WT: 4D? I Hardly Know Thee!

Episode Notes

The episode opens with Andrew Mayne describing a YouTube creator, Mashpo, who made a one-dimensional game and then pushed the idea toward a 4D Minecraft project. The hosts unpack what it would mean to perceive a world as a thin strip or a slice, compare it to barcode-like visuals and anaglyph 3D, and discuss whether humans could adapt to higher-dimensional perception using new sensory cues and neuroplasticity. Andrew also ties the discussion to educational games and how such projects could help people, especially younger learners, build intuition for higher-dimensional thinking.

The conversation then moves into AI image generation, especially DALL·E. Andrew says newer versions can generate convincing red-blue anaglyph 3D images and can respond to camera-lens and macro-photography language, while the group debates whether the system is just a toy or something more useful. They discuss prompting strategy, reporting outputs that miss expectations or show bias, and the value of tools that better understand user intent. The episode closes with a Starship update, including the booster on the mount, possible testing, the 33-engine system, and the idea that early Starship missions may be used to deploy Starlink satellites.

Brian also plugs the Weird Things Patreon, and the hosts use the mid-episode AI discussion to reflect on how playful demonstrations can help the public understand AI. Andrew and Brian briefly mention the 1960 George Pal version of The Time Machine as a film Andrew was watching with his wife, but only as a conversation point rather than a recommendation.

Key topics

• One-dimensional and four-dimensional game demos: Andrew describes Mashpo's one-dimensional game and a 4D Minecraft project, and the hosts discuss how such worlds might be perceived as slices, strips, or projections.
• Perceiving higher dimensions through altered visual slices: Brian and Andrew compare the 1D game to a barcode-like view and discuss 4D as something visible only through slices or shadows in lower dimensions.
• Neuroplasticity and sensory substitution: The hosts discuss vibrating vests, tongue electrodes, and the idea that humans might adapt to new sensory channels over time.
• DALL·E generating anaglyph 3D images: Andrew says DALL·E can now make convincing red-blue 3D images and that it appears to have learned this from training data.
• Image models understanding specific visual language: Andrew says DALL·E understands camera lens types and macro photography cues, suggesting broader learned visual concepts.
• Prompting strategy and user intent: Justin says DALL·E works better when given room rather than over-described instructions, and the group discusses tools that understand what users want with less prompting.
• Reporting bad outputs and handling bias: Andrew explains that DALL·E has a reporting tool for outputs that do not match expectations or are inappropriate, and he says users should report biased representations so the system can improve.
• AI as a public-facing entry point to understanding AI: Brian argues that fun image generation helps people develop a functional understanding of AI and makes the technology easier to grasp.
• Educational value of 4D thinking: Andrew suggests that a 4D Minecraft-like game could help middle schoolers understand higher-dimensional thinking and later physics concepts.
• Starship testing and launch readiness: The hosts discuss Starship booster mounting, possible fuel or ignition tests, the 33-engine booster, and the uncertainty around launch reliability.
• Starlink as an early Starship payload: Andrew says a mysterious white box at Starbase looked like a Starlink dispenser, and the hosts discuss Starship likely being used first to deploy Starlink satellites.

Picks

• Brian Brushwood: Weird Things Patreon — He explicitly plugs the Patreon and frames it as the place to get more Weird Things content.