Automating Academic Writing?

The advent of ChatGPT has completely upended academic research writing. Authors love it. Editors hate it. Everyone does it. Yet all deny it.

But have you ever wondered on exact the reearchers spend time on while writing papers? Is it really the concept or hunting for the right verb, adjective or noun to make the writing ‘seamless’.

And yet when using AI the same problems persist. Ask it to write a manuscript and it will:

-Put the name of your variables in the first sentence

-Explain mediation in the second line

-Talk about your sample in the first line of the second paragraph.

As a results, most people I have spoken to find that AI is at best ok-ish at academic writing. It would write the paper that reflects the maturity of a first trimester Ph.D. Student.

However, this is where prompting plays a big part. A lot of academic writing depends upon the ability of the person to prompt the AI well.

Don't trust me?

Try ZeroDrafter - our AI introduction writer (It's free!)

Just give it the names of your study variables and you are done!

Imagine the time savings; you do not need a co-author to write a first draft. You can write 3 manuscripts in the time of one!

-Did you notice that the bot never stated the hypothesis in the first paragraph - a typical problem when you ask AI to write.

-It also managed the transitions well!

It even grounded the hypothesis in theory!

It also cited papers correctly (mostly) and when it was unable to it just installed a placeholder <CITATION>

And all this…

-Without knowing any papers

-Without using any sophesticated models! I am not even using the best model out there (GPT-4) to write it. The bot uses Claude v 1.0 which is even worse than GPT-3.5 by many counts.

Then how is it able to manage the transitions well?

That's the magic of ‘prompting’. Prompting is not ‘just’ English. It is much more. It is your ability to learn about how the AI processes it's outputs - how to prime it's latent space so that it gives you the best and most predictable output.

Want to know more? Click here

Stay tuned for next week!