How do ChatGPT and other “Generative AI” work? Andrew Long attempts to cut through the hype and explain what is happening in the background.
ChatGPT is built on the Generative Pre-trained Transformer (GPT) architecture, a key innovation in natural language processing (NLP) in recent years, and that can generate new content instead of analyzing existing data.
OpenAI, the company behind ChatGPT, trained the model using Reinforcement Learning from Human Feedback (RLHF) and fine-tuning from a model in the GPT-3.5 series. Available publicly now as a vast prototype experiment, the best path forward is to try it for yourself and retrain the way you think about pretty much everything...
I consider applications to #softwaredevelopment, competing AI from Google, the ethical debate around the authorship of technical #geoscience publications, and related Generative AI to create images.