Generative AI Platforms and Ecology

Posted by on March 19, 2024

By Kathleen Carroll

Conversations about AI and its potential ramifications for ecology have been abound in my lab, department, and on Twitter. Given the number of conversations I have been privy to, I wanted to share my insights with other Early Career Folks. However, I am by no means an expert on using AI in science and would love to have a conversation with others who have more experience! Please reach out to both me (@GuloToTaxidea) and the section (@ESA_EarlyCareer) on Twitter to continue this conversation!

Free generative AI platforms I have explored (there are so so so many more) and what I view as their main pros and cons:

  • ChatGPT
    • Pros:
      • Great for text editing text and code – a non-native English speaking labmate to me they found that ChatGPT made their editing process much less painless for scientific publications
      • It is widely used and recognized – so it links to many other programs (e.g., Glasp for note taking is powered by ChatGPT as is Grammarly)
    • Cons:
      • Makes up fake citations in default settings (why?!?!)
      • If you ask it a question, it may provide some incorrect information
      • limited knowledge of events that occurred after September 2021
    • Consensus
      • Pros:
        • Designed for scientists
        • Consolidates articles
        • Can provide information on scientific consensuses from papers
        • Simple links to full text
      • Cons:
        • Less flexible than elicit so article filtering isn’t as easy
      • Both?
        • Gives a summary of top 5-10 papers which can be great if you only need some literature but is also limiting for many nuanced questions
      • Elicit
        • Pros:
          • Designed for scientists
          • Consolidates articles
          • You can add your own columns based on important search criteria
        • Cons:
          • Only seems to search abstracts of papers for key search terms
        • Both?
          • Elicit seems to find different papers than Google Scholar (which also uses AI) and I have no idea why – I still recommend checking both

I also compiled a list of what seems to be the consensus on how we should (or should not) be using generative AIs in our daily work. Remember though, that this list is not comprehensive or definitive. Before using AI check your university, company, and publisher’s requirements for using and reporting!

Do:

  • use generative AIs to help you find papers/citations (not ChatGPT –see cons above)
  • use generative AIs to find cool coding solutions to problems you haven’t been able to solve
  • clearly acknowledge any contribution made by AI software in Methods and/or Acknowledgements – as required by most publishers and jounrals
  • use generative AIs for grammatical editing
  • use generative AIs to brainstorm new ideas by asking questions
  • upload confusing papers and ask generative AIs questions about them (I find it helpful to ask things like, “what are the five main points of Article XYZ?”)
  • ask generative AIs for mind-maps of topics you want to learn more about
  • ask generative AIs for an outline of steps to master a new skill or topic/or for new class preps – see image
  • one of the funnier uses I have seen is using generative AI to get help writing professional emails – you can ask an AI to write you a more professional version of an email you have composed (especially if you have a tendency to write snarky emails like one of my close friends who then sends them to me for editing – guess I’m out of a job)

Don’t:

  • use generative AIs to write original content (e.g., new sentences, whole paragraphs, papers, or lit reviews) and publish it as your own
  • assume that the information provided is automatically correct – I asked ChatGPT about a paper after only providing the DOI and it summarized a completely different paper

 

Share this article