Tabular data, the backbone of countless applications from finance to healthcare, is a treasure trove of insights waiting to be unearthed. However, unlocking this potential has always been hindered by the tedious and costly process of human annotation. Imagine trying to teach a machine learning model about customer behavior using a massive spreadsheet – without labels explaining what each column represents. It's like trying to read a book in a foreign language without a translation. This is the challenge that researchers have long grappled with: how to efficiently and effectively annotate massive tabular datasets so AI models can learn from them. Now, a new research paper introduces AnnotatedTables, a groundbreaking dataset that leverages the power of Large Language Models (LLMs) to automate this crucial process. By using LLMs like ChatGPT to understand the structure and content of tables, researchers have created a massive collection of over 32,000 databases with automatically generated annotations, including over 400,000 executable SQL queries. This innovative approach not only drastically reduces the need for human intervention but also opens up new possibilities for research. The paper demonstrates the versatility of this dataset by exploring two key applications: teaching LLMs a completely new programming language called Rel and evaluating the performance of cutting-edge tabular classification models. The results are promising. In the first study, LLMs learned to translate SQL queries into Rel with surprising accuracy, demonstrating their potential to quickly grasp new coding languages. In the second study, the dataset allowed researchers to rigorously evaluate a novel neural network called TabPFN, which tackles tabular classification with a unique approach. The evaluation, scaled up significantly thanks to AnnotatedTables, showed TabPFN performing on par with state-of-the-art AutoML methods but with a significantly faster processing time. AnnotatedTables signifies a significant leap forward in the world of AI and tabular data. By harnessing the power of LLMs for annotation, it offers a cost-effective and scalable solution to a long-standing bottleneck, paving the way for more sophisticated and insightful analyses of tabular data across diverse fields. While the dataset and the accompanying research show incredible promise, there are still challenges ahead. Further research is needed to refine the LLM annotation process, improve the complexity of generated SQL queries, and explore the full potential of this massive new resource. But one thing is clear: AnnotatedTables has unlocked a new door in the quest to unlock the full potential of tabular data, bringing us closer to a future where AI can truly understand and learn from the wealth of information hidden within the rows and columns of our digital world.