ChatDKG - an open framework for trusted AI based on OriginTrail DKG 🚀
According to Goldman Sachs Chief Information Officer, Marco Argenti, "the impact of advances in generative artificial intelligence on society could be comparable to the printing press" and with over 91% of top businesses currently investing in AI (NewVantage, 2022), those opportunities might be seized at a very rapid pace. However, with the internet already flooded with misinformation, AI is likely to amplify the problem.
Two major challenges with AI models such as OpenAI's ChatGPT are:
- Hallucinations or errors in AI-generated outputs,
- the ability to verify the provenance of data that was used to generate those outputs.
Using such outputs could lead to incorrect decisions/actions and the rise of bias, misinformation, and propaganda - all of which could have severe consequences for individuals and society. Therefore there is a clear need for AI systems to be rooted in verifiable data sources to mitigate those risks and ensure they reach their full potential.
What is ChatDKG?
ChatDKG is an open framework for efficient search, access, and utilization of trusted knowledge with AI, based on OriginTrail Decentralized Knowledge Graph (DKG) and Knowledge Assets. The ChatDKG framework is designed as an open community project to explore a diverse range of possible applications.
The key objectives of ChatDKG are to:
- Create technical tools based on Knowledge Assets and AI
- Establish best practices for discovering, accessing, querying and verifying trusted knowledge with AI tools
- Generate a trusted, open knowledge base by creating Knowledge Assets through crowdsourcing and iterations on OriginTrail DKG Mainnet
1 million TRAC grant pool available for ChatDKG builders
For contributors that want to get involved and support the development of the ChatDKG framework a 1 million TRAC token grant pool was made available by Trace Labs. There are three levels of grants available, depending on a project’s potential:
- 🐰 Rabbit level: 25k TRAC
- 🦅 Eagle level: 50k TRAC
- 🔥 Phoenix level: 100k TRAC
The potential of each project is evaluated by the Steering committee during the application process, based on three categories:
- Utility of OriginTrail DKG and Knowledge Assets for AI systems
- Breadth (widely vs narrowly applicable)
- Scalability of project components
There are four key challenge areas that you can apply for:
- Semantic search over Knowledge Assets (e.g. Knowledge asset vector similarity search tools)
- Natural language querying over OriginTrail DKG (e.g. GPT-based SPARQL query generator)
- AI-based Knowledge publishing tools (e.g. Agents for KA semantic enrichment)
- General AI system integrations (e.g. integrations with Huggingface, BARD, Langchain, etc.)
- Other ideas relevant to the application of AI and DKG (Knowledge Assets)
How do I start?
Applications to the ChatDKG grant program are done on a rolling basis, meaning you can apply at any time until the 1 million TRAC grant pool has been fully distributed.
Your project needs to meet the following criteria for a successful application to the ChatDKG grant program:
- Public, open-source project
- Has to apply to OriginTrail DKG and Knowledge Assets
- Has to solve a meaningful problem
- Active development & engagement in ChatDKG Discord
How does the grant process look like?
🏅 Step 1: Apply with your idea
- Open a GitHub repo for your ChatDKG idea with your project description, milestones, timeline, and expected impact
- Fill in the ChatDKG grant application form here
- Review your idea with the Steering committee
- Get your application and grant level approved, and collect 15% of your grant amount
💻 Step 2: Work on your idea
- Milestone review with the Steering committee (according to the milestones you defined in your project plan)
- Active development through the ChatDKG Discord channel for ongoing support and knowledge exchange
- Ad-hoc guidance from Steering Committee members if required
🌟 Step 3: Submit your results
- Submit your final deliverables when ready
- Deliverables should be in the form of code in your GitHub repo
- Review and demo with the Steering committee
- Demo needs to be done with Knowledge Assets on OriginTrail DKG mainnet
- Public presentation of your project and demo in the Open Office Hours
- Collect the rest of your grant upon successful review
The Steering committee for ChatDKG is comprised of 5 members with the following responsibilities:
- Guiding ChatDKG development
- Supporting ChatDKG builders: advising, grant application selection, support in project execution
- Facilitating discussions through the ChatDKG channels
Steering committee members:
- Oshani Seneviratne
- Assistant Professor at Rensselaer Polytechnic Institute (RPI), NY
- Ph.D. & S.M. in Computer Science, MIT (under the supervision of Sir Tim Berners-Lee, the inventor of the World Wide Web)
- Primary Research Focus: Decentralized Systems
- More SC committee members TBA
Documentation and decks
- DKG Copilot
- DKG Copilot is designed to help users retrieve, analyze, and manage information about scientific papers from the OriginTrail Decentralized Knowledge Graph (DKG). Users can ask questions in natural language about scientific papers that can be found in the DKG by generating SPARQL queries based on an ontology.
- DKG Azure Function API
- Truthchain Extension repository
- Chrome Extension
- Truthchain is an open-source project dedicated to combating misinformation and promoting truth in an increasingly digital and interconnected world. By harnessing the power of blockchain technology, data verifiers, and user contributions, Truthchain aims to create a decentralized platform where users can collect, verify, and store data, ensuring its authenticity and credibility. This will bring to the forefront quality information, with reputable sources, as a way to build on shared knowledge and individual insights.
- OTHub seeks to incorporate and augment the hallmark features that distinguished OTHub, paving the way for broader adoption. It not only continues to provide a comprehensive overview of the Decentralized Knowledge Graph V6 network statistics to all participants but also introduces bespoke dashboards for asset publishers and DKG network stakeholders. The main focus is to enhance accessibility, efficiency, and user-friendliness for the publication of new knowledge assets on the ecosystem. Expanding OTHub's scope, the platform will serve as the foundational layer - the OTHub portal, enabling other projects and individuals to seamlessly get, update, and publish knowledge assets via OTHub-api and OTHub-bot.
- NatLangKG aims to harness the capabilities of advanced language models to create a natural language interface for the OriginTrail Decentralized Knowledge Graph (DKG). The project plans to empower non-technical users to query Knowledge Assets with ease, transforming natural language input into SPARQL and providing understandable natural language outputs. The initiative promotes a broader adoption of the DKG by making it more user-friendly, while also contributing to the democratization and trustworthiness of AI by rooting the model in a verifiable data source. To facilitate integration with other projects and maximize synergy, the project will utilize existing transformer-based text embeddings.
- PolyCode Frontend
- PolyCode SDK
- PolyCode Documentation
- Dev3 is a blockchain low-code development and business process automation framework. It’s enterprise-grade, yet simple to use. It can scale from small individual projects to millions of daily users. Whatever you are building in blockchain – Dev3 makes it easier and more secure! All Dev3 apps have amazing user experience right out of the box from years of research and development on how to provide usable blockchain software to regular, non-technical users.
ChatDKG - Open Office hours videos
ChatDKG Open office hours are organized on a frequent basis and are by default recorded. All office hours videos will be posted here. Everyone interested in ChatDKG topics is invited and welcome to join the Open Office Hours - join the ChatDKG Discord or follow us on Twitter to stay informed for the next editions.
Join the ChatDKG community
ChatDKG framework is supported by NGI Search.