Skip to Main Content
Chat With Us

CHEM6611: Scientific Communication in Chemistry

:

2. Generative Artificial Intelligence (AI)

Overview of library resources supporting graduate research
Know Your Copyrights
Check the Terms of Service or Terms of Use statements describing copyright.
  • Understand how and where you can use content you create via generative AI
  • Determine who owns the copyright on content you create via generative AI
  • Learn how the company could use the content you created
Protect Your Privacy

While generative AI services claim to protect user privacy, researchers continue to discover new ways to bypass those safeguards. Assume that anything you input to generative AI could leak elsewhere.

  • Check the Terms of Service or Terms of Use statements describing privacy practices
  • Modify your entered text to mask sensitive information, or use human assistance to improve your writing in those sections
  • Ask generative AI how to create or improve a type of graph, instead of using it to visualize your unpublished data
Always Check Policy

Academic and publisher policies on appropriate use of AI will continue to evolve.

  • Document your use of AI just as you document other aspects of your research
  • Check ITS, CTE, and other BC offices for current policy about using AI in research and instruction
  • Check publisher policies on AI before submitting manuscripts

Overview

Weaknesses and Drawbacks

Stochastic Parrot

Digital origami bird created by the prompt 'stochastic parrot'

Generates language
without understanding
Semantic and/or
Sentient Parrot

Perching African grey parrot holds cucumber slice in its beak

Generates and
understands language

 

"Stochastic parrot" image generated by deepai.org.
Parrot African Grey Bird Cucumber by 12019 from Pixabay.

  • Content and quality of generative AI output are limited by the training corpus (date range unknown for many tools) and techniques used to train language models
  • Language models like ChatGPT are notorious for creating "hallucinations"—such as literature citations that sound realistic but don't exist
  • Language models can produce text that includes incorrect information
  • Language models can produce text that plagiarizes existing work from the training corpus or from data entered by other users
  • Proprietary services may not generate reproducible and replicable results

Prompts

Some questions to consider when developing a prompt, or in issuing a series of prompts:

  • What role or persona do you want the tool to use?
  • How will you present the source text in your prompt(s)?
  • When you don't know if training material for an AI tool included paywalled journal articles, what could you present as an example?
  • How will you describe your audience?
  • What specific suggestions can you provide for making the text more concise?
  • If needed, what voice and verb tense will you include in your prompt(s)?
  • How can you assess or verify the output generated by your prompt(s)?

Quote as reported in Improving Your Scientific Writing [PDF] by Frederic D. Bushman, p. 46.)

I want you to act as an AI writing tutor for academic research grants. I will provide you with text and your task is to use artificial intelligence tools, such as natural language processing, to re-write the text to be as clear and concise. In addition, the text should be easy to read and flow naturally. Furthermore, the text should use the active voice. You should also use your rhetorical knowledge and experience about effective writing techniques in order to improve the text without changing the underlying meaning of the text. You should consider at least 5 different versions and show me only the version that you think is the best one.

Learn More About Generative AI