Finding the right tool for the job

Understanding how LLMs work is just the first step in AI literacy. In order to be a competent user of this technology it is important to understand the affordances and limitations of each tool and how it stacks up against some of the others both in the same category and across several categories. For example, ChatGPT, Copilot, and other LLMs are known to produce incorrect or bogus citations. That means that they may not be the most reliable research assistants when citations are important. On the other hand, a platform like Elicit was specifically built to be a research companion and will crawl all available journal databases.

Trial & error

Another crucial consideration for effective and competent AI use is figuring out your own preferences. It is important to work with different AI tools and compare results to see for yourself which one produces more useful content in fewer iterations. Working with several different tools on the same tasks will also allow you to figure out the best prompt format for each.

Using LLMs

One important thing to keep in mind when working with any generative AI is to verify the information the tool generates. Inaccuracies are possible regardless of how good your prompts are.

What is ChatGPT and other general use LLMs good for?

• Brainstorming ideas
• Explaining information in ways that are easy to understand
• Summarizing and outlining
• Exploring research questions and topics (be sure to fact check the results).
• Translating text to different languages (be aware that results may not be questionable in some languages)
• Proofreading & editing
• Planning & projecting
• Visualizing information
• Writing or debugging simple computer code (some machine languages may be handled better than others)

What is it not so good for?

• Library research (not yet). For now, it's best to use Library search, Library databases, or Google Scholar. See I can’t find the citations that ChatGPT gave me. What should I do? for further guidance.
  
  Note: You may want to try one of these tools that summarize web search results with generative AI. 
  • Perplexity AI   
  • Microsoft Copilot 
  • Google Chrome with Gemini (coming soon)
    
    You may also want to try one of these tools that combine generative AI with academic searching, in order to summarize and find more sources with semantic searching.
     
• Asking for any information that would have dire consequences if it was incorrect (such as health, financial, or legal advice, and so on). This is because of its tendency to sometimes make up answers, but still sound very confident.

The results you get are as good as your prompt

The "prompt" is what you type into the search or task box of an AI tool. It is the instruction that you give to the program which will produce or generate the result/output.

You can use a single prompt or go through a series of prompts to complete a task.

The more relevant information and specific instructions you can provide typically the better the results.

Tips for writing effective prompts

1. Try to specify context as much as possible.

2. Give the AI tool detailed instructions, including how you would like the results formatted.

3. Use an iterative approach by asking for additional revisions until you get the result you want.