Comment by yowzadave

Musk has been fixated on this idea that Twitter is a huge treasure trove of data for AI training, often complaining about AI companies using its data for training purposes. I had just assumed that companies like OpenAI were just crawling the web, which included Twitter, rather than targeting Twitter in particular. Is the Twitter data really that valuable for training AIs? What particular qualities does it have that make it particularly useful compared to any other freely available data set?