1. Sorry but that's still far less bias than ConvNets for vision. And far less bias than previous methods for language. The less bias the better. It's simple.
2. Assuming the rumor is fact, GPT was trained for multiple epochs and books are a small percent of what trains LLMs lol.
2. Assuming the rumor is fact, GPT was trained for multiple epochs and books are a small percent of what trains LLMs lol.