The Matrix

In his De Tijd column, Jeroen explains why world models, AI that simulates objects, physics and causality, are the next leap after ChatGPT. Think DeepMind’s Genie 3 generating rich 3D worlds to train other AIs, Nvidia’s Earth 2 forecasting weather at street level, and Meta’s V-JEPA 2 learning how actions change the world. Paired with language models, this points to more capable robots, safer autonomy and faster product development through increasingly realistic digital twins. The obstacles are real (new breakthroughs, massive compute and better data), but the direction is clear. There is even a mind-bender at the end: if we can simulate reality perfectly, what does that say about our own?

Read Jeroen’s column in De Tijd.