Putting different kinds of thing into the same space of numbers so they can be compared directly. Once an image and a sentence sit in the same space, you can search photographs with words, or find images similar to other images, without any labels at all.
Technical view
Trained contrastively on paired data so that matching pairs are close and non-matching pairs distant, as in CLIP for image and text, with extensions to audio and video. Enables zero-shot classification by comparing an image against embedded label descriptions, cross-modal retrieval, and conditioning of generative models. Limitations include sensitivity to prompt phrasing for the text side and inheritance of dataset biases from web-scraped pairs.
Related
Part of AI Terminology Index, section: Modalities and Applications.