Why You Need An AI Use Policy (And How To Make One) with Kim Snyder

Understanding Generative AI

• 💡 The current "game-changer" in AI is generative AI, exemplified by tools like ChatGPT, which burst onto the scene in November 2022, making AI more accessible.
• 🚀 Unlike complex machine learning, generative AI focuses on creating things like text, images, video, and translations, making it ubiquitous.
• 📈 AI's growth is exponential and not a fad; its impact is likened to the internet's introduction, fundamentally changing how work is done.

Why Nonprofits Need an AI Policy

• ✅ Many nonprofit staff are already using AI tools (e.g., Bing Chat, GPT-4) because they are accessible and often free, even if not officially sanctioned.
• ⚠️ Ignoring AI is risky; organizations that don't address it will be at a disadvantage as AI is not going away.
• 🧠 It's crucial for human-centered people, especially in nonprofits, to engage with and understand these tools to ensure they are used for good.

Key Risks of AI Tools

• 🤥 Confabulation: AI tools can sound convincing and confident but be entirely wrong, making up facts or quotes, so always keep a human in the loop for review.
• ⚖️ Bias: Generative AI is trained on world data, which is inherently biased, potentially leading to skewed or inappropriate outputs, especially in image generation.
• 📝 Copyright: There are significant concerns around copyright infringement, as AI models are trained on existing works, raising issues for artists and writers.
• 🔒 Privacy & Cybersecurity: Inputting personal, proprietary, or identifying information into prompts can lead to data breaches, emphasizing the need for caution.

Essential Elements of an AI Policy

• 🛠️ An AI policy should define sanctioned tools for organizational use to prevent "Shadow AI" and ensure consistency.
• ✍️ It must address attribution (whether to credit AI for first drafts) and establish a clear review process to prevent unverified content from being released.
• 🛡️ Guidelines for anonymization are vital when using AI for data analysis, especially with sensitive information like donor data.
• 🔄 Policies should be living documents, requiring regular review cycles (monthly/quarterly) to adapt to the rapid evolution of AI technology.

Identifying Beneficial AI Use Cases

• 📊 Nonprofits should conduct a risk-benefit analysis to identify low-risk, high-benefit tasks suitable for AI assistance.
• 🚀 AI can significantly increase productivity, with studies showing tasks completed faster and with higher quality, and reducing the "draining" nature of certain tasks.
• 💡 Examples of effective uses include generating first drafts for marketing personas, campaign pitches, grant proposals, summarizing reports, or drafting difficult emails.
• 🎯 Practice is key to learning how to effectively interact with AI, starting with tasks that are not high-stakes.