Hands-on Workshop: Quick tests to spot when AI is fooling you
Details
## Details
Food and drink from 5-5:30. Come early!
We're partnering with Human-Centered AI to present a hands-on workshop in Critical AI Literacy. Bring your laptop and something you're actually working on. Together, we'll run through a series of 3 tests that will check whether your LLM is invisibly:
- Telling you what you want to hear
- Giving confident responses that aren't backed by evidence
- Leaving out important details
Your work can stay confidential! The tests work regardless of domain and will reveal more than you're expecting. Even if you have your responses "locked down" with a facts-only system prompt, you'll learn more about the potential fallibility of the systems your work is based on.
We will follow with a facilitated discussion about how the product design choices made by foundation model companies keep us engaged, sometimes at our own expense. Then we'll explore various techniques to shore up your defenses and create more reliable and robust workflows and systems.