Nvidia unveils in January 2025 new products such as chips to better train robots and cars. These will also enhance the gaming chips and its first desktop computer.
Cosmos foundation models generate photorealistic video which can be used to train robots and self-driving cars. This will be more economical than using conventional data.
These models create synthetic data for training. They make the robots and cars understand the physical world similar to the way LLMs have helped chatbots to generate responses in natural language.
Today the data is generated by putting the cars on the road to gather video. Robots are taught by humans to teach repetitive tasks. Cosmos takes a text description to generate a video that obeys the laws of physics.
Cosmos is available on open license similar to Meta’s Llama3 language models.
Nvidia also unveiled gaming chips that use Nvidia’s Blackwell AI technology. These chips are called RTX50 series. They make images more realistic. There are more accurate human faces.
Toyota will use its Orin chips and automative operating system to provide advanced driver assistance in several models.
Leave a Reply