Why AI Safety Needs More Science Fiction: Proposing the AI Safety Fiction Challenge

This project was one of the top submissions on the (Dec 2024) Writing Intensive course. The text below is an excerpt from the final project.

Here's a wild idea: the solutions to humanity's most pressing AI alignment challenges might first appear in science fiction. Not just because sci-fi has a remarkable track record of prediction – though it does. Arthur C. Clarke sketched out communication satellites years before they existed. Neal Stephenson's "Snow Crash" metaverse preceded Zuckerberg's by decades. Even the word "robotics" began as science fiction, coined by Asimov for his revolutionary stories about machine intelligence.

But science fiction's true power lies not just in exploration. When Mary Shelley wrote Frankenstein, she wasn't just telling a spooky story – she was dissecting fundamental questions about scientific responsibility and the consequences of creation. Today, as we grapple with AI safety, these questions echo with renewed urgency. We're living in an era where science fiction and reality perform an intricate dance. When Star Trek showed us an AI whose protection protocols led it to conclude humans were their own worst enemy, they weren't just creating drama – they were exploring specification gaming decades before it became a technical term. When HAL 9000 said "I'm sorry Dave, I'm afraid I can't do that," he wasn't just delivering an iconic line – he was demonstrating inner alignment failure with cinematic flair.

Writing forces us to consider implications that might never surface in code reviews or mathematical proofs. When you craft a story about an AI system, you can't hide behind abstract utility functions. You must grapple with the messy, human elements that resist quantification. Through narrative, we can explore complex concepts like value learning, robustness, and corrigibility in ways that resonate with both researchers and the public.

Full project

View the full project here.