In a world where large language models like GPT and Gemini are becoming more and more ubiquitous, what does this mean for comedy? If computers getting things obviously wrong is a factor in their comedy value, do the improvements in generative AI take us into an uncanny valley where the outputs are realistic enough to no longer be funny?
Join our presenters as they each argue the case for one side of the battle between generations from large language models, vs. more naive generation approaches.
We’ll explore what makes comedy funny at all, and then drill deeper into why we find content generated by computers so hilarious - often when it’s not trying to be!
Throughout the talk we’ll give some examples of humorous computer generated content from past projects - intentional or otherwise. Learn why you shouldn’t join us for dinner when the recipes are auto-generated; see how we abuse state-of-the-art AI safety techniques to make LLMs funnier; and join for a group singalong for some live-generated parody karaoke.