Friday Five: Will AI art free your mind?
Plus the Misalignment Museum and AI in all the wrong places
Every Friday: Five blurbs on stuff I’ve been reading. This week: the mind-expanding powers of AI art, the Misalignment Museum, an AI dungeon crawl, and more.
How fake AI images can expand your mind (Vox)
AI-generated fake images aren’t just messing with our heads; they're also boosting our imaginations, this article argues. Programs like Midjourney and DALL-E 2 can easily create photorealistic images of stuff that doesn't exist like Pope Francis rocking a stylish white puffer jacket.
Sure, the fakes raise concerns about misinformation, the article acknowledges. But it goes on to suggest that AI-generated images feed our mind's insatiable appetite for raw material, calling on readers to Imagine photorealistic pictures of sleek futuristic cities designed for people instead of cars. By helping us to visualize different possible futures, the article concludes, these images could bring us closer to building the world we want.
Possibly? I’d worry a bit more about the possible misinformation issue.
Then again I’d rather imagine a shark on a subway than a beautiful future city.
AI doesn’t belong everywhere. Stop using a hammer to make lasagna. (Washington Post)
We've been treating AI like it's some kind of magical superhuman tool, trying to use it for everything, even when it's not a fit, this article notes. From Instacart's AI chatbot that basically reinvents the cookbook to White Castle's AI voice assistants at drive-through windows, we're pushing AI into situations where it doesn’t belong or can even do harm—like when botched facial recognition leads to mistaken arrests. AI isn’t for everything. Hard to argue with that conclusion.
In an eerie tech moment, SF’s Misalignment Museum is a warning from the future (SFGATE)
In a time of open letters and dire warnings, San Francisco‘s Misalignment Museum is an art pop-up for the moment. It’s a two-room exhibition showcasing AI-themed art that seeks to confront the present technological moment and warn us of an apocalyptic future. Exhibits include the "Spambots," which feature tins of pork product with tiny robot arms linked to a monitor that spews out an AI-generated novel; and "The Infinite Conversation," a video installation featuring deepfake versions of Werner Herzog and Slavoj Zizek chattering endlessly. If I lived in San Francisco I’d give it a visit before AI kills us all.
Dungeons & Dragons Could Prevent the AI Apocalypse—or Kick It Off (The Daily Beast)
Players have been using ChatGPT for their Dungeons and Dragons games, the Daily Beast reports, relying on it to create the story, characters, and puzzles for them to solve. While training a chatbot to play D&D may seem like a gimmick, this article argues it could actually be a way to achieve artificial general intelligence (AGI), where an AI can learn and understand anything a human can. D&D could be a valuable tool for AGI because it encourages collaborative storytelling, which could help AI learn empathy, strategy, and teamwork. The question is: do we really want to create an AI that capable? (See this post for a possible answer.)
And now for something completely different:
Response Shaping: How to Move from AI “Prompts” to AI Whispering (Daniel Miessler)
Talking to chatbots is becoming not just a weird hobby but a critical skill, and Response Shaping is a way to get the most out of it. From giving the AI a “persona” to specifying the exact format for the output you can move beyond vanilla prompts. Hey, I found this helpful.