applications.
2. Fostering Innovation: Our project serves as a pioneering example of how blockchain technology can be leveraged in the field of AI. This encourages innovation within the Cardano community, inspiring developers and stakeholders to explore new and diverse applications on the blockchain.
3. Strengthening Position in Tech Sectors: Deploying a decentralized LLM on Cardano positions the network at the forefront of a new intersection between blockchain and AI. This enhances Cardano's reputation as a versatile and cutting-edge technology platform.

Measuring Impact:

• Quantitative metrics will include the number of transactions related to the LLM, the growth in the number of developers building on Cardano post-implementation, and network performance metrics pre and post-implementation.
• Qualitative measures will involve community feedback, the level of innovation sparked by the project, and the overall sentiment within the Cardano ecosystem regarding the project.

Sharing Outputs and Opportunities:

• Documentation and Reporting: Regular progress reports and final results will be published on community platforms, including Cardano forums and social media channels.
• Open Source Contributions: Code and documentation will be made available on platforms like GitHub, allowing developers to access, use, and build upon our work.
• Community Engagement: We plan to hold webinars and workshops for the Cardano community to educate and encourage participation in both using and developing further on the LLM platform.
• Partnerships and Collaborations: We will actively seek partnerships within the Cardano ecosystem to expand the use cases of our project and collaborate with AI research groups to share insights and findings.

[CAPABILITY & FEASIBILITY] What is your capability to deliver your project with high levels of trust and accountability? How do you intend to validate if your approach is feasible?

Capability to Deliver with Trust and Accountability

1. Academic Credentials and Research Experience: As a PhD student in machine learning in Switzerland, you possess a strong academic foundation in a highly relevant field. Your experience and knowledge in machine learning, evidenced by multiple papers published in prestigious venues like ICML, ICLR, and NeurIPS, demonstrate your capability to understand, innovate, and apply complex ML concepts, which are integral to the success of this project.
2. Partnerships with Project Catalysts and SingularityNET: Your active partnerships with Project Catalysts Fund 8 and SingularityNET's Deep Funding Round 1 highlight your ability to collaborate effectively with significant entities in the field. These partnerships not only bring external validation but also offer access to resources, networks, and expertise that are critical for project delivery.

Validating Feasibility

1. Prototype Development and Testing: Leverage your machine learning expertise to develop a working prototype. Test the prototype in real-world scenarios to evaluate performance and identify areas for improvement.
2. Peer Review and Feedback: Submit your methodologies and findings for peer review within your academic and professional networks, including your connections at ICML, ICLR, and NeurIPS. This will provide critical, expert feedback on the feasibility and potential of your approach.
3. Collaborative Pilot Projects: Utilize your partnerships with Project Catalysts and SingularityNET to set up pilot projects. These projects will serve as practical tests of your approach's feasibility in operational environments.

[Project Milestones] What are the key milestones you need to achieve in order to complete your project successfully?

Initial Integration and Network Testing

Description: The first milestone involves integrating a basic prototype of the Large Language Model (LLM) with a select group of Cardano stake pool operators (SPOs). This stage focuses on establishing the initial connection, ensuring data synchronization, and testing the network's capacity to handle the added computational load.

Output(s):

• A prototype LLM deployed on a subset of SPOs.
• Initial performance and load test reports.

Acceptance Criteria:

• Successful deployment of the LLM prototype on the selected SPOs.
• Positive initial performance metrics indicating the network can handle the computational load without significant disruptions.
• Documentation of any issues encountered and strategies for mitigation in further development.

Decentralization and Security Enhancement

Description: This milestone aims to expand the decentralization of the LLM across a larger network of SPOs and enhance security protocols. Key focus areas include optimizing data distribution, improving resilience against potential attacks, and ensuring data integrity across the decentralized network.

Output(s):

• Expanded deployment of the LLM across more SPOs.
• Enhanced security protocols and systems.
• Comprehensive testing reports on decentralization efficiency and security.

Acceptance Criteria:

• Demonstrable distribution of LLM components across a wider network of SPOs without loss of data integrity.
• Implementation of robust security measures with successful test results against simulated attacks.
• Detailed documentation of decentralization strategies and security protocols.

Nunet Integration and Demonstrative Use Case (Key Milestone)

Description: The final and most crucial milestone involves integrating the LLM with Nunet, providing a complete ecosystem for a decentralized LLM. This milestone will also include the development of a demonstrative use case, showcasing the practical application and benefits of the decentralized LLM on the Cardano network.

Output(s):

• Fully integrated LLM with Nunet.
• A working demonstrative use case of the LLM in action.
• Comprehensive performance and utility reports.

Acceptance Criteria:

• Successful integration of the LLM with Nunet, confirmed through rigorous testing.
• A demonstrative use case that clearly shows the LLM's functionality, benefits, and potential applications on the Cardano blockchain.
• Positive feedback from a select user group trial, confirming the LLM's practicality and utility.

[RESOURCES] Who is in the project team and what are their roles?

Justin Diamond, PhD Student in Machine Learning. Codeveloping Hetzerk, a protocol for decentralized physics simulations on Cardano (Fund8)

https://www.linkedin.com/in/justin-sidney-diamond-881798193/

Justin Diamond - Machine Learning SpecialistKey Responsibilities:Driving advanced research in machine learning algorithms specific to language models.

• Implementing cutting-edge techniques like Normalizing Flows and Graph-based Neural Networks.
• Enhancing the model's accuracy and efficiency through innovative approaches.
• Notable Experience:Currently a Machine Learning PhD Student at the University of Basel, Justin brings the latest academic insights to the project.
• His tenure as a Research Assistant in Physics and Machine Learning at the University of Luxembourg, with a focus on complex modeling techniques like Density Functional Tight Binding and Many Body Dispersion, showcases his deep technical expertise.
• Experience as a Machine Learning Research Scientist at Toyota Technological Institute at Chicago, working on transformer-based 3D protein generative models and RESNET for contact map prediction, highlights his proficiency in applying machine learning to diverse and challenging problems.

Floriane Le Floch - Co-Founder & Chief Technology Officer (CTO) of Evermore a Web3 startup

Key Responsibilities:Steering the project's technological strategy, particularly in integrating advanced AI and machine learning techniques.

• Pioneering the development of the decentralized large language model's architecture.
• Advocating for sustainable and ethical AI practices.
• Distinguished Experience:Floriane's tenure as Co-Founder & CTO at Evermore Labs showcases her expertise in leading innovative tech solutions.
• Her experience as Head of Technology at Thousand Faces and a Founding Member at Lili.ai highlights her proficiency in AI-driven project management and technology implementation.
• Her initiative in founding MAIA, an AI solution in healthcare, demonstrates her ability to create impactful, socially beneficial AI applications.

Floriane Le Floch's leadership and technical acumen position her as an invaluable asset to the project, ensuring a blend of innovation, sustainability, and ethical AI deployment.

[BUDGET & COSTS] Please provide a cost breakdown of the proposed work and resources.

Each Milestone is allocated 25000 ADA

Milestone 1: Initial Integration and Network Testing

• Development and Integration (60%): Resources for coding, integrating, and setting up the LLM with the Cardano stake pool operators.
• Testing and Quality Assurance (40%): Funds for initial testing and ensuring the prototype functions as intended.

Milestone 2: Decentralization and Security Enhancement

• Security and Decentralization (70%): Major focus on enhancing security and effectively decentralizing the LLM across more SPOs.
• Community Engagement and Feedback (30%): Interaction with the Cardano community for insights and improvements.

Milestone 3: Nunet Integration and Demonstrative Use Case

• Nunet Integration and Use Case Development (80%): Predominantly for integrating with Nunet and developing a practical use case.
• Documentation and Outreach (20%): For preparing comprehensive reports and disseminating findings to the community.

[VALUE FOR MONEY] How does the cost of the project represent value for money for the Cardano ecosystem?

Each milestone's cost is strategically allocated to cover essential development phases, from foundational setup to final demonstration. The project promises to enhance Cardano's technological capabilities, attract new users and developers, and position Cardano as a leader in innovative blockchain solutions. The investment in this project thus represents a strategic opportunity for Cardano to expand its influence and utility in the broader blockchain and AI domains.

[IMPORTANT NOTE] The Applicant agrees to Fund Rules and also that data in the submission form and other data provided by the project team during the course of the project will be publicly available.

I Accept