The possibilities are broader if generative AI is trained on other data types, not just text. Some models are already capable of training to some extent on unlabelled videos or images.

Personally identifiable information has been found in DataComp CommonPool, one of the largest open-source data sets used to train image generation models.

DeepSeek-R1 released model code and pre-trained weights but not training data. Ai2 is taking a different approach to be more open.

Former OpenAI researcher reveals company's questionable data collection practices that may have violated copyright raising ethical concerns about AI development ...

Ai2’s MolmoAct open-source robotics system brings 3D reasoning and real-time planning to robots, offering a transparent alternative to black-box models.

On opening day, RealMan unveiled the RealBOT Embodied Intelligence Open Platform, which it designed for high-quality data acquisition.