Google has introduced Gemini, a new multimodal artificial intelligence model that it calls the most advanced it has ever released. However, these words alone make it difficult to understand what Gemini is really capable of and how it can be used.

To lift the veil of secrecy over the new model’s capabilities, Google DeepMind created a short video showing Gemini’s work, in which the AI model used a computer vision system to recognize images.

It’s hard to tell how heavily edited the video is because you can see the transition between different frames. But even if Gemini doesn’t always give the right answers, its capabilities are still quite impressive.