Talks at AI x Voices
AI x Voices #1 · Upcoming
Tuesday, 6 October 2026, Hevre
“LLMs and their manners, or what a model does when nobody is watching”
One of the fundamental traits of LLMs is that they generate answers in line with our moral and ethical principles, a property known as AI alignment. Przemek will show what AI knows about morality, how models are trained for alignment, how surprisingly easy it is to break them and how little their guardrails actually protect. He will also cover the phenomenon of one model having many personas and what that means for all of us.