November 25, 2024
Introducing AI Research Engine (RE0): Technology Preview
Today, we are proud to unveil the AI Research Engine (RE0), an innovative platform designed to revolutionize how research ideas are generated and developed. Currently in Technology Preview, this system combines the power of cutting-edge AI with the richness of lived experience and community insights to propose groundbreaking research ideas.
RE0 stands apart in its ability to synthesize multiple streams of input—scientific literature, podcast interviews, user-generated questions, and structured community feedback—into well-crafted, actionable research proposals. It moves beyond traditional AI tools that summarize and retrieve, offering a deeply collaborative and idea-driven approach to research generation.
Our initial use case focuses on misophonia, leveraging extensive podcast interview archives and lived-experience data to propose fresh directions for research. However, this is just the beginning. The system is modular and scalable, ready to address complex challenges in countless fields.
You can listen to a discussion about the platform on a special episode of the podcast. And below, you’ll find actual sample generated proposals as well as a deep dive into how RE0 works. Sign up to learn more about the platform, give feedback, and see a demo.
Sample Proposals | ||
---|---|---|
Adverse Childhood Experiences and Misophonia: Identifying Audiological and Psychological Factors Modulating Severity | Full Proposal | Student Project |
Effects of Remote Work on Misophonia Triggers and Coping Mechanisms | Full Proposal | Student Project |
Epigenetic Markers and Transgenerational Effects in Misophonia | Full Proposal | Student Project |
Exploring Neurological and Physiological Overlaps in Misophonia and ADHD: A Neuroimaging Study | Full Proposal | Student Project |
Exploring the Cross-Modal Triggers for Misophonia | Full Proposal | Student Project |
Exploring the Influence of Community Noise Policies on Misophonia and Resident Well-being | Full Proposal | Student Project |
The Role of Virtual Soundscapes in Mitigating Misophonic Responses During Quarantine | Full Proposal | Student Project |
Validation Study on Decreased Sound Tolerance in Tinnitus Patients: Emphasizing Psychological and Audiological Factors in Misophonia Research | Full Proposal | Student Project |
How It Works
Stage 1: The Data Stage – Aggregating Knowledge Across Boundaries
The journey begins with the Data Stage, where diverse sources of information are collected, processed, and indexed to form the foundational knowledge base. This stage ensures the system has access to the most comprehensive and up-to-date resources available.
Diverse Data Inputs
The system taps into multiple knowledge streams:
- Scientific Literature: With access to over 220 million articles and papers, the system continuously retrieves the latest publicly available research in real time. PDFs, metadata, and abstracts are automatically fetched and processed.
- Podcast Interviews: Transcripts from podcasts, such as The Misophonia Podcast, are processed using cutting-edge transcription tools. These conversations often reveal nuanced, lived experiences that are missing from traditional research.
- Community Questions and Input: Insights and recurring themes from community discussions provide unique, ground-level perspectives that guide future exploration.
Processing and Summarization
Once collected, the data undergoes extraction and processing:
- PDFs are parsed for relevant text, which is then summarized using advanced large language models (LLMs).
- Summaries condense complex research papers and hours of interviews into clear, accessible insights.
Data Storage and Retrieval
All processed data—raw and summarized—is indexed and stored across a robust infrastructure, combining:
- Postgres for structured storage.
- Elasticsearch for fast and flexible search capabilities.
- Vector Databases for embedding-based similarity search.
Tracking Sources for Transparency
To maintain credibility and reduce hallucinations, we track the sources of every piece of information throughout the entire process. This allows users to inspect citations, verify claims, and evaluate the system’s reliability, ensuring that the research remains grounded and trustworthy.
Stage 2: The Rumination Stage – A Think Tank for Ideas
The Rumination Stage is where the system begins to think like a researcher, synthesizing diverse inputs into preliminary research ideas. This stage mimics the creative process of brainstorming, guided by human oversight.
Strategic Inquiry
The system employs strategies inspired by scientific inquiry, blending:
- Open-ended ad hoc questions submitted by researchers.
- Structured themes derived from community discussions or podcast content.
- Data-driven prompts generated from trends in the stored knowledge base.
Idea Generation
Using these inputs, the system identifies potential research directions. Rather than simply retrieving information, it builds hypotheses and frameworks, ensuring the suggestions are not only relevant but also novel.
Human Collaboration
At this stage, human input plays a critical role. Researchers review the system’s initial ideas, selecting the most promising ones to develop further. This collaborative refinement ensures that the final proposals are both innovative and practical.
Stage 3: The Generation Stage – From Ideas to Detailed Proposals
With the selected ideas in hand, the system transitions to the Generation Stage, crafting detailed research proposals designed to inspire action and inform decision-making.
Comprehensive Outlines
The system begins by creating detailed outlines, including:
- A clear abstract summarizing the research focus.
- A literature review connecting the new proposal to existing knowledge.
- A methodology outlining how the research will be conducted.
- A recruitment plan to identify participants, if needed.
- A timeline to guide implementation.
Proposal Expansion
Each outline is then expanded into a fully fleshed-out proposal. These proposals are designed to meet the rigorous standards of academic publishing while remaining accessible to non-experts.
Stage 4: The Presentation Stage – Tailoring Insights for Every Audience
The final step is the Presentation Stage, where RE0 formats its research proposals to resonate with distinct audiences. This stage is not just about communicating results—it’s about ensuring that the outputs are relevant and actionable for researchers, curious youth, and the general public alike.
Academic-Level Outputs
For researchers and academics, RE0 generates detailed proposals that include comprehensive methodologies, literature reviews, and citations. While these outputs may not yet match the depth of proposals crafted entirely by human experts, they serve as strong starting points—offering inspiration and laying the groundwork for further refinement. With collaboration, these proposals can evolve into submission-ready documents for journals, conferences, or institutional review boards.
High School Science Proposals
For students and educators, RE0 creates simplified research plans suitable for high school science projects. These proposals maintain scientific rigor while being achievable within school environments, fostering engagement with meaningful research. They are designed to inspire the next generation of researchers by making complex ideas accessible and actionable.
AI-Generated Audio Summaries
To bridge the gap between technical research and public understanding, RE0 produces conversational audio summaries. These AI-driven dialogues feature virtual personalities discussing the key aspects of the proposals, presenting complex ideas in a clear and engaging way. This format makes research accessible to a broader audience, sparking curiosity and fostering a deeper appreciation for science.
By tailoring outputs for these specific groups, RE0 ensures its research pipeline delivers value to the academic community, young scientists, and the general public—encouraging action, understanding, and inspiration at every level.
Why This Matters
RE0 represents a paradigm shift in research innovation. Here’s what sets it apart:
- Integration Across Data Sources: By blending traditional scientific literature with podcasts, community input, and AI-driven insights, the system captures a broader, richer perspective.
- Modularity: The platform is designed to be flexible, allowing users to replace or enhance individual components to suit their needs.
- Beyond Summarization: Unlike retrieval-augmented generation (RAG) tools, the system actively synthesizes new ideas and frameworks.
- Catalyst for Progress: Our ultimate goal is to inspire research that addresses critical challenges, particularly in areas shaped by lived experience, such as misophonia.
A Note on the Technology Preview
As part of the Technology Preview, RE0 is an evolving system. Some outputs may feel experimental or overly straightforward at this stage, but this is an intentional step in our development journey. By gathering feedback and iterating rapidly, we are laying the groundwork for a stable and polished release in early 2025.
Our roadmap includes exciting enhancements:
- Improved citation validation and provenance tracking systems.
- Deeper integration of community insights and multimedia sources.
- Expanded tools to support the entire research lifecycle, moving beyond proposal generation.
Be Part of the Future of Research
We’re working with forward-thinking partners to help refine RE0’s capabilities and explore its potential. By joining the Technology Preview, you’ll play a key role in shaping this groundbreaking system, ensuring it meets the needs of researchers, educators, and communities worldwide.
Stay tuned for updates, and if you’re interested in collaborating, contact us to learn more about how RE0 can transform your research process. Together, we can redefine what’s possible in the world of ideas.