[tcmb]

Meta says they have an advantage because all the data on their platform, which they can use to train models. Yet, later in the article is mentioned that Meta so far has used the Common Crawl dataset.

If their trove of data is so valuable, why don't they use it? Could the problem be that none of Meta's "hundreds of billions of images" is labeled, making it far less useful for training models?

Also, I imagine that the content on FB and Insta is highly skewed in a certain direction, like selfies and food images, which restricts its usefulness further.

So, Zuckerberg's statement sounds like pure marketing to boost the share price.