OpenAI Finishes Training Next-Gen Model, Code-Named Bel

According to a tipster, OpenAI has finished pre-training its next-generation model, code-named Bel, with over 10T tokens of training data. Bel is the direct successor to Doug and is expected to be the base model after GPT-6, possibly pushing toward AGI. AGI is a hypothetical system that can learn, understand, and perform any intellectual task like a human, with cross-domain and self-learning abilities.