Is Cloning the World Possible? Videogame Engine to Train AI to Effectively Handle Realistic Scenarios

A tech company recently said it wants to utilize its videogame engine to develop naturalistic clones of the actual world, hoping it will make training artificial intelligence safer, not to mention easier.

Unity, the company powering famous videogames like Cuphead, Pokemon, and Hollow, is being used to create "digital twins" or virtual replicas of all things from rollercoasters to all cities, a similar Daily Star reported.

@twitter|https://twitter.com/dailystar/status/1521145486861762560?ref_src=twsrc%5Etfw@

This is for tech firms to be able to create realistic simulations to test new products and even teach AI to handle real-world scenarios more efficiently.

However, now, according to company executive David Rhodes, Unity wants to go a step ahead and construct a "digital twin of the world."

ALSO READ: AI Clerk Serves Customers at 7-Eleven South Korea for the Opening of Its First-Ever Unmanned Convenience Store

(Photo: Spencer Platt/Getty Images)
Companies such as Lyft and Tesla are training their vehicle AIs by driving around real roads, although this has resulted in several accidents as the artificial intelligence is still ‘learning how to behave.’

Developing Non-Player Characteristics

According to some news sites, a similar report from the Silver Screen Beat said that Unity is utilizing its engine to develop realistic virtual humans called non-player characteristics or NPCs.

These can be used to test everything from the rollercoasters' safety to real-world' traffic situations for a self-driving vehicle, minus anybody getting hurt.

According to Unity's supervisor, senior vice president of AI Danny Lange, one can recreate a world that is better than the actual world for training systems in a synthetic world.

He added that he can create many more scenarios "using that data in unity." For instance, this could someday help make fully self-driving cars a reality.

Training Artificial Intelligence on Data

Companies such as Lyft and Tesla are training their vehicle AIs by driving around real roads, although this has resulted in several accidents as the artificial intelligence is still learning how to behave.

Moreover, AI requires thousands of types of training data to learn how to handle itself, although this is currently limited by the length of time a driver can take a vehicle out on the road.

With this game engine, engineers can virtually produce as much data as they could drive a car 500 million miles every 24 hours.

Once the AI is trained on this data, then gets uploaded on an autonomous vehicle, it will be much better positioned to handle roads minus the driver.

Referring to the new technology, Lange asked about how many times it would take to train an AI, "a thousand?" In a game engine, he explained, one can, in fact, have an NPC "trying to get killed in front of a car." and you can see if it can avoid it.

While the tech firm is still a way off from developing a digital twin of the entire planet, use cases such as self-driving show the potential for the technology that could someday result in something like the movie, The Matrix.

Related information about games and Matrix is shown on Jacksepticeye's YouTube video below: