Here are some useful things to bear in mind when using a Large Language Model.
1: A LLM is not a search engine
ChatGPT and other LLMs will almost always answer any query you give them, but you cannot guarantee its response is helpful or even accurate. Remember, LLMs understand the structure of language, not its meaning!
2: Do not use LLMs for critical information
You should not use them to answer health and medical queries; to obtain legal advice and other topics where your own or other peoples’ wellbeing might be at stake. At times like this you must be able to guarantee you are receiving the correct information.
3: LLMs make stuff up
If you are using them for your work, use other, authoritative resources to double-check the output from an LLM. Use internet search engines, Wikipedia, and subject-specific sites to confirm (or deny) what the LLM has told you.
4: LLMs learn from you
Many LLMs use the information entered by users to improve their performance. This data is impossible to recall, so be very careful about providing personal or sensitive information to an LLM.
5: You might be breaking the law
You should not use LLMs to upload personal data or data that you do not own to LLMs. These forms of data are protected by various laws and, unless you have specific permission to do so, you cannot legally use them with an LLM. Do not send the contents of books, magazines, newspapers and other publications to an LLM as you do not have permission to do so. Likewise, do not upload questions from your homework or examinations to LLMs, you might not just be breaking the law, you may be breaking the rules for your school, college, or university.
6: LLMs have their own writing style
The text produced by LLMs is slightly different from the way most people write. Whilst the style of writing varies between LLMs, they tend to produce slightly pompous, unfocused material and use words that are less frequent in human writing. Software tools can, to some extent, distinguish LLM text from human written text. These tools are used by many institutions, including schools and universities to flag students who may be passing AI content off as their own work.
7: If you use an LLM for your work – tell people
Be honest with others, tell them what AI you used, how you used it and how much of the finished piece is the AI’s work rather than your own.
8: Have fun!
Are you stuck writing the outline of your novel? Do you want to summarise a complex argument? Are you trying to squeeze a lot of information into a limited number of words? Are you struggling with that first paragraph? Do you want to write a book of AI-generated recipes? LLMs can help you get started with these complex tasks.