ChatGPT quoted me a section of California law that doesn’t exist.
Claude summed a list of numbers and... just left one out.
When I checked their work and pushed back, both said "You're right." (So glad we cleared that up, thanks.)
Gen AI gives answers based on probability, and it's tuned to give pleasing results. Without proper guardrails, those results can be anything from pandering to hallucinatory.
SO: LEARN HOW TO BUILD THE GUARDRAILS.
This Thursday, October 15, I'm proud to be running a hands-on workshop for the STEMinist Summit: "AI for Deep (and Accurate) Analysis." We'll cover choosing the right LLM, building prompts that work, and agentic reviews.
The STEMinist Summit runs online October 15 and 16, hosted by AJ Lauer and Aleksandra Lemańska. It's all about building confidence, visibility, and leadership for women and nonbinary STEM professionals. steministsummit.com
Want 15% off? I've got 5 discount codes. Hit reply and I'll send you one.
Know a woman or nonbinary STEM pro who should be there? Send this their way.
Also speaking: Mozhgan Kabiri Chimeh, Wendy Wilhelm, Ewelina Skowron, Alyssa Whitcraft, Aleksandra Łoś, Anastasiya Zygmantovich