2025 is the year AI stops being “mostly text” and becomes truly multimodal and mission-capable. Breakthroughs range from far more capable multimodal foundation models (text + voice + images + video), dramatic progress in text-to-video and Sports Harmonicode video understanding, faster and larger protein-folding models accelerating drug discovery, practical on-device foundation models, and a hardware …
Latest AI Breakthroughs — 2025 Update on Multimodal Models, Video & Science