HKSYU ChatGPT
HKSYU ChatGPT
Pilot Run Version 1.0
Last Updated: 13/Aug/2023
Last Updated: 13/Aug/2023
About HKSYU ChatGPT
- HKSYU ChatGPT provides a private and secure chatbot platform to members of the University, including faculty members, administration staff, and students.
- To access the platform, use the link https://chatgpt.hksyu.edu/ and log in with your HKSYU account.
- Future improvements and advancements will be based on actual usage and feedback received. Unlike publicly available ChatGPT apps, the HKSYU ChatGPT WILL NOT use users’ prompts for the language model training. Users can also clear their chat history at their discretion at any time.
Usage Token
At rollout, the monthly usage quota is 500,000 tokens for Staff members and 300,000 tokens for Students. According to OpenAI, general token usage approximations are as follows:
https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them
These usage quotas will be reviewed based on actual usage and feedback during the pilot run.
- 1 token ~= 4 chars in English ~= ¾ words
- 100 tokens ~= 75 words ~= 1 paragraph
- 2048 tokens ~= 1,500 words
https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them
These usage quotas will be reviewed based on actual usage and feedback during the pilot run.
ChatGPT Limitations
Here are some known limitations of ChatGPT and usage advises in academic settings:
- ChatGPT lacks professional, specific and contextual information, especially when there are ambiguous words in the users’ prompt and language other than English. It could provide a general overview and outlines for certain topics, but evaluating the usefulness of information depends on the user
- ChatGPT may fabricate quotations of news, academic articles, and law cases. Also, the user MUST validate and confirm the truthfulness of content generated by ChatGPT.
- ChatGPT is an implementation of AI which lacks common sense, emotions, humour, time, and contextual understanding. Also, for security reasons, ChatGPT will not respond to some sensitive words.
- ChatGPT may produce grammatical and contextual mistakes. The user is responsible for proofreading and correcting the generated content instead of using it as-is.
Third-party ChatGPT and Cautions
Alternatively, members of the University could register and access third-party-provided ChatGPT services, like OpenAI ChatGPT and Quora POE. However, given the Large Language Model (LLM) algorithm will analyse and train using users’ input and uploaded information, members of the University should exercise caution when using these third-party apps, especially for information involving privacy, internal documents, unpublished manuscripts and proposals, etc.
↑