Overview of artificial intelligence in Google search to be fully released this week

Google announced at its GoogleI/O developer conference at Shoreline Amphitheater in Mountain View, Calif., that Artificial Intelligence Overview, the next evolution of search-generated experiences, will be available in the U.S. starting on May 14 and more soon. National launch. Google showed off several other changes to Google Cloud, Gemini, Workspace and more, including AI actions and summaries that can run across apps, providing interesting options for small businesses.

Google search will include an overview of artificial intelligence

AI Overviews is an extension of the Google search generation experience where AI-generated answers appear at the top of Google searches. You may have seen SGE in action, as some US users have been able to try it since last October. SGE can also produce images or text. AI Overview adds AI-generated information to the top of any Google search engine result.

With AI Overview, “Google will do the work for you. Instead of piecing together all the information yourself, you can ask questions and “get answers immediately,” Liz Reid, Google’s vice president of search, said at I/O.

Reid said that by the end of this year, Artificial Intelligence Overview will be available to more than 1 billion people. Google hopes to answer “ten questions in one” and link tasks together so that artificial intelligence can make accurate connections between information; this can be achieved through multi-step reasoning. For example, someone could ask not only about the best yoga studio in the area, but also about the distance between the studio and their home and the studio’s introductory offer. All of this information will be listed in a handy column at the top of the search results.

Soon, AI Overviews will also be able to answer questions about videos fed to it.

An overview of the usability of artificial intelligence

AI Overviews will begin rolling out in the United States on May 14 and is expected to be available to everyone in the country within a week. Availability in other countries is “coming soon.” Previously, an overview of artificial intelligence has been provided in Search Labs.

Is the AI ​​overview useful?

Will the Artificial Intelligence Overview Really Make Google Search More Useful? An AI overview may diminish the usefulness of a search if its answers prove to be incorrect, irrelevant, or misleading. Google says it will carefully keep track of which images are generated by artificial intelligence and which come from the web.

Gemini 1.5 Pro gets upgrades, including 2 million contextual windows for specific users

Google is expanding Gemini 1.5 Pro’s contextual windows to 2 million for select Google Cloud customers. For a wider contextual window, join the waitlist in Google AI Studio or Vertex AI.

Google CEO Sundar Pichai said that the ultimate goal of the team working on expanding the Google Gemini context window is “infinite context.”

Gemini 1.5 Pro, Google’s large language model, is getting API quality improvements and a new version, Gemini 1.5 Flash. New features available to developers in the Gemini API include video frame extraction, parallel function calls, and context caching. Native video frame extraction and parallel function calls are now supported; context caching is expected to drop in June.

Now available globally, the Gemini 1.5 Flash is a smaller model focused on fast response times. Users of Gemini 1.5 Pro and Gemini 1.5 will be able to enter information for AI to analyze in 1 million contextual windows.

Gemma 2 parameter size is 27B

Google’s small language model Gemma will undergo a major overhaul in June. Gemma 2 will have a 27B parameter model in response to developer requests for a larger Gemma model, but the model is still small enough to fit into compact projects. Google said that Gemma 2 can run effectively on Vertex AI’s single TPU host.

In addition, Google launched PaliGemma, a language and vision model for tasks such as image captioning and image-based questioning. PaliGemma is now available in Vertex AI.

Features like Gemini aggregation will be added to Google Workspace

Google Workspace has received several AI enhancements enabled by Gemini 1.5’s long context windows and multimodality. For example, users can ask Gemini to summarize long email threads or Google Meet calls.

Gemini will be available next month in the desktop Workspace side panel for businesses and consumers using the Gemini for Workspace add-on and the Google One AI Premium plan. Gemini side panels are now available for Workspace Labs and Gemini for Workspace Alpha users.

Several new Gemini features will be available to Workspace and AI Advanced customers going forward, starting this month for lab users, with general rollout in July:

• Summarize email threads.
• Q&A in your email inbox.
• Use longer suggested replies in Smart Replies to extract contextual information from email threads.

Gemini 1.5 can establish connections between applications in Workspace, such as Gmail and Docs. Aparna Pappu, vice president and general manager of Workspace at Google, demonstrated this by showing how small business owners can use Gemini 1.5 to organize and track their travel receipts in an email-based spreadsheet. This feature “Data Questions and Answers” will be launched to laboratory users in July.

Next, Google will add virtual teammates to Workspace. Virtual teammates will have identities, workspace accounts, and goals just like their AI colleagues (but won’t require PTO). Employees can ask questions about their work to assistants, who will have access to the “collective memory” of the teams they work with.

Google has yet to announce a release date for Virtual Teammate. The company plans to add third-party features to it in the future. This is just speculation, but Virtual Teammate could be particularly useful to enterprises if it connects to a CRM application.

Voice and video capabilities coming soon to Gemini app

The Gemini app will launch voice and video capabilities later this year. Gemini will be able to “see” through your camera and respond instantly.

Users will be able to create “Gems,” custom agents that do things like act as personal writing coaches. The idea is to make Gemini a “real assistant” that can, for example, plan trips. Gems will be coming to Gemini Advanced this summer.

The timing of Gemini’s addition of multimodality is interesting compared to the ChatGPT demonstration using GPT-4o earlier this week. Gemini and ChatGPT have very natural-sounding conversations. ChatGPT responded to the outage but misread or misinterpreted some of the circumstances.

See: OpenAI shows how the latest version of its GPT-4 model responds to live video.

Imagen 3 has improvements in generating text

Google has released Imagen 3, the next evolution of its image-generating artificial intelligence. Imagen 3 is designed to better render text, which has been a major weakness of AI image generators in the past. Today, select creators can try out Image 3 in ImageFX at Google Labs, and Imagine 3 is coming soon to developers at Vertex AI.

Google and DeepMind showcase other creative AI tools

Another creative artificial intelligence product announced by Google is Veo, their next-generation generative video model from DeepMind. Veo is available to select creators starting May 14 in VideoFX, an experimental tool on labs.google. Google plans to add Veo to YouTube Shorts and other products at an unspecified date.

Other creative types may want to use the Music AI Sandbox, a set of generative AI tools for making music. No public or private launch date for the Music AI Sandbox has yet been announced.

Sixth-generation Trillium GPUs enhance capabilities in Google Cloud data centers

Pichai launched Google’s sixth-generation Google Cloud TPU, called Trillium; Google claims that the TPU is 4.7 times more efficient than the previous generation. Trillium TPU is designed to add higher performance to Google Cloud data centers and compete with NVIDIA’s AI accelerators.

Time on Trillium will be available to Google Cloud customers by the end of 2024.

TechRepublic reports from Google I/O.