Amazon Digitizes Rare Books to Enhance Its AI Models
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Texts published before 2022 are considered particularly valuable for training language models, as they cannot have been written by artificial intelligence. Training on AI-generated writings carries the risk of "model collapse," meaning a degradation of output quality after ingesting too much artificial content. Rare books, especially those that are no longer published or impossible to find online, thus represent a sought-after new source of data.
A Tracked Book Leads to Amazon Facility VGT3
A tracking device placed in a rare book allowed its journey to be traced to an Amazon facility in Las Vegas, identified as VGT3. This location is distinguished by a logo featuring a dinosaur holding a book. According to 404 Media, Amazon acquires a large number of rare works, removes their bindings, and then scans them to feed the training of its AI models.
Amazon Claims to Purchase Through Commercial Channels
Amazon asserts that it buys books through commercial channels to improve the products and services used by its customers. Companies developing language models require vast amounts of text, and these models are said to have already absorbed everything available on the Internet.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.