I recently attended a focus group discussion on the use of artificial intelligence in scientific writing and publishing, which led me to write this blogpost!

For this blogpost, I asked ChatGPT (my personal favorite AI tool) to define what AI is. In short, AI refers to computer systems or machines that perform tasks that would “normally require human intelligence.” Some things that I have used AI for in my personal life include what to cook for dinner that evening, workout routines, trip itineraries, or what astrology says about my compatibility with my partner’s star sign (…I wish I was kidding about that last one). For my work – I’ll be honest – I probably do use it more often than some of my other colleagues…but I don’t think that’s a bad thing?

I first started using AI more for my work about a year ago when I was preparing for my qualifying exam. As I would ask myself what the pros and cons of my experimental model are, I would sometimes ask ChatGPT the same questions to see what other ideas it could come up with! Luckily I was warned very early on to NOT feed ChatGPT any unpublished or proprietary information, so I would never state the specific chemical or gene I was interested in, I’d simply say “gene X”, for example. Using ChatGPT as a study tool was very useful for me, as I had a million and one questions to ask. After today’s focus group discussion, what I think helped make ChatGPT an even more useful tool for me was that I already had a deep foundational knowledge of the things I was asking about, so I could identify when the output didn’t make logical sense.

Something incredibly important to point out is that AI does not always give you accurate information!!! I have asked ChatGPT to calculate the number of days between two dates so that I could note the exposure time for my experiment, but the number it spit out didn’t make sense to me, so I asked, “Are you sure that’s accurate?” To which it responded, “Great catch! [blah blah blah – insert correct answer].” And note, asking how many days are between two dates is a fairly low-level, simple calculation…so people should definitely be critical of the responses they get when asking “higher level” questions to the chatbot. One of the big pieces of advice we received in this workshop was to use AI as a guide, rather than to use it as a rule.

This advice becomes a lot more critical when considering writing and publishing. I think one of the biggest critiques of AI in all different workspaces is how it can take away from human creativity and personal style. Something that makes humanity so beautiful is our uniqueness and diversity – we don’t want to lose that. A fear of using AI for creative things like writing is that there will be a “homogenization” of writing styles, and everything will become overall average…and boring.

Now, I am writing this as someone who has actually used AI to help me with writing – both with personal things (like this blog) and with science, so I am going to share HOW I use AI, and where I (personally) think it is beneficial and where I think it is not. I have often been told that I am a good writer, and I don’t think that my use of AI makes this any less true! This – like pretty much everything else on my personal blog – is my own opinion, and not necessarily a reflection of the views of my workplace, colleagues, etc.

When talking about writing, I mainly use ChatGPT to help me edit and refine text that I have already finished writing myself. Thus, all of the original content was generated by me. First, I will read through my own text and identify spots that I think read weird, don’t transition well, or don’t fully communicate whatever it is I am trying to communicate. I will then ask ChatGPT to help me edit those specific sentences. Whenever ChatGPT gives me multiple options (which it usually does), I will then reply with which versions I liked better and why so that we can come up with a newer version that contains all the components that I liked. I do not ever just simply accept all the edits that ChatGPT gives me – in fact, I often don’t really like the initial edits that ChatGPT provides me – instead I almost have a “conversation” with ChatGPT to help me get close to what I want. Once I think ChatGPT is close enough, I will still edit the sentence (or sentences) so that it contains my voice and my writing style. I consider this method as “using AI as a supplementary tool” rather than “using it as a necessary crutch.”

As an example of a somewhat embarrassing, though admittedly true, way that I have attempted to use ChatGPT where it was entirely unhelpful and actually gave me more work was when I asked ChatGPT to help me make a journal club presentation for a paper I had to present for class. I am in my fourth year of my PhD, I’m 27 years old…I’m really tired of taking classes and having to do work for them 😅 I’m not ashamed to admit that. I gave ChatGPT the journal article link and it quickly spit out suggestions for each slide. I have done journal club presentations many times before, and so I knew that what ChatGPT gave me was, well…trash. I spent about 10-15 minutes trying to make sense of it before I decided this method was turning out to take more time and be more work than it was to simply read the paper and make the slides myself. Additionally, the work that I created myself was much better AND it maintained my personality and style of presenting – I actually like presenting and using my own creativity to make slides…so I don’t really know why I tried to give this task to AI in the first place (I was just being lazy).

All this to say, I think that ChatGPT, and its similar counterparts, are a great tool for editing, refining, and thinking of clearer ways of communicating things. However, I think “natural human intelligence” (a term I am stealing from my amazing mentor for my Writing Internship) is still the greatest tool we have!! Not only this, but natural human intelligence is also an essential tool. AI can only be as good as the prompts we – as humans – feed it. Another quote I really appreciated from one of the speakers at this workshop was that, “Prompt writing is still writing” and can help show how much someone knows about stuff. I really liked this quote because, as I was sitting through this workshop, I began to feel a little bit insecure about my use of AI for writing and felt like I might be doing something wrong…but this statement helped validate me that how I use AI is at a “higher level” and reminded me that the quality of the outputs that AI gives me is a reflection of the quality of input that *I* feed it.

Did I write this entire blogpost to make me feel better about using AI for writing? Maybe. But, you know what, maybe…just maybe…I helped show you that it is okay to use AI for writing, it’s just a matter of how you use it to help with writing. At the end of the day, the outputs that AI provides you will only be as good as the inputs you feed it…so it works at its best when you have a strong foundational knowledge of what you are asking it about.

I focused this blogpost on the use of AI in writing and publishing since that is what the workshop was focused on, but obviously there are many, many other uses for it beyond that! Additionally, I think that the use of AI in writing, publishing, or any creative art is a space where people are really concerned about its use because of the potential to lose human creativity. So please remember, your own personal knowledge, style, and personality, however, are still the best things out there…even a tool as powerful as AI couldn’t make someone as unique or cool as you!!!