FAQ about ChatGPT (BETA)
What is ChatGPT?
ChatGPT is an advanced language model developed by OpenAI. It’s designed to generate human-like text based on the input it’s given.
Here are some key points about ChatGPT:
- Conversational Agent: It can carry on a conversation with users, making it useful for tasks like drafting emails, writing code, creating written content, tutoring, translating languages, simulating characters for video games, and much more.
- Trained through Machine Learning: ChatGPT is trained on a diverse range of internet text. However, it doesn’t know specifics about which documents were in its training set or have access to any personal data unless explicitly provided during the conversation.
- Improvement over Time: OpenAI continuously works on updates and improvements to make ChatGPT more useful and safe.
- Safe and Secure: Measures are in place to ensure user interactions are private and secure. It also has a system to refuse generation of inappropriate content.
How is ChatGPT implemented in Minerva Messenger?
The current model version implemented in Minerva Messenger is GPT-4-1106-vision-preview. ChatGPT is implemented as a chatbot via an open source Mattermost API plugin. The chatbot emulates the conversation with another Minerva Messenger user. To start conversations, users must first invoke ChatGPT via a direct message using the @ChatGPT name. A quick guide on how to get started with ChatGPT in Minerva Messenger can be found here.
Why is ChatGPT offered as Beta?
Generative AI is a rapidly evolving technology, we are still in the early days in how ChatGPT is being used to assist researchers. After focusing on its technical implementation in Minerva Messenger, we are now in need of more input from users to get a better idea on the stability and functionality of ChatGPT. Because of its experimental nature for both developers and users alike, we are branding it as Beta. Please feel free to reach us with your comments or suggestions at minervamessenger@mpdl.mpg.de.
What privacy and security concerns exist with ChatGPT?
The terms of service of ChatGPT are those stipulated by OpenAI. For Minerva Messenger users, the following points are important to keep in mind:
- Data stored within Minerva Messenger (e.g. chats) cannot be accessed by ChatGPT, OpenAi only receives data which is optionally sent by Minerva Messenger users via the chatbot API.
- Data which is voluntarily sent to ChatGPT using our bot implementation is sent to OpenAI servers, and is retained for 30 days before being deleted.
- The API implementation in Minerva Messenger was done in a way that data transmitted to OpenAI is not linked to the user, in other words, data is anonymized and cannot be tracked to a specific user ID. This ensures another of layer of privacy for Minerva Messenger users.
Nevertheless, we strongly suggest that sensitive data e.g. passwords, financial and medical information, as well as any propriety information of a confidential nature, should NOT be shared with the ChatGPT bot.