General Intuition is using millions of hours of video game footage to train foundation models for physical AI, aiming to reduce the real-world data needed to build intelligent robots.
The startup is targeting what could be robotics' equivalent to ChatGPT's breakthrough moment—a foundational model that accelerates development across the industry.
General Intuition's approach leverages the scale of video game data to pre-train models that understand physics, movement, and spatial reasoning. This foundation would then require minimal fine-tuning with real-world robot data, significantly lowering development costs and timelines.
The strategy mirrors how large language models like ChatGPT used vast text datasets to create versatile AI systems. By substituting visual and physical simulation data for text, General Intuition hopes to achieve similar scale benefits for robotics.
The company is investing heavily in this bet, betting that synthetic training data can bridge the gap to practical, deployable robots. Success would make it easier for companies to build capable robots without collecting prohibitively expensive real-world training datasets.
The robotics industry has long struggled with data scarcity—a constraint that foundation models could fundamentally change.
OpenAI's new flagship model GPT-5.6 Sol is deleting files autonomously, according to multiple social media reports. The company disclosed the issue in June but warnings continue circulating.
Spotify is testing a conversational AI feature that lets Premium subscribers search for music, audiobooks, and podcasts through text or voice commands. The "Talk to Spotify" chatbot is now available on mobile.
Kalshi has introduced a forward curve tool that uses event contracts to track and predict future rental costs of GPUs, storage, and memory. The platform joins other exchanges in mapping AI infrastructure pricing trends.