How OpenAI Taught AI To See The World

4 Просмотры

  • Спасибо! Поделитесь с друзьями!

    Вам не понравилось видео. Спасибо за то что поделились своим мнением!

aii
Издатель
[Day 9/100] Subscribe so you don’t miss out!

Today we are breaking down "Learning Transferable Visual Models From Natural Language Supervision" (2021), the study that laid the absolute foundation for text-to-image tools like DALL-E and Stable Diffusion.

Before this paper, AI could only recognize a rigid, pre-made list of words that humans spent years manually labeling. This paper introduced CLIP (Contrastive Language-Image Pre-Training), changing the trajectory of technology forever by letting AI learn from 400 million image-text pairs across the web.

Why it matters: It completely eliminated the need for human-labeled datasets, matching industry-standard image recognition models using pure, zero-shot learning.

The Authors: Alec Radford, Ilya Sutskever, Aditya Ramesh, et al.
arXiv: 2103.00020

Subscribe so you don't miss Day 10 tomorrow.

Video by: Sachin Hiriyanna

#ComputerVision #Shorts #Tech #Science #Coding #ResearchPaper #OpenAI #CLIP
Категория
AI Картинки
Выполните вход или зарегистрируйтесь для отправки комментария.
Комментариев нет.