The article delves into Meta's internal deliberations around sourcing high-quality writing for training its AI model, Llama 3. Faced with unreasonably expensive and slow licensing options for legal content, employees considered pirating from Library Genesis, a massive online repository of pirated books and research papers. The urgency was underscored by a senior manager's assertion that books were more crucial than web data. Ultimately, after discussions, permission was obtained from leadership to utilize the materials, highlighting a significant ethical dilemma in AI development and content sourcing.
In an internal chat, a research scientist noted that licensing options for acquiring the high-quality writing needed for Llama 3 were "unreasonably expensive," leading to discussions about piracy.
A senior engineer highlighted the need for immediate access to books, stating, "books are actually more important than web data," guiding Meta's unethical decision.
Documents revealed that the Meta team turned to Library Genesis, a vast online repository of pirated materials, after facing challenges with legal licensing routes.
Meta employees recognized that licensing a single book would limit their ability to utilize fair use strategies in training AI, adding to their dilemma.
Collection
[
|
...
]