Google I/O 2024: AI Overviews Go Live and Other Product Announcements

The future of search is decidedly getting an AI-powered reboot. The tech giant stunned everyone at its annual Google I/O 2024 event on May 14th with its revolutionary AI-powered search capabilities.

Will AI Overviews finally go live?

The event’s highlight was the long-awaited launch of SGE under the “AI Overviews” moniker. While AI Overviews will debut first in the U.S. market, Google guaranteed an eventual expansion, stating “more countries coming soon.” 

To learn more about SGE’s early stages, check out our article: Inside Google SGE: Uncovering the beta version of AI-driven search.

Danny Sullivan, a Google spokesperson, assured that AI Overviews is near launch-ready. He also mentioned that users engage more with search results and report higher satisfaction rates when AI Overviews are available.

AI Overviews will begin rolling out to everyone today in the US, with more countries coming soon. They provide both a quick overview of a topic and links to learn more. We’ve found that with AI Overviews, people use Search more, and are more satisfied with their results.…

— Google SearchLiaison (@searchliaison) May 14, 2024

This is an exciting announcement, although there is a tiny “but” worth considering. 

SGE previously generated results with “questionable quality,” as highlighted by Barry Schwartz. 

That being said, we hope Google will make more progress in optimizing its AI capabilities to enhance the AI’s output quality and align it with ethical principles.

By the way, OpenAI’s recent GPT-4o release has shown promising improvements, setting the new standard for AI language models. Check it out!

What capabilities will AI overviews have? 

Well,  the announcement that AI Overviews is now going live was just the appetizer of this event. Google offered a full multi-course menu of impressive AI search innovations that AI Overviews is capable of. Initially, these features will only be available through Google’s Search Labs for the U.S. market in English. 

Here are some of the features they revealed:

More Google I/O 2024 highlights 

Gemini 1.5 Pro enhancements

Google has upgraded its Gemini AI model to handle much larger inputs, including lengthy documents, code, videos, and audio files. They announced that the new Gemini 1.5 Pro will have a 2 million token context window later this year. This will greatly expand its ability to understand long-form content. It also processes more than any other commercial AI model on the market.

Project Astra: Google’s multimodal AI assistant

Google also announced Project Astra, a multimodal AI assistant capable of engaging in semi-conversational interactions while using the device’s camera to identify objects, people, and their actions in real time. The AI assistant is part of Google’s Gemini project and is currently in its early stages of development.

You can watch the video of Project Astra in action here. 

Imagen 3 update

Next up in Google’s string of paradigm-shifting announcements was Google’s launch of Imagen 3, its most advanced image generation model. Imagen 3 creates detail-rich, realistic images with fewer flaws than earlier models. This model understands natural language better and can include small details from longer prompts, enabling it to work well with different styles and generate text more accurately. 

Google has made the model available to some creators through a private preview on ImageFX and plans to offer it soon on Vertex AI. The public can join the waitlist for early access. 

Trillium TPU

Google announced Trillium, the sixth and latest generation of its custom AI accelerator, the Tensor Processing Unit (TPU). Trillium TPUs deliver a 4.7x boost in peak compute performance per chip compared to Google’s previous TPU v5e version. Trillium has already become Google’s most sustainable TPU generation, with over 67% greater energy efficiency than TPU v5e. 

Google plans to make Trillium TPUs available to cloud customers later this year.

Ask with Video in Google search

Google also gave a live demonstration of their groundbreaking new search feature that showcases advancements in video understanding. Users can now ask questions in video format. Search can analyze the complex visuals, understand your query, provide explanations and next steps, and offer relevant resources through an AI-generated overview.

Gemini mobile app

Google is enhancing the Gemini app with a slew of new capabilities, including: 

Gemma

Google is improving its Gemma language models at scale. The next generation, Gemma 2 (to be launched in June), will feature a 27-billion-parameter model. Google has massively expanded its capabilities with this large model size, which Nvidia has optimized to run efficiently across Google’s AI infrastructure, including next-gen GPUs and TPU hosts in Vertex AI. 

LearnLM

Google has introduced LearnLM, a series of AI language models optimized for educational purposes. These Gemini-based models are currently being used to improve learning features in several Google products, such as Search, YouTube, and Google Classroom. The majority of its features, however, are not available yet. 

Few final remarks

The advancements introduced during Google I/O 2024 are a clear indication that the AI revolution is in full swing. The way we search, learn, and interact with technology will never be the same. From AI Overviews and video searches to custom AI assistants, Google is pushing the boundaries of what’s possible. 

The future is now. Are you ready to embrace it?

Written by

Anna Postol

Anna Postol is a content marketer at SE Ranking, where she finds joy in creating content for SEO, digital marketing, and social media. Her focus is on writing accurate, easy-to-digest, informative pieces that turn complex topics into engaging narratives that resonate with readers. Outside of work, she enjoys stretching exercises, workouts, planning her next travel adventure, and spending time with her beloved cat-daughter.