(Perugia) Shailesh Prakash, general manager of Google News, admitted Thursday that current models of artificial intelligence “are not perfect at all” but “are progressing rapidly”, a few weeks after the suspension of a tool by the American group due to historical inaccuracies.
To understand where we are with AI, “let’s use an analogy with football”: “a game lasts 90 minutes, I would say that we are in the first 5 minutes”, he said during an intervention at the international journalism festival in Perugia, Italy.
“So, it’s very early. These models are progressing rapidly. They are not perfect at all,” added Mr. Prakash, the general manager of Google News (Google’s free online service that automatically presents news articles from the Internet).
In February, shortly after launching Gemini, its new generative AI interface, Google suspended the creation of images of people following numerous user reports. A query on a German soldier from 1943, for example, resulted in images of Asian or black-skinned soldiers.
“This is completely unacceptable and we were wrong,” reacted Google CEO Sundar Pichai.
Last year, Google also indicated that it was developing a new artificial intelligence tool intended to help journalists write their articles, in partnership with several big names in the press.
“We are working hard” and “I am an optimist,” assured Mr. Prakash on Thursday. “These tools are tools with humans in the loop,” he stressed, in the face of fears sometimes expressed about the possible replacement of journalists by machines.
The new tools will allow us to be “more creative, more ambitious”, according to him.
The Google manager explained that he was working on “marking” in particular: it will be possible to consult AI sources to verify that the machine has not had “hallucinations”, inventing elements.
So, “in the end, the responsibility lies with you,” he insisted.