Frequently Asked Questions

Last updated: 28 January 2025
Are my chats in the CityUHK GPT Chatbot secure? / Will my chat history be shared
with others?
- The CityUHK GPT chatbot provides a private and secure environment to support the work and studies of CityUHK users. Your content will not be shared with others within or outside CityU.
- As the CityUHK GPT chatbot is powered by Microsoft Azure OpenAI service, the input (prompts) and output (completions) are stored securely with Microsoft for up to thirty (30) days for monitoring the use that violates the applicable product terms. [Details]
- The chatting history is stored in the CityUHK server for six (6) months to support recall of the conversations by individual users. As this is one of the IT services of the university, the same policies and guidelines apply.
How does the CityUHK GPT Chatbot's quota work?
- The quota is deducted based on the total number of words in the input (prompts) and the outputs (completions). Nevertheless, previous inputs and outputs, up to 10 consecutive iterations, are included in the new prompt to make the chatbot context-aware.
When I asked the CityUHK GPT Chatbot questions about CityU, it answers
incorrectly,
why?
- The CityUHK GPT Chatbot is trained on a lot of text from different sources. It uses patterns in this text to give human-like answers, but it doesn't truly understand the information. This can lead to mistakes or unexpected answers.
- The CityUHK GPT Chatbot uses a general language model (GPT-4o and GPT-4o Mini) that doesn't have specific knowledge about CityU.
Does the CityUHK GPT chatbot have the GPT-4o Mini model? I asked it about GPT-4o
Mini and it said it did
not exist.
- Yes, all our staff and students can access both the GPT-4o and the GPT-4o Mini model in CityUHK GPT chatbot.
- The GPT models for 4o & 4o-Mini are using the default training data up to October 2023, when GPT-4o Mini was not officially released. You will have the answer that GPT-4o Mini model does not exist even when using it.
How can I access the CityUHK GPT Chatbot?
You can access the CityUHK GPT Chatbot via:
- the CityUHK Mobile App on your phone, or;
- the CityUHK Portal on the web.
Who should I contact if I have enquiry/suggestions about CityUHK GPT chatbot?
- We welcome any comments/suggestions to help us improve our services, please contact us at DigitalLearning@cityu.edu.hk.
Do all Gen AI Tool have real time access to the internet?
Not all Gen AI Tools have access to real time access to the internet. Gen AI Tools that have
access to the internet tend to be less secure. You may refer to the table below:
Gen AI Tool | Internet Access | How to Use? | Security |
---|---|---|---|
CityUHK GPT Chatbot | No | CityUHK Mobile App | High |
Gemini | Yes | CityUHK Google Workspace | Moderate |
Perplexity | Yes | Public Access | Low |
Which Gen AI tools allow file upload? What type of files do they accept?
Not all Gen AI Tools allow file upload. You may refer to the table below:
Gen AI Tool | File Upload | Accepted File Types |
---|---|---|
CityUHK GPT Chatbot | No | N/A |
Gemini | Yes | Image files (PNG, JPEG, etc..) |
Perplexity | Yes | TXT, DOCX, PDF, CSV, XLSX and many others |
What is the difference between GPT-4o and GPT-4o mini?
There are some key differences between the two models. You may refer to the table below:
Feature | GPT-4o | GPT-4o Mini |
---|---|---|
Size and Efficiency | Larger and more powerful, but uses more resources. | Smaller and more efficient, using fewer resources. |
Speed | Slower in generating responses. | Faster in generating responses. |
Performance | Better at handling complex tasks and providing detailed answers. | Good for everyday tasks and quick responses, but not as detailed. |
Use Cases | Ideal for high-stakes, complex applications. | Suitable for general use, customer support, and tasks requiring quick responses. |
Quota | Same (Up to 30,000 words each per month) | Same (Up to 30,000 words each per month) |
The Gen AI I am using is not giving me the output I want, what can I do?
- Clarify Your Input: Make sure your prompt is clear and specific. Providing more context can help the Gen AI understand your request better.
- Use Examples: Include examples of the kind of output you expect. This can guide the AI to produce more accurate results.
- Break Down Your Request: If your request is complex, try breaking it down into smaller, more manageable parts.
- Adjust Parameters: If the Gen AI tool allows, adjust parameters like temperature or max tokens to refine the output.
- Review and Edit: Sometimes, the Gen AI's output may need a bit of tweaking. Review the results and make any necessary edits.
- Seek Help: If you're still having trouble, consult the tool's documentation or contact us at DigitalLearning@cityu.edu.hk.