‘Ghosts in the Machine: Large Language Models Are Mostly Full of Crap But Still Surprisingly Useful If You Dare’

Alternative title: ‘There is No Artificial Intelligence, Only Zuul; or, now that this thing is here, what can/should we do with it?’

We are so pleased to host again our colleague from Carleton University: Dr. Shawn Graham! Dr. Graham been playing with models of language since the days of Markov chains and Recurrent Neural Networks. In this talk, he’ll explain poorly how they work, why they work, and why you might want to just think of them as ‘calculators for words’. Alternatively, we can think of them as ghostly echoes of our web culture back to itself. In which case, he’ll walk us through some of the ways he’s found listening to these echoes to be useful in his research, in his teaching, but will also share some of the dead ends too. This isn’t a boosterism talk but rather an attempt to dispel hype. He didn’t write this abstract with the help of an LLM. Location: Univeristy of Ottawa CreatorSpace (Pérez Hall, room 302).