A year ago, the Google I/O conference was an opportunity for the American giant to reassure those who thought it was overwhelmed by the incredible success of ChatGPT. This year, the opening keynote of its annual developer conference allowed it to go on the attack. “There are still so many opportunities ahead of us,” promised Sundar Pichai, its general director.
Main announcement: the large-scale deployment of generative artificial intelligence within its search engine. These new features, tested since the beginning of 2023, will be accessible this week to all American users. Other countries will follow “shortly”. By the end of the year, a billion people will have access to it.
Advertisement
Concretely, a new module, called “AI overview”, will be displayed at the top of the results page for certain searches, above the traditional links. It will include an answer provided by Gemini, Google's large language model. This structured text, citing several sources, will notably replace the “knowledge graphs”, the small extracts from websites displayed until now to answer certain questions.
A rival designed by OpenAI?
Beyond that, Google promises “new, completely redesigned search experience”. The site will, for example, be able to plan a trip using Google Maps and by fetching necessary information (hotel, flight) directly from the users' Gmail account. For “inspiration” searches, such as suggestions for restaurants, movies, recipes or products, the AI will generate the entire page of results, sorting them into different categories. Finally, it will also be possible to ask a question with a video, for example to understand why an electronic device no longer works.
Advertisement
Despite the emergence of generative AI, Google has taken its time to change how its engine works – unlike Microsoft, which very quickly integrated AI-generated responses into Bing. Interest around start-up Perplexity.AI, which claims more than 10 million monthly users, and rumors around the launch of a search engine by OpenAI, the designer of ChatGPT, has certainly accelerated its timetable.
Faced with these new competitors, Google will be able to count on its brand image and the well-established habits of Internet users. The group also highlights its database of more than “1000 billion pieces of information about people, places and things”, explains Liz Reid, the new director of Google Search. And also on his “incomparable ranking system” which should make it possible to provide the most relevant information.
Gemini everywhere
These changes, however, represent a threat to Google's business model, which relies on sponsored links that companies buy to appear at the top of search results. They could in fact result in a drop in advertising revenue, while causing an increase in costs because of the computing resources necessary to run the large language models. According to Financial Timesthe company is considering paying for access to certain AI-powered features.
During the two hours of presentation, entirely devoted to artificial intelligence, Google also announced the deployment of Gemini on numerous in-house services. The large language model will notably be added to the next generation of Android, to replace the current Google Assistant. Integrated at the operating system level, it will be able to interact with different applications and also listen to telephone conversations. Some of these new options will be able to run locally, without using a cloud platform. Gemini will also arrive on Google Photos, to offer even more advanced search options. And on the Workspace suite (Gmail, Drive, etc.) but only for paying users.
New glasses?
Google also unveiled new AI models. First, a Flash version of Gemini, announced as faster and more efficient for certain requests (summarizing a conversation, captioning photos, etc.). Then, Imagen 3, the latest version of its image generator. And finally Veo, a model capable of creating videos, rival of OpenAI's Sora, currently only available in the testing phase.
Finally, the company also lifted the veil on the Astra project, supposed to represent “the future of AI assistants”. In a presentation video, filmed in real time, a user interacts with an AI using her smartphone camera to obtain information. Google promises fast response times, unlike the AI Pin, the AI pin designed by the start-up Humane. The video ends with a possible teaser: connected glasses…
Selected for you