What link can be between the search for NASA and Google? The need to understand and the mission of helping. Therefore, Pandu Nayak, former researcher at the American Space Agency and now a scientist in Chief of Google, defines a motivation that explains his strange professional career. For more than 20 years, he has linked his destination with those of the Google search engine. From a keyword reflection until the arrival of AI through the boom in the mobile, he experienced all the disorders, revolutions and innovations that have shaped the search for Google, the reason for being since 1998 of one of the world’s technological giants.
“Our mission is to organize global information and make it universally accessible and useful,” he explains to Tech & Co Pandu Nayak. “And this mission has never changed, even if the tools to achieve it have evolved radically.” Because behind the simplicity of a research bar, it is a technological war machine that Sergey Brin and Larry Page designed at the end of the 1990s, giving birth to Google in the lower part of Susan Wojcicki’s garage.
“The search on Google is based on two fundamental principles: relevance and utility with results that really meet their expectations, but also on the quality and reliability of information. And this is a priority for us from the beginning,” recalls Search’s head of thought.
The revolution is walking
Initially thought as a “simple” search engine, Google Search has become a real multimodal tool to now search images and videos. A tool capable of analyzing sounds, photos and now understanding natural language thanks to AI. Despite “the investigation of one billion received every day”, Pandu Nayak says that it is still amazed: “15% of the applications made every day in Google have never seen each other before. A vertiginous figure that reflects the inexhaustible human curiosity and the way in which it is stimulated.”
For Google chief scientist, constantly improve available tools encourages users to make even more applications. It should be said that, according to a recent fusion water study, we are a social study, one in three adults is mainly asking for search engines. And artificial intelligence brings a part of new characteristics to improve all this.
Since 2023, a new effectiveness for Google has begun with the introduction of general views of AI (seen, it is not yet available in France), automatic summaries to answer complex questions in the search for Google. “Users ask more questions, they spend more time in the results, they explore many more sources. This improves the quality of their understanding,” said the search manager.
In the 230 countries where the function is available, IA descriptions now appear at the top of the answers, long before sites, media articles or official organizations on the subject. Enough to be questioned by observers and users who see it, for some a sufficient answer, for others an answer, whose sources are not clear enough.
In addition, many editors and content creators are also concerned about the use of their texts by AI and the fact that their information is disturbed in the answers. “We are creating a search for users, but a rich and dynamic content ecosystem is essential for them. We both want,” replies the scientist. A balancing position that will undoubtedly be difficult to maintain in the long term, when the sources, deprived of visiting, will see their economic model once more agitated.
The chief scientist hit him for three decades: the quality of the sources is essential for Google, either classified within an initial pageank or by a classification whose “algorithm is changed thousands of times a year” to ensure that the result of the requests is optimal. The best possible and the most varied for “the user chooses the result you like.”
“Arrange for reliable information is our founding value. And today we also help users judge for themselves, with context tools,” he recalls. This refers in particular About this result which allows us to understand the results obtained.
Search while we talk
But artificial intelligence acquires a substantial scale in the operation of the search, of the very admission of one of its architects. “We are witnessing a dazzling growth of deep learning and AI technologies, with new features in the field such as Google Lens or Weusing to Miring. This has opened new visual research perspectives,” summarizes Pandu Nayak. “And obviously there is the current generation of generative and everything that brings as the discovery assisted by AI, the AI stand out, etc.”.
If it becomes preponderant and “incredible in its way of understanding language”, the AI “is not perfect”, with “hallucinations problems inherent to its functioning.” “We have developed techniques to limit hallucinations and maintain a quality equivalent to the best web extracts,” he admits, recognizing that AI is not magical and that everything must be done to be rigorous. Because user’s trust remains the nerve of war. But the arrival of the multimodal, the ability to develop vocally or in writing applications while we talk, the use of the camera to point information and translate it instantly are giant steps in the investigation.
“Before joining Google, 20 years ago, I thought the search worked perfectly and asked me what we could do more to improve it. I would never have thought that we would get there. What we were doing only touched the field of possibilities. With ia, we really go to the upper stage and there are still tons of things that we can do in the future nayak banks. The permanent search for the perfect research has no end …
Source: BFM TV
