ChatGPT Writes Convincing Medical Study for a Fictional Wonder Drug

ChatGPT is a culmination of Reinforcement Learning optimization strategies, specifically Proximal Policy Optimization (PPO). OpenAI leveraged AI trainers to rank the model and shape rewards based on model ranking. Make no mistake: reinforcement learning requires constant iterations, trial and error, rewarding unintended behaviors, etc. The computational barrier to entry is costly in compute costs and time to train. However, it is one of the most effective conversational AI's to date.