Google I/O heavily featured generative AI in its annual showcase of new and upcoming products and features. We’ve put together a list of the generative AI highlights from the avalanche released by Google executives and their slickly produced preview videos. That’s before even considering the Pixel smartphones, tablets, and other hardware introduced in the same keynote. Look for more detailed breakdowns of some of these highlights soon.
Google AI and LLMs
- PaLM 2
- The updated version of the PaLM large language model is more powerful and flexible.
- PaLM 2 is trained on data in more than 100 languages and allows collaboration on coding, mathematics, and creative writing projects.
- Related Med-PaLM 2 LLM aimed at diagnosing medical conditions and reached 85% accuracy on the US Medical Licensing Examination.
- Next-generation multimodal foundation model Gemini is currently undergoing training.
- Google’s conversational generative AI LLM is now accessible to the general public in over 180 countries, has learned Japanese and Korean, and is on the road to speaking 40 languages.
- Expanding visual capabilities are coming soon, including images in answers and the option to include images in prompts by leveraging Google Lens to get captions for pictures and other commentaries.
- Adobe Firefly generative AI models will run Bard’s text-to-image feature and streamline editing in Adobe Express.
- Integrations with other third-party services in the works, including Kayak, OpenTable, Instacart, Wolfram, and Khan Academy.
- Bard chat is making citations more precise and adding the ability to export and run code with Replit, starting with Python.
- Vertex AI
- Three new foundation models are available via API, Generative AI Studio, or data science notebook.
- Text-to-code foundation model Codey can be added to SDKs or applications for code creation and editing.
- Text-to-image foundation model Imagen provides synthetic media production and has been in a closed demo for year.
- The new Chirp speech-to-text foundation models designed as a kind od universal translator for customer service and voice-enabled AI assistants.
Search and Creation
- Google Search
- Google will include AI-powered snapshots to search results and with question summaries and conversational mode to continue the search without the AI forgetting relevant context.
- Google Photos
- Google Photos is launching a new Magic Editor tool with the ability to move and edit objects, extending the Magic Eraser tool and allowing manipulation of lighting and other elements.
- The new “About this Image” tool evaluates visual content to trace its history and whether it might be a deepfake.
- Google Duet
- Portfolio of generative AI features for Google Workspace labeled Google Duet.
- New Gmail “Help Me Write” tool will draft emails, including automatically filling in names and other relevant information.
- Text-to-image tool will be embedded in Google Slides.
- Google Sheets will automate data classification and create custom plans.
- Google Meet leverages generative AI to make unique backgrounds and mask a user’s surroundings.
Follow @voicebotaiFollow @erichschwartz
Google Teaches Generative AI Chatbot Bard to Write and Explain Software Code
Google Demos New Conversational AI Model and Opens AI Test Kitchen
Google Assistant Unveils New ‘Capabilities’ and ‘Widgets’ for Developers at Google I/O
The post 20+ Generative AI Features Announced at Google I/O appeared first on Voicebot.ai.
Read the full article here