Natix and Valeo are working together to build a decentralised, self-driving camera model that will provide a clear foundation for the safe use of physical AI in everyday life.
Valeo, a company that makes parts for cars, and Natix Network, a decentralised physical infrastructure (DePIN) provider based on Solana, have teamed up to build an open-source AI multicamera model that will help make self-driving cars better.
The two businesses revealed on Thursday that they are working on the World Foundation Model (WFM), which they say would be able to learn and predict how things move in the real world and change based on traffic conditions.
Advancing physical AI beyond text-based models
The multi-camera model wants AI models to go beyond comprehending text to recognising real-life situations in physical spaces. It also wants to help research into self-driving cars.
Valeo and Natix promised to make their models, datasets, and training tools available to the public so that developers may improve these features.
Wayve, a startup that makes self-driving cars, is already employing WFMs in its cars. For example, one of its cars drove around portions of Las Vegas without any prior training in the city, according to information supplied by the company’s CEO, Alex Kendall, on Friday.
DePIN infrastructure and decentralised data networks
WFM is part of the larger DePIN industry, which combines blockchain technology with community-owned physical infrastructure to form decentralised networks where people can give resources, such computing power, in exchange for cryptocurrency.
Marc Vrecko, CEO of Valeo’s Brain Division, claimed that the WFM self-driving camera model’s main purpose is to securely and responsibly “advance mobility intelligence” and the arrival of self-driving cars.
Alireza Ghods, the co-founder and current CEO of Natix, believes that WFMs represent a unique opportunity, similar to the emergence of large language models (LLMs) between 2017 and 2020.
Predictive world models and safety considerations
The multi-camera world model is different from current AI models that only see things. It wants to include predictive capabilities to speed up the use of self-driving cars in the real world.
Natix added that making the WFM decentralised and open-source could let physical AI systems be taught and evaluated in more real-world situations before they are put to use. An official from the firm stated that transparent frameworks expedite the ecosystem’s progress. They also said that thorough testing is very important for safety.
Alpamayo is a collection of open-source vision-language-action models that chipmaker Nvidia made. They are one of Valeo’s and Natix’s biggest competitors. The answer uses camera and sensor data to make decisions through reasoning-based autonomy.
Natix was created in 2020 and runs a decentralised, multi-camera data network that has hundreds of thousands of participants and millions of kilometres of recorded driving data.