AI and Research Practice
Upcoming
The opening talk. How large language models actually work, and what that means for getting output you can trust in research.
When
Tuesday, October 13, 2026, noon to 1:30 pm ET (90 minutes), in person at Michigan State University. The recording will be posted on this page after the talk. IBR also hosts the recording.
What to take with you
THE CHALLENGEFluency tells you nothing. LLMs predict, and your output is one draw. A wrong output reads exactly like a right one, and sounds just as certain.
THE RESPONSE, PART 1Set up the input. Check the output. Context is front-end judgment. Verification is back-end judgment.
THE RESPONSE, PART 2Adjust both to the type of question. When a right answer exists, an external source settles it. When it doesn’t, your judgment does.
Go further
The guide covers everything in the talk and more, and every demo from the talk has its own page under Demos. Questions from the room will be answered on the Questions page.