Transform scientific tasks into solutions
Define your task
Get a solution
Report
Use case: Solve complex research challenges effectively
Designed for you: As research scientists, we share a passion for exploring beyond the known. However, even simple scientific tasks often require extensive studying, analyzing, coding, testing, and more, typically taking much longer than expected. That’s where RASSAI comes in.
Fly with RASSAI: Provide a task description and relevant materials like books, papers, or code repositories, and let RASSAI make your research fly by:
- suggesting the best action at each step
- extracting relevant knowledge typically scattered across the materials
- suggesting available strategies when multiple options are available
- breaking down complex tasks into into manageable steps and assisting in their resolution
- integrating all parts and reporting your results along with your decision process
Example: Pearson's rho conveniently measures the strength of dependence between two variables when their distribution is Gaussian. However, in fields like finance and environmental science, data often show extreme dependencies and heavy tails, deviating from the Gaussian model. In these cases, Gumbel copulas are effective alternatives, capturing complex dependence structures and tail dependencies.
Task: Generating samples from these models is crucial, as the Monte Carlo approach is often used to compute quantities that are analytically difficult to determine.
Solution: Providing RASSAI with relevant books and papers, we access the solution within minutes, including a detailed technical report.
Use Case: Step 1
Manage your tasks
Outline your research task with support for LaTeX formatting.
Active Mode lets you engage with RASSAI to conduct your research actively.
View Mode allows you to review all your previous steps from Active Mode and share your progress with others. Whether it's for:
- Colleagues or teammates: to report completed work
- Students: for didactic reasons
- Project Lead: to highlight progress in broader research initiatives
Use Case: Step 2
Plan the next step
RASSAI assists you in planning the next action for task resolution by ranking available actions based on their suitability and providing detailed descriptions and reasoning for each option.
Retrieve acquires more knowledge from reliable sources.
Ask enables RASSAI to ask you for clarification if the task description is unclear.
Decompose approaches complex tasks hierarchically.
Strategize chooses the best method for task resolution when multiple options are available.
Complete turns granular tasks into solutions.
Use Case: Step 3
Retrieve
RASSAI searches reliable knowledge sources and gathers relevant information, providing Retrieved Knowledge Extracts (RKEs) (right panel).
Each RKE is linked to its source document and paragraph for validation (e.g. 2_13 or 4_463).
You approve or reject each RKE using the check box. This step mitigates AI model hallucinations and avoids confusion from irrelevant knowledge.
All your previous decisions and actions are appended to the task representation (left panel).
Use Case: Step 4
Plan the next step after retrieval
Once an action has been finished, RASSAI assists you in planning the next step.
After retrieval, the task representation is augmented by:
- AI-generated RKEs approved by the user (bottom).
- Knowledge from reliable sources, Retrieved Knowledge Sources (RKS) (top), identified by ISBN or DOI.
Based on the new knowledge, RASSAI typically reranks the available actions.
Use Case: Step 5
Complete
The synergy between you, AI, and the provided knowledge sources via RASSAI comes to fruition: The task transforms into a solution draft.
The solution draft follows RASSAI's internal format:
- An algorithm composed of a set of instructions (1-4).
- Each instruction has a description (green background), with the color indicating the AI model's confidence in its correctness.
- If the model lacks knowledge required to complete an instruction, this is reported.
If code is part of the solution, it is also provided. Notice that, as it closely follows the draft, it is not optimized for speed (vectorized). However, this requirement was not a part of the task description. But it could be additionaly vectorized, which is rather a routing task for AI models.
Expert's highlight: What is far from being a routine task for AI models is solving this task correctly. An important observation here is that RASSAI effectively merged knowledge from a book (978...), a paper (10.1016...), and a knowledge extract (inverse...), resulting in a correct solution. With any of these parts missing, the draft fails to be a correct solution. We believe that this is the future of AI-assisted research, showcasing how RASSAI seamlessly integrates diverse knowledge sources to deliver accurate solutions quickly and effectively.
Finally, you can either approve the solution draft or regenerate it based on your expertise.
Notes:
- While this example reflects our expertise, RASSAI is task-agnostic and can be used for any scientific research area.
- OpenAI's GPT-4o failed to solve this task correctly in 100 attempts when prompted by the task description.
- All presented screenshots are directly from RASSAI, unedited. In this example, the user only steered RASSAI and oversaw its intermediate outputs, without any textual inputs.
Use Case: Final Step
Report
An approved draft is considered a solution to the task, and a research report can be generated.
Since RASSAI has observed all your decisions, it can include all gathered information:
- Retrieved knowledge sources and extracts
- References
- Code or visualizations based on the code (Figure 1)
Summary
Stainding on the synergy of your expertise, AI, and reliable knowledge sources, RASSAI effectively assists you in solving complex research challenges. Key features include:
- Task Agnosticism: RASSAI can handle tasks from any scientific area.
- Efficient Knowledge Generation: Generates new knowledge efficiently based on refining provided knowledge sources.
- AI-Assisted Task Resolution: Guides you through the task resolution process with state-of-the-art AI assistance.
- Comprehensive Reporting: Observes and includes all decisions and gathered information, generating detailed research reports.
- Interactive Planning: Assists in planning and executing each step, ensuring a structured approach to complex tasks.
Standing on the synergy
Human Intelligence
Role: As the human expert in RASSAI's research pipeline, you are the ultimate decision-maker, fully responsible for the outcomes. You guide the research, manage each step, and tackle challenges that AI cannot handle.
Advantages: Humans bring unique strengths perfected over billions of years of evolution:
- Deep understanding of the physical world
- Persistent and reliable memory
- Advanced reasoning capabilities
- Ability to plan and execute complex, hierarchical tasks
- Intuition and common sense
Weaknesses: Despite their strengths, humans have limitations:
- Slower processing speed compared to AI
- Prone to errors and biases
- Expertise often limited to specific areas
- Restricted by physical and mental endurance
Synergy: The efficiency and tirelessness of AI compensate for human limitations in speed and endurance, while the vast data resources and AI tools assist in managing information overload.
Artificial Intelligence
Role: The AI model serves as your assistant, providing suggestions and aiding in routine tasks like literature surveys, coding, comments, and test design
Advantages: The AI model offers speed and efficiency, vast knowledge across numerous fields, remarkable language processing capabilities, and never gets tired.
Weaknesses: Despite its strengths, the AI model has limitations. According to Meta's Chief AI Scientist Yann LeCun, essential characteristics necessary for intelligent behavior, such as deep understanding, common sense, and contextual reasoning, are still missing in even the best current AI models. The AI can produce incorrect or nonsensical outputs ("hallucinations") and may struggle with tasks requiring deep contextual knowledge.
Synergy: The human expert’s advanced reasoning, intuition, and common sense compensate for the AI’s lack of deep understanding, while reliable knowledge sources helps guide the AI’s outputs towards accuracy.
Reliable Knowledge Sources
Role: Reliable knowledge sources serve as the foundation of your research, providing you, the human expert, with insights and wisdom to solve complex tasks.
Advantages: Reliable knowledge sources offer the accumulated wisdom of the brightest minds over the years, allowing you to stand on the shoulders of giants. Sources such as scholarly papers, reputable parts of the internet, and comprehensive databases like Wikipedia provide a vast amount of knowledge, grounding your research in well-established information.
Weaknesses: The vast amount of information can be overwhelming, requiring AI tools to quickly digest and identify relevant details.
Synergy: The AI's capability to sift through vast amounts of information and the human expert’s ability to interpret and apply this information effectively combine to overcome information overload challenges.
Designed by research scientists, for research scientists
Collaboration: RASSAI is designed to work with you during research, not to replace you. Being a research scientist, you love to explore, and there’s no benefit in being replaced by AI. Instead, RASSAI frees you from routine tasks by transferring them to AI and engages your unique human strengths that AI lacks, empowering you to achieve more.
Organization: Research can often be chaotic, with papers, code, and notes scattered across various places, slowing progress. RASSAI organizes your research in a structured and hierarchical way. Every decision is supported by AI suggestions and reliable knowledge sources, significantly accelerating your research process.
Cutting-edge Technology: The AI area is rapidly evolving, with numerous models and techniques emerging constantly. It's challenging for researchers to keep up. RASSAI handles these difficulties for you, providing the best available AI models and integrating the most promising techniques. This allows you to focus solely on your research.