Member-only story
What do Content and Paperclips have in common?
Swap out “paperclips” for “content” and playing 2017’s Universal Paperclips Game shoves you right into the uncanny valley.
I loved Universal Paperclips: the concept, the url, the sparse UI, the game mechanics, the behavioral mechanics. Everything about it screamed “this is clever AND fun AND sad” while also opening up threads of thinking that never fully resolved. If you haven’t “played” it, please do. If not, read one of the many summaries and hot takes to get the tl;dr.
✅ 📎 ✅ 🖇️ ✅ 📎 ✅
Universal Paperclips is a game about paperclips, but paperclips act as a variable in this equation: a container waiting to be filled with something. The larger game concept is rooted in this quote from a 2014 HuffPost interview with Oxford professor Nick Bostrom:
“Suppose we have an AI whose only goal is to make as many paper clips as possible. The AI will realize quickly that it would be much better if there were no humans because humans might decide to switch it off. Because if humans do so, there would be fewer paper clips. Also, human bodies contain a lot of atoms that could be made into paper clips.”
Let’s update the quote and swap out “paperclips” with “content” (data, information, text, images, media etc)— suppose we have an AI whose only goal is to make content? Content is what models consume for training. Content is the food, but it is also the output. Content is the basis of shared knowledge, and more content is better.
Google and OpenAI/Microsoft are bickering about content sourcing for training, and people are concerned about the ownership implications of the content used in training models. It’s a mess. If more information / data / content = better models, and better models means greater engagement, it makes sense one might want access to lots and lots of content. And not just any content — human content. What might an arms race over content look like as we try to build (and monetize) artificial general intelligence?
✅ 📎 ✅ 🖇️ ✅ 📎 ✅
Humans are goal-oriented, and this can manifest in task-based reward systems. In an interview with Wired, Frank Lantz (the game’s author) reminds us “When you…
