2 ms·
This makes sense, OpenAI has been experimenting with human-in-the-loop type refinement systems for GPT-3 (see InstructGPT https://openai.com/blog/instruction-fo
by amilios 5y ago
This makes sense, OpenAI has been experimenting with human-in-the-loop type refinement systems for GPT-3 (see InstructGPT https://openai.com/blog/instruction-following/ https://openai.com/blog/instruction-following/). It would make sense that they would use something similar for Dall-E 2 as well.