Part 5/9:
Historically, AI training relied on labeled images to develop machine learning models. However, NVIDIA has shifted the paradigm by utilizing video data—allowing machines to learn directly from video feeds rather than relying solely on manual annotations. This methodology opens new avenues for efficiency and accuracy in AI training, resembling the techniques employed in large language models.
The next generation of AI will need to possess a firm grounding in the physical world, achievable primarily through extensive video training. NVIDIA positions itself as a leader in this space, anticipating a surge in computing demands necessary to support this multimodal training approach.