Embedded Files
Welcome to the BeamPath AI Wiki! This is your go-to resource for understanding and utilizing Large Language Models (LLMs) like ChatGPT and how it intersects with medicine and radiation oncology. Here, you'll find a collection of guides, tutorials, and best practices, videos, and papers curated for your learning and exploration.
1. Getting Started with LLMs
A Very Gentle Introduction to Large Language Models without the Hype - Riedl (Medium) Link
How Large Language Models work - Stöffelbauer Link
How to Use ChatGPT - Beginner’s Guide Youtube tutorial Link
Primer on Large Language Models for Health 101 by Stanford AIMI Link
What Is ChatGPT Doing … and Why Does It Work? - Stephen Wolfram Link
How to use ChatGPT Plugins - Lifehacker Link
2. Prompt Engineering
Basics of Prompt Engineering: Learn how to craft effective prompts to get the best results from ChatGPT. Link
Video Tutorials:
3. Lectures on AI in Medicine
Educational Lectures:
AI Revolution in Healthcare - Nature Link
Radiation Oncology in the Age of AI - ACR Video Link
4. Dos and Don'ts When Using LLMs
Best Practices: Essential tips for effectively and safely using LLMs in a medical setting.
Do
Do
Report LLM Utilization in Publications: If you're using insights or content generated by an LLM in your research or publication, disclose this for transparency and ethical considerations.
Exercise Discretion with Internet-Enabled LLMs: When using models like ChatGPT-4, which can access real-time information, verify the accuracy and use them judiciously.
Use for Non-Confidential Tasks: Employ LLMs for brainstorming, drafting non-confidential documents, or generating educational content.
Leverage for Data Analysis: Utilize LLMs in analyzing and interpreting non-sensitive data, beneficial for research and statistical analysis.
Editorial Assistance in Research Papers: Utilize Generative AI for drafting, editing, and refining research papers, ensuring alignment with publication guidelines.
Summarization Aid in Literature Reviews: Use Generative AI for summarizing literature but verify the accuracy against original sources.
Writing Code and Data Analysis: Apply Generative AI for code generation in data analysis tasks, ensuring to verify the output for accuracy.
Non-Technical Summaries and Translations: Use Generative AI to create summaries or translations, while ensuring they accurately represent your work.
Selecting Appropriate AI Model: Choose a Generative AI model that best fits your research needs, considering factors like open-source availability, accuracy, and cost.
Don't
Don't
Avoid Using Personal Health Information (PHI): Never process PHI through any LLM due to non-compliance with HIPAA and risk of data breaches.
Refrain from Uploading Confidential Data: Avoid using LLMs for tasks involving confidential information, such as peer-reviewed papers or grant applications.
Do Not Use as a Reference Manager: LLMs are not designed to manage bibliographic databases or references.
Cautious Handling of Sensitive Data: Do not input text, data, or PDFs into LLMs that you are not comfortable with potentially being shared.
Avoid Creative Contributions in Papers: Refrain from using Generative AI for original idea generation in research papers to maintain academic integrity.
Grant Proposal and Paper Review: Do not use Generative AI for reviewing grant proposals or papers due to confidentiality concerns and lack of expert judgment.
AI as Co-Author: Never list Generative AI as a co-author in publications.
Beware of Bias and Security Risks: Always be vigilant about potential biases in AI outputs and aware of security risks in their usage, particularly in research contexts.
5. Additional Resources
List of LLMs available for use:
ChatGPT by Open AI
Use case examples: Summarizing text, generating code for data wrangling, developing chatbots, assisting in task-oriented conversations.
Use case examples: Chatbot style search-engine.
Claude by Anthropic
Use case: Summarizing text, generating code for data wrangling, developing chatbots, assisting in task-oriented conversations.
Gemini by Google
Use case: Summarizing text, generating code for data wrangling, developing chatbots, assisting in task-oriented conversations.
Copilot by Microsoft
Use case: Assitant like chatbot, can help with summarizing text, generating code for data wrangling, developing chatbots, assisting in task-oriented conversations
Notebook LLM by Google (Beta)
Use case: Engaging with and analyzing multiple PDFs through conversation, interactive educational content, document summarization, data extraction from documents.
Please note that this list is not exhaustive, and new LLMs are being developed and released by various companies and organizations regularly.
General Videos:
Difference between ChatGPT 3.5 and ChatGPT 4 Link
Further Reading: Articles, papers, and websites for deepening your understanding of AI in oncology.
Papers related to LLM and Medicine, Oncology, and Radiation Oncology Paperpile folder
Disclosure: While LLMs offer numerous benefits, it's crucial to approach them with a critical eye. They are not infallible and can sometimes "hallucinate" or produce inaccurate information. Always exercise due diligence when using these tools. Additionally, it's important to remember that LLMs like ChatGPT are not HIPAA certified. Therefore, avoid uploading or sharing confidential or patient-specific information. For more information on best practices and responsible use, check our wiki.
Google Sites
Report abuse