The author is curious about companies like Asteria, which claim to train their AI models exclusively on licensed content. They wonder whether it’s feasible for Asteria to acquire enough licensed material to create a competitive model against platforms like Veo, or if they might only be fine-tuning an existing model with their dataset. This raises questions about the practicality and scale of licensing content for training AI, as well as the methods companies use to develop their models in the competitive landscape. The discussion implies a need for clarity on the actual practices of content licensing versus model training.
Source link
Share
Read more