In the middle of a battle against his Microsoft rival, Google unveiled a series of features enriched by artificial intelligence on Wednesday, but an error in an advertisement for “Bard” expected dear to its action.
MO12345LEMONDE with AFP
Robots also make mistakes. To the question: “What last NASA discoveries from the James-Webb telescope can I explain to my 9-year-old child?”, The brand new Google, Bard conversational robot made a blunder. Google’s parent company paid for it by losing more than 7 % on the New York Stock Exchange on Wednesday, February 8.
During an event organized in Paris on Wednesday in front of several dozen European journalists, Google announced that it has launched tests to integrate artificial intelligence (AI) called “generative” in online research, while Microsoft comes To present a new test version of its own search engine, Bing, with its algorithms and those of the Chatgpt conversational robot. For the Californian group, it is a question of not giving in the field to its rival, after two decades of almost without sharing domination of this lucrative sector.
But astronomers reported on Twitter a bard error in an advertisement for the new software. The announcement shows an example of a question asked in Bard, and the latter declares among the answers, that the James-Webb space telescope was the first to take photos of a planet out of the solar system, while the European Giant Telescope had actually already done, in 2004. “We still need massive tests,” recognized Prabhakar Raghavan, vice-president in charge in particular of the search engine, before the blunder is surface.
Current AI technology does not guarantee 100 % fair responses
Three of Google’s executives had made the trip to Paris, a way of giving an international and multilingual dimension to its AI projects. The company did not give precision on the way in which Bard would be integrated into its search engine.
Like his rival the day before, Prabhakar Raghavan considered that the integration of AI would constitute “a new era of research”, the very words of the boss of Microsoft, Satya Nadella. But he did not want to give a delay for a consumer version, repeating that it would be “in a few weeks”.
“It will be when we are satisfied with the result,” he added cautiously, allusion to the many chatgpt setbacks and other generative artificial intelligences, to the sometimes absurd or displaced responses. He also assured not only that AI would not give “a single good answer”, but also that it would not discourage Internet users from going to other sites.
Analysts fear an impoverishment of the amount of information but also of advertising revenues from all websites. “We always want people to explore the web,” said Elizabeth Reid, also responsible for the search engine. “People will always want information from people with whom they can identify. There will always be beauty influencers.”
Google also did not specify if the responses of its AI would give the sources of its responses, and recognized that the current AI technology did not guarantee 100 % fair responses. The group assured that he did not give in to the pressure of Microsoft and the worldwide success of Chatgpt to accelerate its announcements. “We have been working on AI for years,” said its leaders. “No particular event” pushed us to make his announcements now, they said.
The technology giant has also shown augmented reality features during the Parisian event, as well as new 3D representations generated from its street images, new possibilities for finding information on photos and a new application of research in works of art.