5 We shared Astra project: Our vision of the future of AI Assistants.
6. We have announced TrilliumThe sixth generation of our customized accelerator, the tensor processing unit (TPU). It is the most efficient TPU to date.
7 Compared to TPU V5E, Trillium TPU reaches a 4.7x increase in advanced calculation performance per chip.
8 They are also our most durable generation: Trillium Tpus is greater than 67% more energy efficient compared to TPU V5E.
9. And we have demo a first audio overview prototype to NotebookWho uses a collection of materials downloaded to create a personalized verbal discussion for the user.
10 We have announced that putting the ground with Google Search – a tool that connects the Gemini model with global knowledge, a wide range of subjects or up -to -date information on the Internet – is now Usually available on the AI summit.
11 We have added an audio understanding in the API Gemini and the AI studio, so Gemini 1.5 Pro can now reason through the image and audio for videos downloaded in AI Studio.
12 Starting with Pixel, applications using gemini nano with multimodality will be able to understand the world as people – not only by text entrance but also by sight, sound and language spoken.
Generative media models and laboratory experiences
13 We have announced Imagen 3Our highest quality image generation model to date.
14 Imagen 3 includes natural language and intention behind your prompts and incorporates small details from longer prompts. This helps him generate an incredible level of detail, producing photorealist and realistic images with visual artifacts much less distracting than our previous models.
15 Imagen 3 is also our best model to date to make the text – a challenge for image generation models.
16 We have deployed Imagen 3 to trust testers in ImageFX and you can register to Join the waiting list.
17 Imagen 3 will also arrive at Vertex ai This summer.
18 Then we announced VeoOur most competent video generation model to date. It generates high quality 1080p resolution videos that can go beyond a minute, in a wide range of cinematographic and visual styles.
19. We will also bring some of the VEO capabilities to YouTube shorts and other products in the future.
20 We have shown what Veo can help artists to do by collaborating with filmmakers – including Donald Glover, who experienced VEO For a film project.
21 We have highlighted Music Ai SandboxA series of musical AI tools that allow people to create new instrumental sections from zero, to transfer styles between trackers and much more. You can find new songs from these collaborations – including one Jean Wyclef and another of Marc Rebillet – On Youtube now.
22 And be sure to check Infinite WonderlandAn experience where Google artists and creatives have experienced together to refine an AI model to constantly reinvent the visual world of the novel “Alice’s Adventures In Wonderland”. Readers of Infinite Wonderland Can generate apparently infinite images for each of the 1,200 phrases in the book according to the respective style of each artist.
23 We announced VideofxOur latest experimental tool that uses Google Deepmind’s generative video model, Veo, and allows you to transform an idea into a video clip.
24 It is also delivered with a storyboard mode which allows you to make the scene by scene and add music to your final video.