The false promise of AI for PhD research and writing

Should you be using AI for your PhD research and writing? There's a lot of hype, but you've still got to do the work.

0:00 The false promise of AI for PhD research
0:39 Where AI can be used effectively and ethically
1:32 Never use AI to generate text for you
2:10 Anything you submit, you have to be able to defend
2:40 Be very skeptical about what others are saying
3:25 The problem with using AI to summarize academic literature
4:43 You still have to do the work

[James_Hayton]

CORRECTION: In the video i said it was OK to use AI for transcription. Having given it some more thought I'm not sure about this. I would not trust ANY online AI platform with sensitive data, and you may be in breach of ethics regulations if you upload audio files of participant interviews.

If you can auto-transcribe without the data leaving your computer you might be OK. Double check with your ethics committee on the rules, and if in doubt, don't risk it.

[TheAldousc]

I have 3 rules for AI use in academia. 1 always put research integrity above all else 2 never used AI as a quotable source or outsource your thinking to AI 3 Use AI where appropriate to improve productivity.

[kamilazdybal]

AI aside, same thing goes with all those videos "How to write a paper in a weekend" -- they're selling a lie. You can't write a good-quality research paper in a day or two, simply because corrections to your prose, developing clarity of your figures, and getting powerful insights from your results are a gradual process. It has to be separated by rest so that each time you come back to writing/editing, you can look at your paper with a fresh mind. The process of writing necessarily has to be punctuated with periods of time away from your manuscript, and hence takes weeks or months, sometimes even years. There's no shortcut to developing good-quality research.

Keep up the great work, James! I hope that thanks to channels like yours, academics will keep their integrity and will spend time deeply thinking about their research instead of searching for band-aids that sell a lie.

[Dr_Talobaidi]

I've heard of two students who got disqualified before examination because supervisors picked that the thesis was written by AI. So great video to emphasise this point

[httv]

Yeah, for me, deep reading is more valuable than a summary from AI. Sometimes AI gives false perceptions to my mind, but it can give me another new thing to read deeply and make my own statement.

[salganik]

In the early rise of LLMs, I found out that in niche fields they just copied the results of Wikipedia pages on that topic. This motivated me to update some of those pages in my field (sea ice), but also showed me how bad or misleading some answers were (and still are).

P.S. Thanks for supporting ethical research.

[MrZERK]

Thank you Dr. Hayton for shedding some light into the topic. People are easily fooled by AI since it sells the idea that it will do the work for you. Unfortunately nowadays, in a world where everyone is too busy that sounds too appealing to ignore.

[missytyrrell]

I don't want anything /anyone to do the work for me. I enjoy it. 😏

[aneedfortheory]

Hello James. I am very grateful that you published this video. Your arguments are well presented, and I completely agree with you. Specialization requires deep reading on the part of the researcher, whether they like it or not.

I would like to ask, though, what techniques do you have for efficiently reading a paper? I used to think reading 2-3 pages per day of, say, a 20-25 article will suffice, but very quickly you either find yourself losing focus or being unable to connect the sections you read given that you are reading the article over a two-week period. Is there any other method you can suggest that enables a researcher to glean the information efficiently from the text?

[simonclark]

It was disappointing to hear so many questions about AI after your recent livestream. This seems to me to be a distraction from the real priority of writing a dissertation: being clear about what you want to say!

[Chandni_rudra]

Thank you so much for that insightful video. I am editing a few chapters of mine as I have to submit my thesis around the month of October. I have paucity of time. Can I use paraphrasing tools for doing a little stuff of mine? Will it count into plagiarism? It will be helpful if you respond.

[GemmaS]

Thanks for this. How to make sure a proofread text using AI is not considered as containing AI generated text?

[csbc]

Shit hits the fan when you offload your thinking to AI.

[James-oyqc]

My own view is that AI has the potential to serve as a very useful tool, particularly in identifying some of the articles that should be read as part of your research. If used properly, it has the potential to be a very useful servant but a bad master.

I agree there is nothing like reading a paper in its entirety to understand the subtleties and nuances of an argument. As late as last evening, reading through a sixty-five-page peer-reviewed paper [written in 2010 - located by Google Scholar] I found a footnote that would not have been identified by AI that has been invaluable. AI is not a substitute for diligent hard work, which is often tedious. I share your fear that it is going to entrench academic mediocrity; however, for those who use AI “intelligently, ” it does proffer real opportunities.

May I commend your integrity for not becoming one of the paid sycophants on YouTube that proffer an AI-based vision of academic utopia that is utterly misleading.

[gabrielasofia]

Chat gpt cannot be used in Hong Kong :(