In the latest developments within the tech realm, Niantic, the creators of the groundbreaking augmented reality game Pokémon Go, are leveraging the app’s extensive user-generated data to enhance the capabilities of delivery robots. Since its launch in 2016, Pokémon Go has attracted over 500 million installations within just two months, providing a vast dataset that Niantic Spatial, an AI-focused spinoff of Niantic, is now utilizing to construct sophisticated world models. These models serve to ground large language models (LLMs) in tangible environments, significantly improving the navigation accuracy of autonomous robots. This development represents a significant leap forward in the integration of AI with real-world applications, showcasing the potential of crowdsourced data in enhancing machine learning algorithms.
Meanwhile, the quest for extraterrestrial life is intensifying as the U.S. faces increasing competition from China. Following the Perseverance rover’s discovery of a unique rocky formation on Mars in July 2024, which suggested the possibility of alien life, NASA initiated plans to retrieve samples for Earth-based analysis. However, the project is now struggling for funding and support, leaving these promising specimens stranded on the Martian surface. This predicament has allowed China to gain a competitive edge in the race to uncover evidence of life beyond Earth, as they pursue their own ambitious missions to explore Mars, further complicating the geopolitical dynamics of space exploration. As both nations push the boundaries of technology and exploration, the implications for future discoveries are monumental.
Source: The Download: Pokémon Go to train world models, and the US-China race to find aliens via MIT Technology Review