Data Conversion

I came across this use case of ChatGPT while working on this project. I aimed to create a citation for an article in CSL-JSON format. I had a long list of authors, but the formatting needed to be fixed when I copied and pasted them. So I asked ChatGPT the following to help me with citations.

Example

User

Extract and convert this text containing authors…

Enkelejda KasneciKathrin SesslerStefan KüchemannMaria BannertDaryna DementievaFrank FischerUrs GasserGeorg GrohStephan GünnemannEyke HüllermeierStepha KruscheGitta KutyniokTilman MichaeliClaudia NerdelJürgen PfefferOleksandra PoquetMichael SailerAlbrecht SchmidtTina SeidelMatthias StadlerJochen WellerJochen KuhnGjergji KasneciFloridi&Chiriatti, 2020Devlin et al., 2018Tay et al., 2022Vaswanietal.,2017Vaswani et al. (2017)

Into the following JSON format:

{
  "given": "<author-1-first-name>",
  "family": "<author-1-last-name>"
},
{
  "given": "<author-2-first-name>",
  "family": "<author-2-last-name>"
},
...

GPT

Here’s the JSON format for the given text containing authors: [ { "given": "Enkelejda", "family": "Kasneci" }, { "given": "Kathrin", "family": "Sessler" }, { "given": "Stefan", "family": "Küchemann" }, { "given": "Maria", "family": "Bannert" }, { "given": "Daryna", "family": "Dementieva" }, { "given": "Frank", "family": "Fischer" }, { "given": "Urs", "family": "Gasser" }, { "given": "Georg", "family": "Groh" }, { "given": "Stephan", "family": "Günnemann" }, { "given": "Eyke", "family": "Hüllermeier" }, { "given": "Stepha", "family": "Krusche" }, { "given": "Gitta", "family": "Kutyniok" }, { "given": "Tilman", "family": "Michaeli" }, { "given": "Claudia", "family": "Nerdel" }, { "given": "Jürgen", "family": "Pfeffer" }, { "given": "Oleksandra", "family": "Poquet" }, { "given": "Michael", "family": "Sailer" }, { "given": "Albrecht", "family": "Schmidt" }, { "given": "Tina", "family": "Seidel" }, { "given": "Matthias", "family": "Stadler" }, { "given": "Jochen", "family": "Weller" }, { "given": "Jochen", "family": "Kuhn" }, { "given": "Gjergji", "family": "Kasneci" }, { "given": "Floridi", "family": "Chiriatti" }, { "given": "Devlin", "family": "et al." }, { "given": "Tay", "family": "et al." }, { "given": "Vaswani", "family": "et al." }]

Analysis

I could then use the response from ChatGPT in my project without having to do the manual work of parsing each name individually.

This example raises an important question. What if ChatGPT is wrong? I want to counter that question with “What if I am wrong?” or “What if my computer glitches and types the wrong letter?” or “What if < blank > ?”

Hidden in my first question about ChatGPT’s correctness is an implicit assumption that ChatGPT is the only point of failure. However, a better way of thinking about this is “Who is responsible?” If I trust ChatGPT to give me the correct answer and it produces something bogus that I do not verify, I am responsible, not the chatbot.

This is why education around the limitations of GPT is essential so that people can better take responsibility for the output.