Using AI for PhD research and writing: Some cautionary thoughts

For more PhD tips: phd.academy
Also check out my online writing course: phd.academy/the-writing-course
0:00 Introduction- the positives of AI for research
0:37 The risks
2:11 If you put your name on it you're responsible for it
2:53 If you're going to use it, you still need well-developed research skills to check what it has done
3:47 My worry and my hope for the future
4:46 Make sure you can defend it if you're going to put your name on it

[@reflective_shell]

I like your passage about "AI-as-interns". Actually, when discussing some arguments in the text with ChatGPT, I felt like I was talking with a ~2nd-3rd year student who is very good at gathering information by request but quite bad at generating authentically new ideas. When I asked to criticize a certain methodology from the perspectives of Theory1 and Theory2 it did just great (at least, convincing and coherent) but the other time the requested summary of another theory seemed super generic - as if a person was only half-prepared for the exam and tried its best to fill the gaps within an answer.
Anyway, this is how LLMs operate and we should be aware of their advantages and flaws in order to meaningfully integrate them into a writing process

[@James_Hayton]

The point about AI being like an intern isn't mine- it came from Kevin Kelly (I think on the Tim Ferriss podcast)

[@helmutwagabi6688]

I discovered that AI creates non-existent references whenever asked to give a reference for the material churned out. It cannot be relied upon in that sense.

[@affiliatesssw6940]

We need to use it ethically and still do the work ourselve!!

[@James_Hayton]

You can get it to do SOME of the work, for example, if I were doing interviews then I'd use AI to transcribe them, but then go through and listen and correct it.

[@emmanuelandem9400]

You are spot on this. Exactly my thought . I fear that someday it will make researchers lazy. Just my thought

[@James_Hayton]

I'm not sure it's laziness (though sometimes that might be the case)... I think often it's more the pressure to produce results and publications that drives people to cheat.

[@Sorobai]

I started using it in my research it was helping me just fine rewriting my text until I realized the damn thing got even more enthusiastic about the work than I was and wanted to do all the work alone. I felt I was a espectator of my own project so I had to shut it down 😅

[@thailevan3720]

thank you

[@SociotechnicalSystems]

Human-in-the-loop.