OpenAI’s artificial intelligence tool ChatGPT is making waves all over the world. As soon as this chatbot was launched, big tech companies also started working on similar generative AI models. OpenAI has taken the help of YouTube to train GPT-4, the next version of ChatGPT. ChatGPT is designed to answer any question. To improve it, OpenAI has taken the help of millions of YouTube videos.
Use of video is illegal
According to the emerging report, OpenAI has transcribed more than 1 million hours of YouTube videos to train the ChatGPT-4 AI model. According to the New York Times, a leading American media organization, OpenAI was aware that transcribing the video in this way was illegal, but considered it fair use. According to the report, OpenAI President Greg Brockman has been personally involved in collecting the YouTube videos.
However, OpenAI says that it has used publicly available data and non-public data to train its generative AI tools. Any large language model is trained in every aspect, for which any publicly available data can be used. However, Google has given its response to this activity of OpenAI.
use of data secretly
Google said it had found a number of flaws in OpenAI’s activity. The tech company said that both our robots.txt files and service policies do not allow unauthorized scraping or downloading of YouTube content. According to the report, OpenAI has trained its AI model on Google’s video streaming platform YouTube by scraping its data. For this he has secretly used data from YouTube.