Google is extending its Gemini multimodal foundation model to its entire product line, including Search and Android phones.
The tech giant, at its Gemini multimodal foundation model on Tuesday, unleashed a blizzard of new and updated generative AI-infused products and services for not only developers, but also enterprises, consumers, filmmakers, musicians, learners and educators.
"Today, all of our 2 billion user products use Gemini," Google CEO Sundar Pichai said in his opening keynote. "One of the most exciting transformations with Gemini has been in Google Search. People are using it to search in entirely new ways and asking new types of questions or more complex queries, even searching with photos and getting back the best the web has to offer."
Gemini upgrades
Google made generally available its Gemini 1.5 Pro model, which has been in preview for developers for several months. It boasts a 1 million token context window that allows it to ingest large volumes of information. The "long context" window is also now available for consumers in Gemini Advanced, in 35 languages.
Gemini 1.5 Pro, with a 2 million token context window, became available in private preview Tuesday.
The vendor also unveiled a new Gemini mobile app and put native generative AI technology on the operating system used by Google Pixel phones, tablets and other Android mobile devices. The move likely puts more pressure on Apple to improve the iPhone's native AI capabilities, which Apple is expected to do soon.
Google also unveiled new AI chips and AI Agents, or assistants, to help users plan and complete various tasks, such as finding receipts, locating order numbers and filing product return forms.
Generative AI race
Google's flurry of generative AI releases came a day after OpenAI unveiled GPT-4o, a faster, more interactive update of its biggest large language model.
"It's very clear that they're expanding AI across their portfolio, and it's no longer viewed as a separate thing that people would use," said Paul Nashawaty, an analyst at the Futurum Group. "Now it's in everything. It's AI everywhere."
While Google's moves are not a direct response to OpenAI, or Microsoft -- which has pursued a similar strategy by adding its Copilot AI assistant to many of its products -- Google nevertheless signaled that it is ready to match any advance in generative AI.
In the AI arms race, Google has definitely leveled up.
Johna Till JohnsonCEO and founder, Nemertes
With Gemini, Google is drawing on its unique access to the vast troves of data about virtually everything on the internet to enable users to search for information about nearly anything on any device, using voice, text or video, noted Johna Till Johnson, CEO and founder of Nemertes Research.
"In the AI arms race, Google has definitely leveled up," she said. "They are clearly taking advantage of their strength, which is having world data."
However, the search giant still offers few guarantees about the accuracy or veracity of the information it scours, curates and serves up to users, Johnson said. Nor is it clear that enterprises or other users have signaled they need generative AI in all their applications, she said.
"My question is, from an enterprise perspective. I don't think this strategic direction for Google marks a great advance in terms of what people actually need AI for," Johnson added.
More AI updates
The main generative AI update to Search -- the second step Google has made to add the technology to its most-used service, after Search Generative Experience -- is AI Overviews.
AI Overviews, which produces information summaries shown above traditional Search results, will be available to Search users in the U.S. this week. It will become available around the world by the end of the year, according to Google.
In Google Photos, users as of this summer will be able to search for photos by asking detailed questions, going beyond traditional keyword search.
Google also relied on Gemini to build a family of models designed for learning. Introduced at Google I/O on Monday, the LearnLM AI models are based on learning science principles including active learning, managing cognitive load, adapting to the learner, stimulating curiosity and deepening metacognition, according to the vendor.
Beyond new enterprise and consumer releases and updates, the Google DeepMind research unit contributed to the slew of AI moves.
Google's VideoFX -- an experimental tool powered by Veo, Google DeepMind's new generative video model -- is aimed at creatives such as film and video makers who can try out and produce video sequences with text prompts.
The tech giant also released Imagen 3, the latest version of its highest-quality text-to-image model.
Also among the new and updated products is a new generation of Google's line of Gemma open models aimed at responsible AI innovation; a project to use AI technology in red teaming new products; and an expansion of SynthID, Google's digital watermarking tool, to text and video.
Meanwhile, there's a practical advantage for Google in extending AI and generative AI technology across its entire universe of products and services, according to Nashawaty, of the Futurum Group.
"It's using AI as an enabler to drive product adoption," he said.
Shaun Sutner is senior news director for TechTarget Editorial's information management team, driving coverage of artificial intelligence, unified communications, analytics and data management technologies. He is a veteran journalist with more than 30 years of news experience.
