My mind is on speculative fiction involving the possible consequences of the use of AI. My upcoming short comic, Changeling.ai—words by me and Michael Judd, art and letters by Olga Wieszczyk—will be published as part of the Systemic Anthology. Because of that comic and a side of existential dread, I spent some time digging into AI and its possibly dangers to its users. There are many dark scenarios at the end of our AI use but there are also some comical outputs such as the birth of Elias Thorne.
The fictional character Elias Thorne appears in over 25% of stories created by LLMs (or 26.5 % of the 20,000 of them studied by Cornell researchers). So, this means, when you ask a LLM to tell you a story, there is approximately a 25% chance the character might be named Elias Thorne. The odds are, he might also be a lighthouse keeper or a clock maker. His constant reappearances represent a model breakdown or AI inbreeding. As these larger models consume their own slop, more and more of these…

Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.