CAIA Speaker Event: Adam Kalai (OpenAI)
Here are some details on Caltech AI Alignment’s next speaker event:
Who: Adam Kalai (virtual), OpenAI
When: May 8th at 5-6 pm PT
Where: Watch party in Chen 100
Zoom link: https://caltech.zoom.us/j/85682146288
What: Adam Tauman Kalai is a Research Scientist at OpenAI, specializing in AI Safety and Ethics. Adam earned his BA from Harvard University and his PhD from Carnegie Mellon University. His talk will unpack why language models hallucinate and what the field can do to reduce hallucinations. He will show how pretraining induces hallucinations in base language models. Specifically, for any categories of facts where one cannot learn to "fact check" them from the training data, then base models cannot avoid hallucinations. Second with current evaluations, wild guessing increases both accuracy and inaccuracy. Hence, post-training techniques are selected and optimized so as to increase hallucination.
No specific technical background is required - we welcome all interested students who are eager to learn! As with all CAIA events, we will have pizza and boba!