OpenAI encountered a supply issue in late 2021. While creating its most recent AI system, the artificial intelligence lab went through every credible English-language text source available online. To train the next iteration of its technology, it required a huge amount of data. Thus, Whisper is a speech recognition tool developed by OpenAI researchers. It might create fresh conversational text by transcribing the audio from YouTube videos, which would improve the intelligence of an AI system. A few OpenAI staff members talked about how this would violate YouTube’s policies.
According to the sources, an OpenAI team translated over a million hours of YouTube content. President of OpenAI Greg Brockman was one of the team members who directly assisted in gathering the films.
After that, the words were put into GPT-4, a system that served as the foundation for the most recent iteration of the ChatGPT chatbot and was regarded as one of the most potent AI models in existence. The competition to develop AI has turned into a frantic search for the digital information required to progress in the field. Tech companies including OpenAI, Google, and Meta have discussed circumventing the law, taken shortcuts, and disregarded corporate standards to access the data. Google, like OpenAI, collected text for their AI models by transcribing
YouTube videos.
High-quality material, like published books and articles that have been meticulously researched and edited by professionals, is the most valued type of data, according to AI researchers. For a long time, the internet offered a limitless supply of information thanks to websites like Reddit and Wikipedia. However, tech companies wanted additional repositories as AI developed. Due to privacy restrictions and their policies, Google and Meta, two companies with billions of users who create social media posts and search queries daily, were prohibited from using a large portion of that content for artificial intelligence.
By 2026, IT corporations might be able to get high-quality data online.