Large Language Models (LLMs) require large amounts of data for training. Very large. Like the entire textual content of the World Wide Web large. In the case of the largest such models — OpenAI’s GPT, Google’s Gemini, Meta’s LLaMA, France’s Mistral — most of the data used is simply vacuumed up from the internet , if not by the companies themselves then by third-party bot-jockeys like Common Crawl, which provides structured subsets of the data suitable for AI training. Other tranches come from digitized archives like the Books 1, 2 and 3 collections and Z-Library.
In nearly all cases, the hoovering and archive-compiling has been done without the permission or even the knowledge of the creators or rights owners of the vacuumed-up haul.
Companies racing to develop more powerful artificial intelligence are rapidly nearing a new problem: The internet might be too small for their plans. Ever more powerful systems developed by OpenAI, Google and others require larger oceans of information to learn from. That demand is straining the available pool of quality public data online at the same time that some data owners are blocking access to AI companies.
The new horror film Late Night with the Devil hit theaters late last month amid a lot of really good buzz. It has a 96% on Rotten Tomatoes and has broken box office records for its distributor, IFC Films. It seemed poised to become the indie movie success story of the first half of 2024. But that buzz has curdled quite a bit once word started to circulate that generative AI had been used in the film.
A new tool from OpenAI that can generate a convincing clone of anyone’s voice using just 15 seconds of recorded audio has been deemed too risky for general release, as the AI lab seeks to minimise the threat of damaging misinformation in a global year of elections. “We hope to start a dialogue on the responsible deployment of synthetic voices, and how society can adapt to these new capabilities,” OpenAI said in an unsigned blogpost.
Just as in film and TV, only more so, AI represents a gathering storm for video game actors. Some studios are experimenting with tools that can clone voices, alter voices and generate audio from text. In interactive, multi-choice games, this can generate a potentially endless number of characters and conversations – and is far more efficient than asking performers to record huge quantities of dialogue.