A recent report has shed light on the questionable practices of tech giants like OpenAI, Microsoft, Google, and Meta in their pursuit of training their AI models by unlawfully obtaining content at scale.
The report from The New York Times revealed that OpenAI, in collaboration with Microsoft, used a tool called Whisper in 2021 to transcribe audio from YouTube videos to develop text for training its ChatGPT AI. Despite objections from employees regarding the violation of YouTube’s rules, OpenAI went ahead and generated over one million hours of YouTube video transcripts. This led to the creation of ChatGPT-4, considered a leading large language model.
Google was not far behind, as the report also exposed its practices of transcribing YouTube videos to train its own AI, known as Gemini, thereby infringing on creators’ copyrights. In response to the success of ChatGPT-4, Google modified its terms of service in 2023 to allow training of Gemini on various online materials within its ecosystem.
Meta, previously embroiled in the Cambridge Analytica scandal, also resorted to dubious methods to gather data for its AI chatbot. This included hiring contractors in Africa to aggregate summaries of copyrighted fiction and nonfiction works. Despite internal opposition, Meta justified its actions by referring to OpenAI’s activities as a market precedent.
The race for data among these tech giants underscores the importance of acquiring vast amounts of information swiftly to advance AI capabilities before regulatory bodies intervene. The goal is to reach a point where synthetic data creation eliminates the need for further data theft, thereby evading accountability.
The CEO of OpenAI, Sam Altman, has emphasized the significance of surpassing the synthetic data event horizon to ensure continued progress in AI development.
This revelation raises concerns about privacy, copyright infringement, and ethical boundaries in the AI industry. It highlights the need for stringent regulations to govern data practices and protect the rights of content creators in the digital landscape.
