Connection to DriversCloudCreate a DriversCloud.com accountReset your DriversCloud.com passwordAccount migration
NVIDIA reportedly trains its AI with YouTube videos... without worrying about terms of use
Artificial intelligence is progressing, but not always in the most ethical of ways... or even illegal?
If there's one brand involved in artificial intelligence, it's NVIDIA. The company founded by Jensn Huang in 1993 first made its mark with graphics cards, before its GPUs (or graphics processing units, the chips at the heart of graphics cards) were put to good use by companies specializing in AI: as it happens, the computing power and data processing capabilities of GPUs are perfect for AI. That's all it took for NVIDIA to become a billion-dollar giant in this new trend.
To consolidate its dominance, NVIDIA has no intention of resting on its laurels, and whether it's a question of new GPUs or perfecting the performance of its AI models, the American company is firing on all cylinders... sometimes to the limit of what the law allows? That's the question posed by 404 Media, which has published a report claiming that NVIDIA has a tendency to make cavalier use of the catalogs of entertainment giants such as Netflix and YouTube to "feed" its artificial intelligence. Indeed, in order to respond as effectively as possible to user requests, and to produce results that are as successful as possible, AI models need "inspirations". NVIDIA could think of nothing better than to draw this inspiration from the content broadcast by these two world-famous platforms, but without really asking them.
YouTube is said to have reacted to 404 Media, explaining that NVIDIA's actions " constitute a flagrant violation of its terms of use ". This is not the first time that YouTube CEO Neal Moha has made such a comment: last April, he made the same remark about Sora, the generative AI created by OpenAI. NVIDIA is said to have used a YouTube video library (over 130 million videos) with a rather special license, since it is dedicated to academic research. Problem is, in NVIDIA's case, it's hard to deny that the AI model "trained" in this way will be commercially exploited. While NVIDIA denies any misuse, there's a good chance that this case will be settled in court, unless an amicable agreement - a way of saying in exchange for a large sum of money - is reached between the companies concerned.