Getting Started with the Stanford AI Playground: A Comprehensive Guide

The Stanford AI Playground offers a fantastic opportunity for faculty, staff, students, postdocs, and visiting scholars to safely explore and experiment with various AI models. This guide provides a quick start to help you navigate the platform and make the most of its features.

What is the Stanford AI Playground?

The AI Playground is a user-friendly platform built on open-source technologies. It allows users to try out AI models from vendors like OpenAI, Google, Anthropic, and Meta in a single, managed environment. University IT (UIT) manages this platform as a pilot program at Stanford.

AI Playground: Safety First

Before you dive in, it's crucial to understand the safety guidelines:

• Avoid High-Risk Data: Do not use high-risk data in prompts or attachments.
• LLMs are Not Flawless: Always exercise caution when using the results. These AI models may produce errors or hallucinations.
• Responsible AI: Familiarize yourself with Responsible AI guidance.

Quick Start Guide

Here's a step-by-step guide to get you started:

1. Open the Playground

• Visit aiplayground.stanford.edu.
• Log in with your Stanford Single Sign-On (SSO).
• If it's your first time, you may need to consent to Information Release settings. Select your preference and click "Accept." This data stays within Stanford systems. The FAQ section has information on "Data privacy and security in the AI Playground."

2. Try Out Prompts

• Type your prompt into the text field at the bottom of the welcome screen and press Enter.
• Refine your prompt based on the results. See guidance on crafting effective prompts.

Interacting with Messages:

• Read aloud: Hear the message read by a computer-generated voice.
• Edit: Modify your prompts or the AI responses.
• Save & Submit: Save your edits and regenerate the AI's response.
• Save: Save your edits without regenerating the response.
• Cancel: Close the edit window without saving changes.
• Copy to clipboard: Copy the message content.
• Regenerate: Force the model to create a new response.
• Fork: Start a new conversation from the selected message.

3. Adjust AI Models and Configuration Options

The AI Playground offers a variety of AI models to suit different needs. You can switch between these models at any point in a conversation.

Available Models:

• OpenAI ChatGPT Models:

  • gpt-4o - Best for complex reasoning, image and PDF analysis, as well as advanced coding; strengths include deep comprehension.
  • gpt-o1 - Good for academic purposes and complex math equations. Slower than GPT 4o.
  • gpt-o3-mini - Best for short questions needing quick answers. Optimized for coding, math, and science tasks.
  • gpt-4o-mini - Good for content creation, basic coding queries; strengths include speed and general use.

• Google Models:

  • gemini-2.0-flash - Best for fast performance and strong reasoning tasks. Excels at math problems and coding assistance.
  • gemini-1.5-pro - Best for advanced reasoning tasks, creative writing, detailed coding, and in-depth research. Great for understanding idioms and nuanced text from other languages. Strengths include high context limits, reasoning, and translation.
  • gemini-1.5-flash - Best for faster performance and quicker responses where speed is paramount. It trades off some of the Pro version's depth and complexity for increased efficiency. It's better for quick queries and snappy back-and-forths.

• Plugins:

  • DALL-E 3: (Not available to students or Visiting Scholars at this time.) Generates AI images from text prompts.
  • Google Imagen 3: (Not available to students or Visiting Scholars at this time.) Generates exceptional photorealistic images based on text descriptions.
  • Google: AI-assisted Google web search.
  • Web Scraper: Reads content from live webpages (single page only).
  • Wolfram: Computational intelligence for complex math.

• Azure Assistants:

  • Data/Code Analyst: Uses OpenAI's code interpreter to process files, generate files with data and images of graphs, solve code problems, and provide data analytics capabilities.

• Anthropic Models:

  • claude-3-5-sonnet - Best for analyzing or creating large bodies of text and for code analysis; strengths include high context limits.
  • claude-3-haiku - Best for quick instruction based tasks with existing data; strengths include speed and high context limits.

• Meta Model:

  • Llama-3.2 - Best for adapting different styles, tones, and analyzing multi-lingual texts; strengths can respond to various types of input.

LLM Configuration Options:

Click the configuration options button to further customize your settings:

• Max context tokens: Maximum input size (1000 tokens ≈ 750 words).
• Max output tokens: Maximum output size (1000 tokens ≈ 750 words).
• Temperature: Controls creativity/randomness.
• Top P: Alternative to temperature, refines responses based on context.

4. Explore the Right-Side Panel

The right-side panel helps you manage prompts and files:

• Open/Close: Use the "Open sidebar" or "Close sidebar" button.
• Prompts: Save prompts for reuse.
• Attach Files: Share and manage files with the AI models.

5. Learn to Share

Share conversations with other authenticated Stanford users:

• In the left-side panel, click the three dots next to the conversation title.
• Click "Share" and then "Create link."
• Important: The conversation will be accessible to any Stanford user with the link, so be cautious about sharing sensitive information.

Removing a Shared Link:

• Click your name (bottom left).
• Click "Settings" then "Data controls."
• Click "Manage" next to "Shared links" to delete links.

6. Customize Your Experience

Access the settings menu by clicking on your name (bottom left) and then "Settings."

• General Settings:
  • Theme: Switch between Light and Dark mode.
  • Auto-Scroll: Automatically move to the latest message.
  • Hide Right-Most Side Panel: Remove the pop-up side panel menu.
  • Archived Chats: Unarchive or delete conversations.
• Messages Settings:
  • Press Enter to Send Message: Enable/disable Enter key sending.
  • Save Drafts Locally: Save drafts of texts and attachments.
  • Default Fork Option: Define information visible when forking.
• Data Controls:
  • Import Conversations: Import from JSON files (e.g., from other GPT chat applications).
  • Shared Links: Manage shared conversations.
  • Clear All Chats: Delete conversations (excluding archived ones).
• Account Settings:
  • Profile Picture: Upload a profile picture (under 2MB).
  • Display Username in Messages: Show your name or "You" in conversations.

7. Try New Features

The AI Playground is continuously updated with new features. Stay updated with the AI Playground release notes.

Conclusion

The Stanford AI Playground provides a rich environment for exploring the capabilities of artificial intelligence. By following this quick start guide, you can begin experimenting with prompts, models, and settings to discover how AI can benefit your work and studies. Remember to always prioritize safety and responsible AI practices.