OpenAI: In order to train its AI model, known as GPT-4, OpenAI reportedly transcribed almost a million hours of YouTube video. OpenAI “thought it was fair use,” according to the New York Times, even though it was aware that this was illegal. Read on to know more.
OpenAI used YouTube videos to train its AI model
Over a million hours of YouTube videos have apparently been used by OpenAI, the business headed by Sam Altman and the creator of ChatGPT, to train their AI model, GPT-4. According to a recent report, the company has deemed it fair usage, despite noting the possible regulatory difficulties. President of OpenAI Greg Brockman was personally involved in selecting the training video content. The Verge story claims that the corporation would maintain its competitiveness in international research by utilising a variety of sources, including partnerships and public data.
Videos on YouTube cannot be used for “independent” apps, according to Google policy.
Concerns about data usage and intellectual property rights have been brought up legally and ethically by this technique. The unlawful gathering of massive volumes of YouTube data will challenge copyright laws and raise concerns about ownership and consent, even if OpenAI claims it will fall under fair use.
Has Google done the same?
A story published in The New York Times claims that Google trained its AI model Gemini using transcribed words from YouTube videos. The creator who uploads the video to the platform owns the copyright to the videos, which is violated if this is true. According to the article, Google extended the terms of service to enable the use of publicly accessible Google Docs files, restaurant ratings on Google Maps, and additional resources for the purpose of training artificial intelligence models.
Keep watching our YouTube Channel ‘DNP INDIA’. Also, please subscribe and follow us on FACEBOOK, INSTAGRAM, and TWITTER.