Tech Giants' Copyright Infringements for AI Training Draw Criticism as Predictable Data Hunger
-
Longtime AI critic Gary Marcus calls AI a "Shakespearean tragedy" as copyright infringements by big tech firms come to light. He warned for years about data-hungry AI models.
-
OpenAI and Google both scraped over 1 million YouTube videos without consent to train AI models, likely violating copyright law.
-
Caught scraping itself, Google doesn't publicly denounce OpenAI's scraping to avoid drawing attention to its own use of YouTube videos.
-
Meta also used copyrighted books without permission to train AI models, deciding potential future lawsuits were worth the risk.
-
Marcus argues the data hunger of AI models was predictable and evidence they are built on shaky ground, needing more data than actually exists.