The research lab is behind Gemini, Google’s family of AI models that can understand and generate text, code, images, audio and videos.