> Personally, I find the fact that it can't say "I don't know" to be much more offputting This is a problem with their PPO "finishing school" not with their ba

> Personally, I find the fact that it can't say "I don't know" to be much more offputting This is a problem with their PPO "finishing school" not with their base model. Check Figure 8 in https://arxiv.org/pdf/2303.08774.pdf https://arxiv.org/pdf/2303.08774.pdf its base model has an uncannily well calibrated confidence in the correctness of its decisions. Basically, their PPO is for training it to talk to normies, and normies don't like uncertainty. Reinforcement learning using Proximal Policy Optimization (PPO): They optimize a policy against the reward model using reinforcement learning. For each new prompt sampled from the dataset, the policy generates an output. The reward model calculates a reward for the output, and the reward is used to update the policy using the PPO algorithm.