Meta's AI assistant is being trained using public Facebook and Instagram posts, with the company's large language model, Llama, considered one of the best in the world. This move has added $200 billion to Meta's market cap in a single day. However, this has led to criticism and legal action from book publishers and news organizations who understand the value of this data, while social media users are left in the dark.
Key takeaways:
- Meta, formerly Facebook, is using all user posts, reels, and comments on Instagram and Facebook to train its powerful AI, making social media profiles one of the most valuable datasets on Earth.
- Meta's CEO, Mark Zuckerberg, claims the company's dataset is larger than the Common Crawl dataset, which is over 250 billion web pages collected over 17 years.
- Meta's large language model, Llama, is one of the best AI models in the world, and the company plans to use it to train products like Meta AI, Imagine, and more, infusing these AI products into Facebook, Instagram, and the Metaverse.
- Despite the value of this data to AI, social media users are not being included in the conversation about data ownership, and Meta has essentially claimed ownership of public social media profiles without any grand announcement or notice to users.