AI Fast Break: Storm Chaser! Using Agentic AI for Basic Research
Image: created in DALL·E3 (fall semester 2023)
GENERAL INTEREST
STORM
What is STORM? “[A] writing system for the Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking. STORM models the pre-writing stage by (1) discovering diverse perspectives in researching the given topic, (2) simulating conversations where writers carrying different perspectives pose questions to a topic expert grounded on trusted Internet sources, (3) curating the collected information to create an outline.” To learn more, read the developers’ research paper, “Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models.”
CREATE AN ARTICLE
In STORM (article generation) mode, users enter a prompt, and STORM responds by sharing an estimated time to task completion, the processing step it is currently on, and a request for the writing’s purpose.
Figure 1 The first stages of STORM's prompt processing sharing.
STORM generates:
A complete article with a summary and hyperlinked in-text citations
The BrainSTORMING Process: the question-and-answer interactions of multiple agents that created the context for drafting the article
The expertise of each agent is explained along with their responses to a moderator’s prompt-related questions
Figure 2 The BrainSTORMING responses of four agents.
Figure 3 The BrainSTORMING responses of four agents.
An interactive Table of Contents (accessed by clicking the dark page icon with a blue dot on the left-hand side of the page)
clicking the icon opens the interactive menu on the left-hand side of the screen
clicking the three horizontal lines next to the STORM dropdown menu collapses the Table of Contents
A feedback form (found at the article’s conclusion)
A PDF version of the article (available via the PDF icon in the lower right-hand corner)
the PDF can be downloaded or printed
Figure 4 The STORM results page showing the article, access to the Table of Contents, and the PDF button.
Sources are referenced by number or via a citation (see below). Clicking on the citation number opens a pop-up box sharing the reference title, URL, and highlight summary.
Figure 5 The reference pop-up box sharing the reference title, URL, and highlight summary.
CO-STORM
In addition to STORM, users can select “Co-STORM” from the dropdown menu. Co-STORM lets users engage in a topic-centered discussion (aka “Roundtable Conversation”) with STORM AI agents.
As with STORM, Co-STORM users are asked to share their writing’s purpose
The double discussion bubbles icon represents the conversational mode, to switch back to the standard article mode, click on the stack of papers icon
Figure 6 The Roundtable Conversation response page.
“See Topic Background Discussions” shares the exchanges between a “Background discussion moderator” and a “Background discussion expert” as they develop an iterated informational context for responding to the prompt
Users can join a roundtable conversation by typing questions, comments, or feedback in the chatbox
They can also extend the conversation by clicking on the “Generate” button associated with an included agent
Responses include in-text citations; clicking on the number triggers a pop-up window including the reference’s title, URL, and a content summary
Figure 7 The Roundtable Conversation response page.
Figure 8 The ability of a user to join the AI agent conversation or extend it by using the "Generate" button.
Figure 9 The "See Topic Background Discussions" page.
Figure 10 A user's engagement in the roundtable discussion.
STORM CONSENT FORM
DESCRIPTION
You are invited to try out a research preview of our NAACL 2024 paper and EMNLP 2024 paper. To use it, you need to first agree with our “Terms of Service” displayed on the web demo and verify you are a real human user by logging in your Google account. On our web demo, you can input the topic you want to learn in depth and your purpose of researching this topic. You can also input questions to our system, and our system will retrieve additional information, update the hierarchical outline and references, and provide a synthesized response. Based on your input, our system will generate a report with hierarchical outline and references. You can read the report on our web demo. If you would like to, you can provide feedback of the generated report using the feedback box on our web demo. Your input (input topic, purpose of writing the article, and follow-up questions) and feedback (if provided) will be securely stored associated with the report generated by our system. Your Google account information will only be used to maintain your login status and will not be combined with data we collected.
USER'S RIGHTS
If you have decided to try out our research preview, please understand you have the right to stop using it at any time. The results of this research study may be presented at scientific or professional meetings or published in scientific journals. Your individual privacy will be maintained in all published and written data resulting from the study. For individuals who prefer not to have their data collected and shared, you may instead use our open-source software available at https://github.com/stanford-oval/storm. For organizations with concerns, please feel free to reach out to us at genie@cs.stanford.edu.
POTENTIAL RISKS
The risks associated with this study are minimal. Study data will be stored securely, in compliance with Stanford University standards, minimizing the risk of confidentiality breach.
CONTACT INFORMATION
If you have any questions, concerns or complaints about this research, its procedures, risks and benefits, contact the Protocol Director, Yijia Shao - (650) 407-9690 - shaoyj@stanford.edu.
Independent Contact
If you are not satisfied with how this study is being conducted, or if you have any concerns, complaints, or general questions about the research or your rights as a participant, please contact the Stanford Institutional Review Board (IRB) to speak to someone independent of the research team at (650)-723-2480 or toll free at 1-866-680-2906, or email at irbnonmed@stanford.edu. You can also write to the Stanford IRB, Stanford University, 1705 El Camino Real, Palo Alto, CA 94306.
Please print or save a copy of this page for your records.