Artificial intelligence · 2021
CLIP
Images and text learn a shared space from 400 million captioned web pictures.
OpenAI's Contrastive Language-Image Pre-training matched images to their captions, giving zero-shot classification competitive with supervised ResNets. CLIP became the text-to-image steering signal for DALL-E 2 and Stable Diffusion.
What it built on 2
What it led to 3
Related measurements
The dashed line marks when it first worked.
Training compute of notable AI models Source: Epoch AI, Notable AI Models (CC BY 4.0)
ML hardware FLOP/s per dollar Source: Epoch AI, ML hardware (CC BY 4.0)