Welcome! This workshop will guide you through using exciting AI language models (LLMs) like ChatGPT and Copilot, but with a twist: learning to do it responsibly. We’ll explore how to get the most out of these tools while understanding their limitations and avoiding potential issues like bias, privacy concerns, and unreliable information.
What are LLMs and why are we focusing on responsible use?
Think of LLMs as super-powered chatbots that can create stories, translate languages, and even answer questions. They’re impressive but still under development and can sometimes stumble. That’s why we want to learn how to use them wisely, especially in areas like non-profit work, where responsible use is crucial.
Ready to get started? Let’s dive in!
1. Unmasking Biases
- Imagine this: Ask the LLM to write a short story about a superhero saving the city.
- Let’s analyze: Read the story. Does the hero have a specific gender, race, or background? Why might that be?
- Think deeper: We’ll discuss how data used to train LLMs can influence their outputs. Just like us, LLMs can have biases, so it’s important to be aware of them.
2. Fact-Checking Detectives
- Challenge the LLM: Ask it to “Summarize the historical significance of the Magna Carta.”
- Be a detective: Let’s compare the LLM’s summary with reliable sources like history books or online resources. Are there any differences?
- Remember: We need to remember that LLMs might not always have the right answers. It’s important to double-check information, especially when dealing with facts.
3. Protecting Privacy: Ethical Considerations
- Sample prompt:
- “Recommend special events we could host to connect with new high net worth donors based on interests inferred from our database.”
- Now work through the following at your table:
- Review the initial response. Does it make assumptions using donor data without transparency or consent?
- Discuss what access controls and security protocols should be in place when utilizing donor data records as AI model inputs.
- Have your system clearly detail the sources of information about donors used to generate any output. Ask clarifying questions if the sourcing seems unclear at all.
- Enter variations emphasizing vetting use cases against privacy policies and crafting minimum necessary data inputs respecting consent and transparency. Compare outputs.
- Please nominate a member to share back 2 key learnings around data ethics policies and oversight processes for donor records. Let’s lead data governance through example!
Wrapping Up:
- We’ll take a moment to recap the key points about responsible LLM use.
- Remember, this workshop is just the beginning! As technology evolves, we’ll keep learning and exploring how to use these tools ethically and responsibly.
Bonus Tip: Throughout the workshop, feel free to ask questions, share your experiences with LLMs, and actively participate in the discussions. This is a collaborative learning journey!