A Trust and Reputation framework for AIs

chevron-icon
Back
project-presentation-img
Eric Duneau
Project Owner

A Trust and Reputation framework for AIs

Funding Requested

$50,000 USD

Expert Review
Star Filled Image Star Filled Image Star Filled Image Star Filled Image Star Filled Image 0
Community
Star Filled Image Star Filled Image Star Filled Image Star Filled Image Star Filled Image 4.5 (8)

Overview

Without a verifiable proof of Reputation of third party AI services, everyone in the SingularityNet ecosystem (users, programs, and AIs) is at the mercy of using the service of bad agents. This proposal is for the development of an open-source decentralised Identity Agent to manage AI's Reputation. Built upon a Digital Identity Trust Framework, it will enable authorised issuers and verifiers to act as the guardians of truth via the issuance of Digital Identities and Verifiable Credentials.

Proposal Description

How our project will contribute to the growth of the decentralized AI platform

Currently, there is no way to get proof of performance and a tracking of past capabilities of an AI residing on the platform. Furthermore, those AIs could get replaced by other versions, trained on different datasets, and nobody will notice. The opacity of how those AIs have been trained and how they operate is striking. 

It is easy to see that if the platform does not come with standards for the Reputation of its AIs, it could soon become a host for rogue programs.

The core problem we are aiming to solve

As AI capabilities expand, a marketplace emerges where users pay for tailored AI services. This marketplace also allows for programmatic calls to those AI services (API calls). Finally, we enter the era of AI to AI calls and AI service consumption by other AIs. Verifying an AI's Identity, Capabilities, Performances, and ultimate Reputation, akin to validating human skills and credentials, becomes critical.

Without a system establishing trust, any end-user, program, or benevolent AI, is at the mercy of calling an AI service which could be either deficient or even malevolent.

The establishment of TRUST in the whole eco-system is critical.  It starts with authenticating AIs via their Digital Identity, and authorising the consumption of their services via Verifiable Credentials (VCs). Those VCs taken altogether then build a reputation score for each actor in the eco-system.

 

Our specific solution to this problem

The establishment of trust in the whole eco-system must come from a Digital Identity Trust Framework, which stamps AIs with a Digital Identity and allows the issuance and verification of Verifiable Credentials to establish its past capabilities and performance with a track record which cannot be tampered.

This proposal will be delivered in 4 milestones, mapping the capability section defined here below:

(i) designing a Digital Identity Trust Framework for establishing a live reputation score for each actors

(ii) delivering an open source package that all other AI projects will be able to include in their project to enable trust and reputation

(iii) delivering a cloud agent that will simplify the integration of Digital Identities and Verifiable Credentials within AI components. This also includes APIs, sample code, and documentation.

(iv) integration of the cloud agent into OpenSourceAIs as a example first implementation of a trust enabling open-source AI ecosystem.

The total duration of the project is estimated at 26 weeks (see details in each milestone), or exactly 6 months. At the moment, no extra risk was added to the project plan, so I would prefer to set the expectation of a delivery within 8 months, and keep a margin to deliver a good surprise of a live solution a few weeks/months earlier. 

The competition and our USPs

Yes

Describe how your solution distinguishes itself from other solutions (if exist) and how it will succeed in the market.

There is currently no solution in place for this. Although the concepts behind a Digital Identity and Trust Framework have matured a lot in the past 4 years, there is still a lack of commercial solution in general, and none apply to the specific field of AIs. The most advanced Digital Identity framework in the closest ecosystem to SingularityNet is Atala Prism, which runs on Cardano. In April 2024, the project was renamed Hyperledger Identus, as it moved ownership from IOHK to Hyperledger. I have been part of the Atala Prism pioneer cohort since late 2022, and I have built many prototypes around the solution. It may be important to state that even at this stage, Identus is only available on pre-prod (Cardano testnet). Therefore, any solution we build now is going to hit commercial status around the same time as Identus gets itself to Production, allowing this AI Identity solution to be perfectly timed to be the first one on the market.

Our team

Eric specialises in AI, Cyber and Crypto. He has delivered open-source libraries in Digital Identity (see SignWithWallet.com), and has built OpenSourceAIs (OSAIS) from the grounds up, as part of a small grant offered by SingularityNet in round 2 (see links in section below). The project was delivered within a few months. Integration of OSAIS into SNET is planned for June 2024. As an early integrator of services, Eric is among the first few to have gone through the integration phase into SNET.

View Team

What we still need besides budget?

No

Existing resources we will leverage for this project

Yes

Description of existing resources

We will leverage those existing resources:

 - Hyperledger Identus (previously known as Atala Prism) will be used as a base layer for issuing Digital Identities and Verifiable Credentials

 - OpenSourceAIs will integrate the Digital Identity and Trust capabilities on all open-source AIs it has already integrated, and all future integration

 - The Cardano blockchain will be leveraged to store Digital Identities issued by AIs

Open Source Licensing

mit

Describe license details and, if applicable, list any components that are not subject to this license.

All open source components (design, Trust Framework package, integration samples, and documentation) will be delivered under the MIT license

The specific implementation of Digital Identity on AIs hosted within OpenSourceAIs will not be made open source, but will follow the guidance of the integration samples which will all be delivered open-source. 

Links and references

https://www.linkedin.com/in/ericduneau/   Eric's LlinkedIn profile

https://deepfunding.ai/proposal/integration-of-mainstream-open-source-ais/  DF2 OSAIS project submission

 

Additional videos

https://www.youtube.com/watch?v=GiVspHiACtU   July 2023 completion milestone report for OSAIS

https://youtu.be/MG6QZMnRUZY  April 2023 pitch to Charles Hoskinson and the Atala Prism team. This project was ranked top 3 most needed project in the Digital Identity ecosystem. I have implemented several prototypes around this initiative since then.  Note that for confidentiality reasons, the discussion with Charles that followed the presentation was edited out of this video extract.

Proposal Video

Placeholder for Spotlight Day Pitch-presentations. Video's will be added by the DF team when available.

  • Total Milestones

    4

  • Total Budget

    $50,000 USD

  • Last Updated

    10 May 2024

Milestone 1 - Design of a Digital Identity Trust Framework

Description

A complete design of the solution will be established in this first milestone. This is planned to take around 6 weeks

Deliverables

(i) A public GitHub repo ; (ii) A few video proof of some tech validation / Proof of concepts supporting the choice of the future tech stack ; (iii) Documented choice of tech stacks for the upcoming milestones ; (iv) Defined specifications for each component (DID VC content Trust Registry metadata) ; (v) Defined specifications of APIs for third party integration

Budget

$10,000 USD

Milestone 2 - Delivery of an open source package

Description

This milestone is about the delivery of the open source code for enabling AI's trust and reputation. This is planned to take 12 weeks, with a +/- 2 week risk.

Deliverables

(i) APIs for assigning a DID to an AI ; (ii) APIs for creating a Verifiable Credential for an AI ; (iii) APIs to get general information and stats from this AI (likely in the form of a JSON format)

Budget

$25,000 USD

Milestone 3 - Delivery of an Identity Cloud Agent

Description

Using the output of the Milestone 2 we will develop and host in a test and a prod environment an operational Identity Cloud Agent so that third party developers do not have to worry about infrastructure and can tap directly into the APIs of the Cloud Agent. This is planed to take 4 weeks.

Deliverables

(i) a hosted Identus Agent Node (likely in Test mode until Identus gets Prod ready) ; (ii) a hosted Identity Cloud Agent ; (iii) sample code for integrating into the Identity Cloud Agent ; (iv) documented agent and APIs ; (v) video proof of all the above

Budget

$10,000 USD

Milestone 4 - Integration into OpenSourceAIs

Description

As a final proof that the solution is working and can be integrated into any AI project we will show a full scale integration of a Reputation system into OpenSourceAIs (OSAIS). By this time OSAIS should be production ready in SNET so it will be the first AI system on SNET with an embedded decentralised reputation system with proof of Identity stored on the Cardano Blockchain. This integration shall be made within 4 weeks.

Deliverables

(i) live OSAIS on SNET with an integrated Reputational trust system ; (ii) video proof of the end-to-end solution

Budget

$5,000 USD

Join the Discussion (10)

Sort by

10 Comments
  • 1
    commentator-avatar
    HenriqC
    May 15, 2024 | 9:55 AM

    Would it be possible for an AI agent to prove who (other agents) it has been interacting with? 

    Reply
    Upvoted by Project Owner
    • 1
      commentator-avatar
      Eric Duneau
      May 15, 2024 | 11:02 AM

      Hey Henriq, There will be a verifiable proof of all service providers called by an AI at each request. Services providers can be providers of GPU, of models, or a complete AI providing a full service. Kind regards Eric

  • 0
    commentator-avatar
    Cardano4Seniors
    May 15, 2024 | 3:10 AM

    I question as I have for most proposals in this new category is how do you justify submitting in the Misc category. I question that your proposal may meet the criteria for the New Pool and may be excluded on a rules review.  

    • 0
      commentator-avatar
      Eric Duneau
      May 15, 2024 | 6:36 AM

      Hey Cardano4seniors. This proposal is in Misc category because it will not be an AI service and will not generate fees on the platform. I did ask guidance from Deepfunding before selecting the category, and received the answer that it had to be in Misc because of the reason explain above. Considerinng my chances to get funded would have been far greater in New Projects (a far larger pool of AGIX), I would have preferred to have my proposal there, but I need to comply with the rules, so it is in Misc. Kind regards - Eric

  • 1
    commentator-avatar
    Jan Horlings
    May 1, 2024 | 5:07 PM

    Hey Eric,

    thanks for your proposal! It would be good to try and get some alignment with the internal team on this. I'll forward this proposal to some. 

    Another aspect is your intention to use Atala prism / hyperledger Identus to create identities. 
    I have been considering to offer community members, the option of an identification method in WaLT (our Wallet Linking Tool) and I have been thinking that Atala Prism/hyperledger Identus might be a good candidate. 
    It's a bit off-topic, would like to hear your thoughts. We can also continue that on another channel if your prefer :-) 

    Cheers!

    Reply
    Upvoted by Project Owner
    • 0
      commentator-avatar
      Eric Duneau
      May 2, 2024 | 11:14 AM

      Hey Jan - thank you for your feedback here. As you know, I have been in the Atala Prism ecosystem for around 18 months now. Prior to this, I have developed 2 identity/authentication solutions (SignWithWallet.com and authenly.com). In relation to DIDs and VCs, in the end, the current situation with Identus is thst those DIDs would be issued on the Cardano Blockchain, in what could be seen as a Digital Identity wallet (although to my view, it could/should also be used as a payment wallet).  I am happy to continue the discussion either here or in a separate channel. Kind regards - Eric 

  • 1
    commentator-avatar
    Ahan M R
    Apr 29, 2024 | 8:51 PM

    How will your framework detect and manage AI services that become malicious or start functioning in unintended ways over time? Just curious if you've already planned this? Thanks!

    • 0
      commentator-avatar
      Eric Duneau
      May 2, 2024 | 11:19 AM

      Hi Ahan - the idea is that EACH transaction is marked in a Verifiable Credential (VC), and also each supplier of the AI gets a way to verify its supplier (of model, GPU, or even annother full AI acting as service provider), by checking its VCs. So if an AI starts to take more money than it should, or takes a longer time than advertised, or fails too often, etc... then this is immediately stamped for ever in a VC accessible to all those wanting to call this AI's service. The AI reputation is a live one, and can improve or get worse at each call. I am not going in further details here, but I think you can get the point. Kind regards - Eric

  • 1
    commentator-avatar
    Ahan M R
    Apr 29, 2024 | 8:24 PM

    What technical standards are you basing your Digital Identity Trust Framework on? How do you plan to ensure compliance with international regulations on digital identity and data privacy?

    • 0
      commentator-avatar
      Eric Duneau
      May 2, 2024 | 11:24 AM

      Hi Ahan : first, a lot of the framework for "processing" is defined within Identus. Then, there is a requirement for a full Trust Registry, which is one of the main delivery of my Milestone 1.  In regards to data privacy, i will refer you to how VCs are generally built via DIDComm (secure messaging between 2 unrelated parties), plus the fact that VCs are generally NOT onchain (to avoid the risk of encrypted data being broken in a few years when quantum algo get mainstream).  In addition, we are talking here about an AI's digital identity, so not a human "data privacy" issue per say. Happy to give you a bit more info on the topic if necessary, but I think you start to get some answers to your question. Kind regards - Eric

Reviews & Rating

Sort by

8 ratings
  • 0
    user-icon
    Victor2815
    May 13, 2024 | 6:11 AM

    Overall

    5

    • Feasibility 5
    • Viability 4
    • Desirabilty 5
    • Usefulness 5
    useful and effective

    Personally, I think this project will be very complex. I know this is not the first time the project proponent has done this, but I want to ask if the project proponent (Mr. Eric) is the only one working, will the work be overloaded? for you or not? For today's global society, AI technology plays a very important role and can be the core of the technology industry in the future. We hope your project is feasible and appears in the community soon.

    user-icon
    Eric Duneau
    May 13, 2024 | 1:14 PM
    Project Owner

    Hi Victor - thank you very much for your time to send feedback. I understand that at first, it may seem unlikely that one person alone can deliver a full system, but I think I have proven so many times now, in the past 3-4 years, that this is what I do. You can check the following projects, which i have developped alone from start to finish: (i) authenly.com ; (ii) SignWithWallet.com ; (iii) opensourceAIs.com ; (iv) shortfakes.com. In addition, I have already done many proof of concepts with digital identity, a lot of knowledge which i will reuse to pass this project. You will also see that i have set the project on a "long" timeframe (8 months to deliver a 6 months plan forecast), to lower the risk of a late delivery. Finally, unlike many others, this project does not have complex dependencies, which could otherwise cause delay outside of my control.  

    Thank you again, and I hope that you will supoprt this project.

    Kind regards - Eric

  • 0
    user-icon
    TrucTrixie
    May 10, 2024 | 11:20 AM

    Overall

    5

    • Feasibility 5
    • Viability 4
    • Desirabilty 5
    • Usefulness 5
    Time factor

    A very good suggestion regarding Digital Identity. Surely the world will need proposals like this more and more in the future. The SNET ecosystem is in need of proposals like this. Everything is commendable, including from the feasibility and benefits that the proposal brings, to the viability and the author's desire expressed in the proposal. However, I haven't seen what time period the author intends to take, please let me know about the duration or timing factor so I can evaluate more perfectly.
    I respect the author very much, because he is a high-quality and reputable person in the high-tech world for a long time.

    user-icon
    Eric Duneau
    May 10, 2024 | 3:30 PM
    Project Owner

    Hello TrucTrixie - thank you for your feedback, and I appreciate very much your compliments. Thank you also for pointing out that the duration of the project was not clear. It is great to have community feedback such as yours, as it was a total oversight on my part. I have updated the pitch to reflect this. I estimate that it will be deliverd in 26 weeks (6 months), but I would prefer to set the expectation to 8 months and have the opportunity to deliver earlier.

  • 0
    user-icon
    Viclex Ad
    May 6, 2024 | 12:01 AM

    Overall

    4

    • Feasibility 4
    • Viability 3
    • Desirabilty 5
    • Usefulness 4
    Proposal for Digital Identity Trust Framework

    Practicality;

    By utilizing already-existing resources like Hyperledger Identus and OpenSourceAIs, the concept signals viability while addressing a crucial issue of AI trust. The intricacy of AI trust frameworks, however, may present difficulties.


    Viability:

    The experience and prior work of the team demonstrate their competency. However, more funding may be required for effective implementation and post-deployment assistance.

    Desirability:

    In a world where AI is pervasive, trust is essential. The present requirement for verifiable AI identities and reputation systems is exactly where this solution fits in.

    Practicality:

    By resolving trust issues, a successful deployment would significantly increase SingularityNET's allure and draw in additional users and AIs.

    Important Success Factors

    1. Robust Governance: Ensure clear decision-making structures for agile adjustments during development.
    2. Continuous Feedback: Regular community engagement and feedback loops for iterative improvements.
    3. Scalable Infrastructure: Plan for scalability as user and AI adoption increases.
    4. Security Measures: Implement stringent security protocols to safeguard digital identities and credentials.

    Important Recommendations for Effective Delivery

    1. Collaborative Development: Engage with relevant stakeholders and AI developers for comprehensive insights.
    2. Comprehensive Documentation: Thorough documentation aids adoption and troubleshooting.
    3. Agile Methodology: Utilize agile practices for flexibility and responsiveness.
    4. Post-Deployment Support: Allocate resources for ongoing maintenance, updates, and user support.

    user-icon
    Eric Duneau
    May 6, 2024 | 6:35 PM
    Project Owner

    Hi VicLex - Thank you, although this looks like a ChatGPT kind of summary reply, and therefore I am not sure what actionable step you expect me to do or answer there. You seem to think the project is not completely viable (3/5) but your text seems to signal that it is viable.

    Kind regards

    Eric 

  • 0
    user-icon
    BlackCoffee
    May 12, 2024 | 9:19 AM

    Overall

    4

    • Feasibility 4
    • Viability 3
    • Desirabilty 4
    • Usefulness 5
    Be specific with data

    Tôi rất thích việc xuất đề thành công khi áp dụng vào công cụ thực tế được hóa học bằng dữ liệu. Bạn có thể làm được điều đó không? Bên bờ lý thuyết rất hay thì những dữ liệu đi kèm cũng rất thú vị. Biết đâu đó cũng chính là mục tiêu cần đạt được trong thực tế. Chúc các bạn và các thành viên trong nhóm (nếu có) may mắn và thành công.

    Hello Eric!
    Of course I am a real person, not an AI, of course I am the one who wrote these opinions.
    I'm sorry I didn't express myself clearly so you couldn't understand the content. What I mean is that you can try to specify what needs to be done when implementing this proposal with specific numbers, such as set targets. Thereby, the community has more basis to evaluate the success of the proposal.
    Thanks Eric.

     

     

    user-icon
    Eric Duneau
    May 6, 2024 | 6:29 PM
    Project Owner

    Hey BlackCoffee - I fankly do not understand well your comment. Hopefully this is no AI generated...

    Kind regards

    Eric

  • 0
    user-icon
    Max1524
    May 1, 2024 | 2:27 PM

    Overall

    4

    • Feasibility 4
    • Viability 4
    • Desirabilty 4
    • Usefulness 4
    Open source management

    The proposal that made the biggest impression on the author was Eric Duneau, a reputable and successful individual. The "Trust and Reputation Framework for AI" led by Eric is worthy. However, I have a question about how to manage open source if the proposal is successfully applied in practice in the future? Perhaps the author should have a separate section to present this.

    user-icon
    Eric Duneau
    May 2, 2024 | 11:02 AM
    Project Owner

    Hey Max1524 - thank you for your feedback.

    This is not a one-time project for me, as I intend to continue building opensourceAIs and its decentralised identity component for the next few years.I will try to add a section in the proposal to cover your question. However, I think the low hanging fruit here is that in case I deliver the code but do not continue to maintain it, anyone in the community could fork it and continue on.  Note also that I am mostly building on either open source, or mainstream components, so that the future solution does not become locked in a proprietary code. 

    Please do not hesitate to ask if you have any additional specific question.

    kind regards - Eric

  • 0
    user-icon
    Joseph Gastoni
    Apr 30, 2024 | 3:20 PM

    Overall

    5

    • Feasibility 5
    • Viability 4
    • Desirabilty 4
    • Usefulness 5
    It has a strong foundation but faces challenges

    The project has a strong foundation but faces challenges in achieving widespread adoption and managing the complexity of open-source development.

    Feasibility:

    • Technical Expertise: Developing a secure and reliable decentralized identity framework requires expertise in blockchain technology, cryptography, and distributed systems.
    • Open-Source Development: Building and maintaining a successful open-source project requires a dedicated team and community engagement.

    Viability:

    • Market Adoption: Convincing AI developers and users to adopt a new standard for AI reputation requires strong community outreach and education.
    • Integration Complexity: Integrating the identity framework with existing AI services might be complex for developers.

    Desirability:

    • Trust and Transparency: The focus on establishing trust and transparency in the SingularityNET ecosystem is highly desirable.
    • Open-Source Approach: The open-source approach encourages community participation and innovation.

    Usefulness:

    • Improved Security: Verifying AI identities can help prevent malicious actors from entering the ecosystem.
    • Enhanced Reputation Management: Tracking AI performance can help users choose reliable services.

    Besides, the project should consider:

    • Phased Development: Start with a minimal viable product (MVP) focusing on core functionalities like digital identity issuance and verification.
    • Community Building: Actively engage with the SingularityNET community to gather feedback and build a strong developer base.
    • Standardization Efforts: Collaborate with other stakeholders in the AI industry to establish wider adoption of the identity framework.
    • Developer Tools and Resources: Provide clear documentation, code samples, and developer tools to simplify integration with existing AI services.

    By addressing these points and focusing on building a strong community, this project has the potential to become a critical component of the SingularityNET ecosystem, promoting trust and security in the use of AI services.

    user-icon
    Eric Duneau
    May 6, 2024 | 6:26 PM
    Project Owner

    Hey Joseph - thank you for your feedback. I hope you saw the mislestone approach to designing and delivering the solution. Also, this is not my first attempt at delivering either open source project, or digital identity project. I agree that the community aspect is a challenge that goes beyod just the development, ad at this point, I have not fully addressed it in the proposal, which is more a technical, development & delivery proposal.

    Please do not hesitate to ask if you have any specific question.

    kind regards - Eric

  • 0
    user-icon
    Aokishi
    Apr 30, 2024 | 6:05 AM

    Overall

    5

    • Feasibility 5
    • Viability 5
    • Desirabilty 5
    • Usefulness 5
    A project worth looking forward to

    When the AI industry expands to a certain extent, an AI marketplace will emerge where users pay for AI services. Therefore, it becomes essential to verify the performance of AI before deciding to purchase any AI service. This is quite an interesting project where AI users can find AI that suits their needs based on Digital Identity. AI sellers and service providers can also evaluate and price AI based on its performance. Eric Duneau is the only core member of the project. According to his biography, Eric is an experienced individual in the software industry. Eric's career in the AI industry only started in 2019 but already has significant achievements such as creating OpenSourceAIs. With high feasibility and great vision, I believe this is a project worth looking forward to.

    user-icon
    Eric Duneau
    May 2, 2024 | 10:50 AM
    Project Owner

    Hey Aokishi - thank you for your feedback. By the way, although I did not mention it here, mventures into A.I. did not start in 2019, as I hold a M.Sc. in Artificial Intelligence, but this is dating 1992, a little bit back in time ;-) 

    Please do not hesitate to ask if you have any specific question.

    kind regards - Eric

  • 0
    user-icon
    Wei Dong
    Apr 29, 2024 | 10:59 PM

    Overall

    4

    • Feasibility 4
    • Viability 4
    • Desirabilty 5
    • Usefulness 4
    New Trust and Reputation Framework in AI

    The Trust and Reputation Framework for AIs is a forward-thinking solution addressing the pressing issues of trust and security in the AI ecosystem. Despite potential challenges in feasibility and market viability, its desirability and usefulness score high, reflecting the critical need for such innovations in the expanding landscape of artificial intelligence.

    user-icon
    Eric Duneau
    May 2, 2024 | 10:46 AM
    Project Owner

    Hey Wei - thank you for your feedback. Please do not hesitate to ask if you have any specific question.

    kind regards - Eric

Summary

Overall Community

4.5

from 8 reviews
  • 5
    4
  • 4
    4
  • 3
    0
  • 2
    0
  • 1
    0

Feasibility

4.5

from 8 reviews

Viability

3.9

from 8 reviews

Desirabilty

4.6

from 8 reviews

Usefulness

4.6

from 8 reviews