Last May, Sundar Pichai, Google’s CEO, said the company would use artificial intelligence to reimagine all of its products.
But because the new generative AI technology carries risks, such as spreading false information, Google has been cautious about applying the technology to its search engine, which is used by more than two billion people and was responsible for $175 billion in revenue last year.
On Tuesday, at Google’s annual conference in Mountain View, Calif., Mr. Pichai showed how the company’s aggressive work on artificial intelligence has finally permeated the search engine. Starting this week, he said, US users will see a feature, AI Overviews, that generates informative summaries on top of traditional search results. By the end of the year, over one billion people will have access to the technology.
AI Overviews is likely to heighten concerns that web publishers will see less traffic from Google Search, putting more pressure on an industry rocked by disruptions to other technology platforms. On Google, users will see longer summaries on a given topic, which may reduce the need to visit another website – although Google downplays these concerns.
“Links included in AI Overviews receive more clicks” from users than if they were presented as traditional search results, Liz Reed, Google’s vice president of search, wrote in a blog post. “We will continue to focus on sending valuable traffic to publishers and creators.”
The company also unveiled a slew of other initiatives — including a lightweight AI model, new chips and so-called agents that help users complete tasks — in an effort to gain the upper hand in an AI slugfest with Microsoft and OpenAI, the maker of ChatGPT.
“We’re in the very early days of the AI platform shift,” Mr. Pichai said Tuesday at Google’s I/O developer conference. “We want everyone to benefit from what Gemini can do,” including developers, startups and the public.
When ChatGPT was launched in late 2022, some in the tech industry saw it as a serious threat to Google’s search engine, the most popular way to get information online. Since then, Google has worked aggressively to regain its edge in AI, releasing a family of technologies called Gemini, including new AI models for developers and chatbots for users. It also implemented the technology in YouTube, Gmail, and Docs, helping users create videos, emails, and drafts with less effort.
All the while, the competition between Google and OpenAI and its partner Microsoft continues. A day before the Google OpenAI conference, it presented a new version of ChatGPT, which is closer to a voice assistant.
(The New York Times sued OpenAI and Microsoft in December for copyright infringement on news content related to artificial intelligence systems.)
At its Silicon Valley event, Google showed how it will bring AI deeper into users’ lives. He introduced Project Astra, an experiment to see how AI can act as an agent, vocally talking to users and responding to images and videos. Some of the capabilities will be available to users of Google’s Gemini chatbot later this year, Demis Hassabis, CEO of DeepMind, Google’s artificial intelligence lab, wrote in a blog post.
DeepMind also introduced Gemini 1.5 Flash, an AI model designed to be fast and efficient but lighter in size than Gemini 1.5 Pro, the mid-range model that Google has implemented in many of its consumer services. Dr. Hassabis wrote that the new model is “very capable” of reasoning and is good at summarizing information, chatting and adding captions to images and videos.
The company announced another AI model, Veo, which generates high-definition videos based on simple text prompts, similar to OpenAI’s Sora system. Google said some creators can preview Veo and that others can join a waiting list to access it. Later this year, the company expects to bring some of Veo’s capabilities to YouTube Shorts, video platform competitor TikTok, and other products.
Google also showed off the latest versions of its Lyria music generator and Imagen 3 image generator. In February, Google’s Gemini chatbot was criticized by users on social media for refusing to generate images of white people and presenting inaccurate images of historical figures. The company said it will turn off the ability to generate images of people until it fixes the problem.
In the past three months, more than one million users have signed up for Gemini Advanced, the version of Google’s chatbot available through a $20 monthly subscription, the company said.
In the coming months, Google will add Gemini Live, which will give users a way to talk to the chatbot through voice commands. The chatbot will respond in natural-sounding voices, Google said, and users can interrupt Gemini to ask clarifying questions. Later this year, users will be able to use their cameras to show Gemini Live the physical world around them and have conversations with the chatbot about it.
In addition to AI Overviews, Google’s search engine will present AI-curated search results pages with generated headings highlighting different types of content. The feature will start with dining and recipe results, and will later be offered for shopping, travel and entertainment queries.
Ms. Reid, the head of search, said in an interview before the conference that she expected the search updates to save users time because Google “can do more of the work for you.”
Mr Pichai said he expected the majority of people to interact with Gemini AI technology through Google’s search engine.
“We will make it ever more seamless for people to interact with Gemini,” Mr Pichai said at a pre-conference briefing.