Google allegedly knew that OpenAI was using transcripts of YouTube videos to train its AI, but didn't say anything because the company was doing the same thing.

YouTube CEO Neil Mohan recently gave an interview to BloombergJoin those teaching AI with YouTube videos. The message was aimed primarily at OpenAI: the company trained the AI ​​of its new video maker, Sora, through content on the platform, thus showing them what the real world looks like.

See also  Index - Economy - A plant that fills oxygen cylinders is being built in Donaugváros

Regarding the topic now A New York times She provided new details, according to which the Mohan family will also have something to sweep around their house. According to sources familiar with the company's inner workings who requested anonymity, both OpenAI and Google trained their AI using transcripts generated from YouTube videos. By doing so, they could have violated YouTube's terms of use and the rights of content producers.

ChatGPT has been removed, and there is a new king of AI

Among other things, Amazon and Google's investments, which can be measured in billions of dollars, were behind the development of CLOUD, America's huge linguistic model called Anthropic. They put in the money because they thought this platform could dethrone ChatGPT. And now it happened.

According to the paper, OpenAI used speech recognition tool Whisper to transcribe more than a million hours of YouTube videos into text format, and used the resulting transcripts to train GPT-4. the the information He also previously wrote about OpenAI's use of data extracted from YouTube.

Matt Bryant, Google spokesperson A New York timesIt said that such use of the contents is prohibited, and the company was not aware that OpenAI was training its system on such contents. According to the American newspaper, at the same time, there were people in the company who knew exactly what was happening, but they did not say anything about it, because Google did the same thing by teaching its own artificial intelligence. According to Bryant, this is indeed the case, except that the company only used the videos of users who gave their permission for this purpose.

See also  INDEX - ECONOMY - Potato pasta can get expensive

The source who spoke to the newspaper also said that Google changed its terms of use in June 2023, giving the company the opportunity to train artificial intelligence on documents and tables created by users in the Google service. According to Bryant, only documents from users who participated in testing Google's experimental features are used for this purpose.

If you want to know about similar things at other times, like it HVG Tech Department Facebook page.