Converging, but unevenly: where physical AI is heading and who can use it
World models and physics engines are converging, but the convergence is uneven, and that unevenness is the most useful thing a leader can understand about the field right now. Aicadium, a Singapore-based industrial AI company, has tracked both technologies and tested them hands-on. Our position is that the two approaches are growing into each other while a second divide quietly widens: who can actually run the best models.
The two fields are growing into each other
Physics engines are gaining learned, AI-like components, and world models are starting to build in physics. Differentiable physics makes a simulator’s calculations open to optimisation, so a system can learn from far fewer trials. Learned simulators use neural networks to approximate physical dynamics. Some run one to two orders of magnitude faster than the simulator they were trained on. From the other side, recent world-model research borrows explicit geometry and physics to make its predictions more physically consistent. The result is two families of tools that increasingly share techniques, even as they keep their distinct strengths.
This is genuine convergence, not marketing. Yet it is uneven. Contact-accuracy tasks, where exact force and geometry decide the outcome, remain the domain of physics engines. Visual-diversity tasks, where many plausible scenes matter more than numerical precision, remain the domain of world models. Embodied AI and autonomous driving are where the two clearly meet. Games and digital twins are still largely physics-engine territory.
A note on a loaded term
The term world model is loaded, and precision matters. Several products marketed as world models are, on close inspection, something else. Often they are a vision-language-action model that maps an instruction to an action. Sometimes they are a generator that produces a scene without being interactive. Aicadium applies a strict definition. A true world model is action-conditioned and predicts the next state of an environment you can interact with. Holding to that definition helps leaders separate genuine capability from confident labelling.
The access asymmetry few are discussing
The most capable models are increasingly open, yet practically out of reach for smaller teams. In our experiments, the largest and most realistic models performed clearly better at following instructions and producing physically plausible results. They were also the hardest to use. Running a model to make a single prediction is far less demanding than training or fine-tuning it. The best models require server-class computing to fine-tune.
This is the access asymmetry. Open weights do not equal open access when the compute bill is the real barrier. The encouraging side is that large labs now release strong pre-trained models. A smaller team often needs only to adapt one rather than train from the beginning. That lowers the barrier, though it does not remove it. Workstation-class hardware keeps improving, which widens what a small team can attempt each year.
Where the field goes next
A clear architectural trend is emerging from the convergence: hierarchy. Fast control and slow reasoning pull in opposite directions. Manipulation needs very low latency, while planning needs a larger, slower model that understands cause and consequence over time. The promising design places a fast action model underneath a slower reasoning model. The central challenge is how the two communicate.
It is worth keeping perspective. Today’s physical AI resembles large language models around 2022, impressive but prone to errors and inconsistency. Much of the progress since then came from the engineering built around those models, not the models alone. A similar trajectory in physical AI could make robots viable in many domains that are out of reach today.
Are world models and physics engines merging into one technology?
They are converging but not merging into a single technology. Physics engines are adopting learned and differentiable components, and world models are incorporating physics, yet each still leads in different tasks. Contact-accuracy work favours physics engines, while visual-diversity work favours world models.
Why does the access asymmetry matter for strategy?
The access asymmetry matters because capability and accessibility are diverging. The strongest models are open in principle but expensive to adapt in practice. A realistic strategy plans for adapting pre-trained models on accessible hardware rather than training from scratch.
Aicadium’s view is that the winners in physical AI will not be those with the largest models. They will be those who match the right method to the task and plan honestly around the cost of using it. Convergence rewards judgement, and judgement is something an organisation can build.
