Get more done with AI

ChatGPT, and more specifically its Advanced Data Analysis capability, allows users to upload a dataset and get it analyzed. However, it has several limitations:

• ChatGPT can only analyze data files like CSVs and Excels; you cannot connect a database or a data warehouse to ChatGPT.

• ChatGPT does not leverage your organizational context and past work to improve the quality of the analysis.

• It’s impractical to review, tweak, reproduce or share the analysis that ChatGPT creates'.

DataLab does not have any of these limitations:

• DataLab can analyze data files, connect to Google Sheets and all common data warehouses and databases.

• DataLab inspects the schema of the connected data source and uses your past activity to enhance the quality of the analysis.

• DataLab features a fully-featured data notebook that contains all the code that the AI Assistant writes to answer your questions. You can seamlessly switch to this code to review, tweak, rerun and share it, without leaving the tool.

DataLab is built on OpenAI’s latest GPT models. Rather than just acting as a simple wrapper that sends your question to OpenAI verbatim, DataLab smartly enhances your question with your unique work context (previous questions, connected data sources, additional organizational context) to improve the quality of the responses. These responses are robustly parsed, which leads to code being written in the background. DataLab sends the output of running that code again to OpenAI to be interpreted.

Large language models, like OpenAI’s GPT model that DataLab is built upon, can make mistakes and hallucinate. The technology is improving at a rapid pace, though, and we fundamentally believe that these problems will be fixed in future iterations of the models. Until then, auditability is key: for important insights that drive important decisions, you want to verify what the AI did to answer your question. With every question you ask in the AI chat, the AI assistant is writing and running code in DataLab’s data notebook, and interpreting the output of that code. You can seamlessly review this code yourself, or share it with a colleague with more technical skills to verify, and tweak it if needed. This is a fundamental advantage of DataLab compared to ChatGPT.

When using DataLab’s AI chat, the quality of the answers to your questions might degrade as the databases you’re asking DataLab to look for answers are getting bigger. DataLab employs smart algorithms to send along the right context along with your original question, but this gets more challenging with larger databases.

• DataCamp nor OpenAI uses customer data to train models: the code or context that you pass to OpenAI will not appear as an answer to someone else’s prompt.

• DataLab sends workbook metadata to OpenAI. Database Metadata like table names, column names and types, and workbook context like previously written code and names and types of variables are sent over to OpenAI to improve the quality of the suggestion.

• DataLab sends the output of code cells to OpenAI. The execution output of code cells is sent over to OpenAI to improve the quality of follow-up suggestions.

• DataLab only sends data to OpenAI when you use AI features. If you decide not to use AI features, or your group admin has disabled the AI Assistant, no metadata or output is sent over to OpenAI.

• DataCamp is ISO 27001 certified, as are these AI features. Your data is protected by solid security practices and policies. For an overview of all our safeguards, visit our security page.