25 Years to World Simulations

Executive summary

• Three media, three clocks. Motion pictures had their first century from the Lumière public screenings of December 1895 to the centenary in 1995. Video games have had their first fifty years from 1976, when the Fairchild Channel F brought microprocessors and swappable cartridges into the home. Story Generators, counted from 2020 when GPT-3 made live generated stories possible, are on course for a first twenty-five years ending around 2045.
• Each medium matured in roughly half the time of the one before. This is an observed pattern, not a law. It holds because each new medium inherits the grammar, audiences, hardware and business models of the last one, so it spends less time inventing them.
• The stages repeat. Each medium moves through attraction, grammar, a sensory leap (sound and colour for film, 3D for games), industry formation, mass culture and a technology plateau. We map all three side by side in the table below.
• Story Generators are at the 1970s text-adventure stage. A handful of teams are working out the grammar, as Crowther and Woods, Infocom, Adventure International and On-Line Systems did between 1976 and 1980, before an industry formed.
• Our projection for 2026 to 2045: text, then illustrated stories drawn from studio canon, then voice, then generative interactive film, then 3D generative worlds and generative VR in the mid-to-late 2030s, then world simulations by about 2045. Every date after 2026 in this report is a projection.
• The main risks are compute cost, canon coherence, regulation and studio consolidation. We also expect the medium to be renamed: once major studios arrive, “Story Generator” will probably get a short consumer name, just as “motion picture” became “film” and “the movies”.
• For creators, the order matters more than the speed: own the canon and the performance layer now, add voice and pictures as the models allow, and stay ready for live 3D.