AI began perceiving a situation as a whole: seeing a screen, hearing speech and responding almost in real time. Robust understanding of long video and causal relationships remains ahead.
About this eventA shared space for text and images — the foundation all image generation grew on.
CLIP learned to match images with internet text descriptions, placing two modalities in a shared representation space. It enabled visual categories to be specified with language instead of task-specific labels and became an important foundation for later generative and multimodal systems.
Source: OpenAIAbout this eventSpeech recognition reaches the level where audio becomes an ordinary input.
About this eventOne model learned to segment arbitrary objects from a point, box or prompt without training for each class.
About this eventVision as a standard part of a frontier model.
About this eventGoogle introduced a model family designed from the start to work across text, images, audio and video.
About this eventOne model began working across text, images and audio in real time.
About this eventAn hour of footage fits, but following the plot and causes inside it does not work yet.
About this eventModalities with neither cheap sensors nor large datasets so far.
Primary material behind this dossier: papers, lab publications and official reports.