How AI can be Restrained to Fight Hallucinations for Scientific Applications
Details
Join NWIRC, Cascadia AI and Cloud Computing, and guest speaker Chelsea Liekhus-Schmaltz
Chelsea Liekhus-Schmaltz PhD — Senior Research Data Scientist - Zap Energy, Founder and CEO Ergo Labs AI
The meetup will include networking time from 5-5:15pm for those who attend in person and an optional tour of the facility before the hybrid event begins at 5:15pm.
NWIRC will provide light refreshments/snacks for in-person attendees.
Unable to attend in person?
Join Zoom Meeting - REMEMBER virtual event does not begin until ~5:15pm
Join Zoom Meeting https://us06web.zoom.us/j/85628523263?pwd=orhkvZADzlBBTb30P4FA4oboJ4iixC.1
Meeting ID: 856 2852 3263 Passcode: 122122
Let us know how plan to attend by registering here: https://luma.com/3ts3pxds
About the Guest Speaker and the Talk
Chelsea has their PhD in Physics from Stanford University and currently works at Zap Energy, a fusion start up, as the lead of the data science team. They have seen how research has been progressively impacted by data science, machine learning, and now generative AI. They are currently working on a new project called "Ergo Labs AI" that works to make AI solutions for science.
One of the biggest problems that generative AI faces is its propensity for hallucinations. This is a particularly difficult problem in "unverifiable" fields like science. At Ergo Labs we are exploring ways to solve these problems in the context of reading and reviewing scientific literature. We will show how specific tools and architectures such as MCP servers, vector databases, and schemas can alleviate some of these problems.
Whether you're a beginner or expert user of AI tools, you'll walk away with practical insights.
Photography Notice: Photos and video may be taken at this event for promotional use. Please contact us in advance, or notify the on-site staff if you prefer not to be included.