ChatGPT goes through several steps to provide an answer to a question. First, it gathers data by accessing web pages and crawling the internet. However, it is at a disadvantage because search engines have been doing this for decades, while ChatGPT is relatively new and has not been able to visit all corners of the internet. One type of information it can gather that other search engines cannot access is chat data, which is collected from the interactions and input of users on the platform. This feeds into their training model, but users should be aware that if they input private data, it may not remain private.
The second step is saving the information, which is similar to indexing in search engines. In the early days of search engines, the data in their index was not updated live, and updates were done every few months due to the high cost in time and money. Large language models like ChatGPT have similar limitations, as updates require retraining the entire model, making it costly and time-consuming. As a result, the information they have may be out of date, and they may not have information on certain topics.
Overall, ChatGPT goes through the steps of gathering data, saving it in a format they can access, and then providing an answer. However, there are limitations to their access to information and the timeliness of their updates, so users should be aware of these factors when using the platform.