ChatGPT is a form of generative AI, an AI chatbot that uses natural language processing to create human-like conversations, trained with human feedback and a reward model for reinforcement learning.
Risk of introducing AI
AI systems typically require large amounts of data to simulate human behavior, and this can be collected in various ways. For example, a company operating on multiple online platforms can get data from many sources. Some are provided directly by the user, such as contact information and purchase history. Others are collected “behind the scenes,” such as through the use of cookies and other tracking technologies.
Looking specifically at ChatGPT, it’s not yet clear how ChatGPT’s underlying data is collected, but data protection experts say it’s illegal to obtain training data simply by searching the Internet. I warn you there is.
A common concern about this process is that AI may “learn” from people’s prompts and provide information that may or may not contain personal data to others who inquire about the issue.
Currently, Language Model AI or ChatGPT does not automatically add information from queries to its model for others to query. In other words, information entered into a query box does not make that data into ChatGPT or language available to others. However, the query is visible to the organization that provided the language model.
These queries are saved and used to develop the next version of the language model. This may mean that a developer or provider or their partners or contractors can read the query and somehow incorporate it into future versions. Therefore, users should fully understand the terms of use and privacy policy before asking sensitive questions.
Questions may be sensitive depending on the data included in the query or who asked the question and when. For example, a manager is found asking, “What is the best way to fire an employee?” Or when someone asks revealing questions about health or relationships. It is important to note that information from multiple queries with the same login name will be aggregated.
Cybersecurity becomes another risk as more organizations create language models. This means that queries stored online can be hacked, corrupted, or even accidentally made publicly accessible. This may include information that can identify you.
