Companies are seeking data through partnerships and acquisitions, with Meta reportedly considering the purchase of publisher Simon & Schuster for more training data. Companies also use "feedback loops" from past interactions and outputs to improve their models. Meta CEO Mark Zuckerberg stated that these feedback loops will be more valuable than any upfront corpus.
Key takeaways:
- Meta's chief product officer, Chris Cox, stated that the company uses publicly available photos and text to train its text-to-image generator model, Emu.
- Users can create images on Meta AI by typing a prompt starting with the word 'imagine', and the AI will generate four images.
- Companies, including Meta, are seeking to acquire more data to train their AI models, with strategies including partnerships with other firms and potential acquisitions.
- Meta uses 'feedback loops', data collected from past interactions and outputs, to train their models and improve future performance.