Physics 130: Home

Most of the AI you are hearing about - like ChatGPT -  is generative AI.  This means simply that the AI tool is creating something "new" from already existing information that has been uploaded (or fed) into the AI tool.  This can be numerical data, images, Wikipedia entries, journal articles, personal information - the options are endless.  What can Generative AI be useful for?

• Jumpstarting your writing (if you are out of ideas or having problems just writing that first paragraph)
• Helping you brainstorm
• Helping you organize your arguments, findings, thoughts, and main/supporting points. 

See caveats and red flags for other important information.

Let's get the big one out of the way:  It's your responsibility to know and abide by University (including the Academic Integrity Policy and Student Code of Conduct) and professor's policy regarding AI. 

Here's something helpful from UMD Libraries (see link below for more):

As of 2023, a typical AI model isn't assessing whether the information it provides is correct. Its goal when it receives a prompt is to generate what it thinks is the most likely string of words to answer that prompt. Sometimes this results in a correct answer, but sometimes it doesn’t – and the AI cannot interpret or distinguish between the two. It’s up to you to make the distinction. 

AI can be wrong in multiple ways:

• It can give the wrong answer 

• It can omit information by mistake

• It can make up completely fake people, events, and articles

• It can mix truth and fiction

Additionally, remember that AI operates on the information if it's been given.  If that information is incorrect, false, inaccurate, biased, misleading, etc., the generative output from the AI can be all of these things as well.  Keep in mind GIGO  (Garbage In, Garbage Out).

How did some of these play out in what we did today?

More info in the links below!

• Talk to the tool like you would a person.  Pretend it is a conversation.
• Don't make the prompt too short:  Provide context, indicate audience and tone, suggest output style and length.
• Don't make the initial prompt too long:  Generative AI generally does a better job if you refine and add as you go along.  Some types of output are best done in parts (like lab reports)
• Don't be too vague:  Use specifics.  For example, don't just ask it to write a short story about crows. Provide more details and you'll get a better developed output.
• Provide corrections for incorrect output.  If you get a factual error, ask the generative AI tool about the error.  For example, if it only names two countries that start with V (and there are four), ask it about the other two.
• When asking generative AI tool to perform calculations, ask those separately from other content.
• Be aware of hallucinations or incorrect data.  Fact check output.