Self-Supervised Learning

TL;DR Current AI systems do not learn autonomously after deployment. The usual process includes fine-tuning, in which a human carefully selects data for a specific task and tunes the model on it. This paper proposes combining learning from observation (System A), learning from action (System B), and a meta-controller (System M) that decides what and how to learn. Why Current AI Systems Do Not Learn The current AI paradigm focuses on scaling LLMs....

TLDR The DINO series advances self-supervised learning for vision transformers through iterative architectural and data refinements. DINOv1 introduces student-teacher distillation on ImageNet-1k. DINOv2 scales to 142M curated images with patch-level objectives. DINOv3 reaches 1.7B Instagram images with register tokens, new Gram matrix based loss, with a custom 7B-parameter ViT, achieving state-of-the-art performance on dense prediction tasks (like instance segmentation) while maintaining a frozen backbone. Motivation Supervised pretraining on ImageNet has dominated vision models, but manually annotating large datasets is expensive and constrains representation quality to label granularity....