The AI That Can Imagine the Physical World

 

 

Most people's mental model of AI is shaped by ChatGPT. You type something, it responds. The AI lives in the world of language — reading, writing, summarising, and answering.

A different kind of AI is now drawing serious investment, and it works on a completely different principle. It does not process language. It simulates physical reality.

 

What a World Model Actually Is

A world model is an AI system that builds an internal representation of how the physical world behaves. It learns — from enormous quantities of video, sensor data, and real-world observation — what happens when objects move, collide, fall, or interact. It develops an intuition for physics.

 

The practical implication: a world model can simulate "what happens next" in a physical environment without having to run the experiment in the real world.

 

For example, Odyssey, a startup founded by engineers who previously built autonomous vehicles, recently raised $310 million to develop this technology. Their approach involves capturing real-world environments at the detail level of satellite imagery — think Google Earth applied to physical spaces — and using that data to build simulations that behave with genuine physical accuracy.

 

Amazon, AMD Ventures, and GV are among the investors backing the company at a $1.45 billion valuation. When investors of that profile concentrate capital on a single concept, it signals something more than a product bet.
 

 

Why This Is Different from AI You Already Know

The AI most organisations have been experimenting with — language models, document summarisers, chatbots — works well for tasks that live entirely in the world of information: writing, searching, classifying, translating.

 

World models address a different category of problem: tasks that require understanding of physical space and cause-and-effect relationships.
Consider what that means in practice. Training a robot today requires thousands of hours of physical trials. The robot attempts a task, fails, adjusts, and repeats. It is expensive and slow.

 

A world model changes this. The robot can run millions of simulated trials in a virtual environment that accurately reflects how physical objects behave before a single physical test takes place.

 

The same logic applies to factory layout planning, autonomous vehicle testing, construction simulation, and any industrial process that currently requires physical prototyping to validate.
 

 

What This Means for Thai Industry

Thailand's manufacturing sector — automotive, electronics, food processing, packaging — faces a straightforward pressure: the cost of automating production lines is high, and the cost of getting it wrong is higher. Most companies delay automation decisions because the validation process is slow and expensive.
 

World model technology directly addresses that bottleneck. If you can simulate a new production line layout, a new robot configuration, or a new logistics flow in a digital environment before committing capital to physical implementation, the economics of automation change.
 

 

 

 

This is not speculative. Automotive manufacturers globally are already using simulation environments to reduce prototype cycles. What is new is that these simulation capabilities are becoming more accurate, more accessible, and available through vendors rather than requiring custom engineering teams to build.

 

The question for Thai businesses is not whether this technology will affect their sector. It is how quickly simulation tools will reach the vendors and integrators they already work with.

 

 

A New Way to Think About Role of AI

For business leaders thinking about where AI will matter most in their operations over the next five years, this distinction is worth holding onto. Language AI will change how knowledge work is done. World models will change how physical work is designed, tested, and optimised — before it happens in the real world.

 

That is a different kind of value, and for industries built around physical production, potentially a larger one.

 

 

 

Oravee Smithiphol is Tech Intelligence & Insights Manager at SCB 10X, where she tracks emerging AI developments across startups, enterprises, and financial services organisations worldwide.
About SCB 10X

 

 

SCB 10X is the disruptive technology investment arm of SCBX Group. With an investment track record since 2016, SCB 10X has deployed over USD 500 million globally into startups in AI, blockchain, and fintech. SCB 10X has backed exceptional companies such as Together AI, Pagaya, Ripple, Fireblocks, Anchorage Digital

Beyond capital, SCB 10X partners with our portfolio founders to test, grow and scale their solutions through SCBX’s network, unlocking commercial opportunities into Thailand and Southeast Asia. Mandated as the group’s speedboat, we discover and ship state-of-the-art technologies and solutions into SCBX group. 

 
For more information, please visit https://scb10x.com/