Genie 4 is now available

Google DeepMind’s release of Genie 4, a generative “world model” that can turn prompts into interactive 3D-like environments, is drawing equal parts awe and skepticism. Commenters question its practical value beyond tech demos, pointing to issues like object permanence, unclear real-world use cases, and Google’s confusing, expensive “Ultra 20x” access model, while others see potential for gaming, simulation, and robotic training. The conversation broadens into a debate over AI’s societal impact, from fears of escapism and job loss to optimism about automation and medical breakthroughs.

Access, Pricing, and Tiers

  • Many commenters are frustrated that Genie 4 is limited to “Google AI Ultra 20x” / high-end tiers, describing $199–$200/month pricing as steep.
  • Confusion over eligibility is common: differences between consumer vs Workspace vs GCP billing, region restrictions, and unclear upgrade paths.
  • Several people say they’d happily pay per token or for a simple trial, but are blocked by opaque plans and “contact sales” flows.

Google’s Product and Account Complexity

  • Strong criticism of Google’s fragmented product lineup and tendency to launch/abandon projects.
  • Multiple horror stories about tangled Google accounts (AdSense, merchant, developer, family plans) make some users hesitant to give Google more money.
  • Workspace admins often restrict AI features, further limiting access even when organizations already pay Google large sums.

Use Cases and Value Proposition

  • Some see Genie 4 as a cool tech demo but struggle to find concrete use cases beyond “lucid dreaming as a service” or gimmicky entertainment.
  • Others suggest potential for:
    • Game or level prototyping, especially if export to formats like FBX/USD/splats or integration with Unity/Unreal/Godot becomes possible.
    • Training environments/world models for robots and autonomous systems (e.g., simulation before acting in the real world).
    • New forms of interactive movies or media where audiences guide the narrative in real time.
  • A few niche ideas include visualizing animal perception (“umwelt”) and realistic synthetic data for robotics or driving.

Technical Concerns and Limitations

  • Demos seem to avoid testing object permanence; users suspect issues when turning the camera or revisiting areas.
  • Debates on whether a neural world model should be the engine vs. simply generating assets/engines.
  • Some criticize this as resource-intensive compared to simpler media (3D panoramas, videos).

Societal and Philosophical Debate

  • Significant thread-wide discussion about AI’s broader impacts: job displacement, creativity, escapism, and regulation.
  • Views range from enthusiastic “let robots do everything” optimism to deep unease and desire to slow or regulate development.
  • Comparisons are made to past tech (TV, internet, nuclear weapons, adtech) and whether AI’s downsides can or should be constrained.