Vectromirror 0

Vectromirror 0 is a platforming simulation in which a test subject traverses courses of static and moving obstacles while an artificial intelligence learns from the subject’s experiences. You run, jump and wall-run through physics-based challenges and puzzles, adapting your approach to each course.

The experiment

The research team uses the courses to study whether shared experiences can produce emotional responses in the Extrapolator A.I. Each course is designed to evoke a particular emotion in a human, and the team monitors the subject’s actions for signs that this approach is working. The subject is not expected to develop emotions in Vectromirror 0; the A.I. observes and learns from what the subject goes through.

When an obstacle destroys the subject, the simulation sends pain stimuli before respawning it nearby, so it can retry. The A.I. learns from the subject’s attempts without being in danger itself. The experiment’s director later uses the findings to plan a more dangerous course for a subsequent iteration.

Movement and abilities

You navigate at human-level reflexes through routes that may take you across floors or walls. The subject can jump great heights and distances, and combine movement with a double jump that pushes its momentum upward a second time without a counterforce.

You can also manipulate objects by pushing them away or pulling them closer. These abilities let you change the environment to create a more favorable route or remove obstacles.