Tech & Appsbreaking news
Niantic sold Pokémon Go and bet on world models
The studio behind Pokémon Go now wants to replicate the real world, and the same technology is reaching robots, driverless cars and gig workers with phones.
A game maker changes course
Niantic, creator of Pokémon Go, sold the game last year and pivoted to world models. [1] If you follow robotics, driverless cars or AI video tools, that shift touches the technology you will see next.
At Niantic Spatial, CTO Brian McClendon hopes to use Niantic's world data to build replicas of the real world. [1]
What a world model does
Unlike large language models, world models use visual inputs and can generate digital environments that simulate physics. [1] Odyssey CEO Oliver Cameron says a world model takes a visual state of the world and predicts the next visual state. [1]
Wayve chief scientist Jamie Shotton says a world model must understand physics, cause and effect, and how other objects and people respond to one's actions. [1]
Training robots and cars in simulation
McClendon says robots can be trained in simulation by photographing their intended environment, generating a 3D model, and running millions of simulations. [1] Driverless cars also improve using world models. [1]
Wayve builds world models designed for cars. [1] Its generative world model is powered in part by dashcam footage. [1] The company uses it to recreate hazardous roads and unsafe conditions that would not be practical to test in real life. [1]
2027 could be a major year for the expansion of home robotics and driverless cars, both bolstered by advances in world models. [1]
Who else is building them
Tencent is investing in its own world models because physics and 3D rendering are core to video games. [1] Yann LeCun, Meta's former chief AI scientist, recently raised over a billion dollars to fund his Advanced Machines Intelligence startup. [1]
Fei-Fei Li launched the startup World Labs in 2024. [1] AMD acquired World Labs this year for over $8 billion. [1]
Google's Genie 4 arrives
Google's Genie 4, a world model that can simulate real and imagined physical worlds, debuted the same week as a new language model. [2] Genie 4 will add audio, letting developers include background noise and dialogue in their AI worlds. [2]
That language model, Gemini 4 Carbon, was announced on a Wednesday, just two weeks after Google's last release. [2] It surpassed Google's Gemini 4 Argon and OpenAI's GPT-6 Astra on benchmarking tests designed to assess overall capabilities. [2] Google is pitching it for cybersecurity, saying it is good at detecting and patching vulnerabilities. [2]
The people behind the data
AI labs need real people to record the physical world in order to simulate it. [1] The author recorded household chores with an iPhone and sold the video data through an app for a fairly small payout. [1]
World models are already changing the labor market behind the scenes, and more gig work may demand this kind of data collection as companies build better world models. [1]
What this means for you
If you want to try Google's new world model, users can join a waitlist for Genie 4 now. [2] Gemini 4 Carbon is a different case: only select cybersecurity experts currently have access, and it is not yet available to the general public. [2]
If you do gig work, more of it may demand data collection like recording the physical world. [1] The author's own experience was a fairly small payout for video of household chores. [1]
Sources
Written from the sources above and checked twice before publishing. How we work · Report an error
.jpg){kind=link}