💭 Some interesting thoughts on possibilities with ChatGPT and GPT-4. Due to its ability to build a model based on a huge corpus of text [aka document data types], many of us leverage it to find answers to questions fast. This is somewhat competing with Google Search Engine, specialty when ChatGPT ELEGANTLY served 5 million users on its first day of being public. 🤯 Here is a thread from multiple CEOs of companies that aim to compete with google. They shared their bold, optimistic predictions for the future. 📊 Some of them refer to managing multiple ML experiments with 1 trillion parameters [ we already had that in 2021], and it will be possible because of fine-tuning of smaller models - aka - more Experiments 🔓 The hidden secret of #LLMs? How much training data you have matters as much as model size. Aka scalable data. big data. and so on. I wonder what do you think about ChatGPT and how it might impact how we build data systems today.