@mathias@gruene.social

This is going to be a long 🧵 but I think it will be worth your time.

Background story: One of the hardest problem in copyright law is actually determining whether a work is in the public domain or not. Many countries have exceptions, sometimes really complex ones.

For example, in the French context, it is relevant to know whether the author fell during wartime on the French side. Many great minds at Europeana tried to come up with flow charts to answer the public domain question for many EU/EFTA countries. The result is gorgeous.

You can access a version of the flow charts online. https://pro.europeana.eu/files/Europeana_Professional/IPR/IPR%20images/Public%20Domain%20Calculator%20Poster.pdf

I used to have a DIN A0 poster hanging in my Parliament office in Berlin and it always spiked the interest of visitors coming by. https://t.co/V8m8ZV4JLS

So it is only natural that I gave this task to #chatGPT because, why not.

Prompt: Write an html page with embedded javascript that will ask a user all the required information to determine if a work is in the public domain

The GPT output is awesome, yet there is a very funny aspect to it (which is openly stated in the comment below the HTML/js text): It didn't do the calculation part

// TODO: Implement the algorithm to calculate whether a work is in the public domain https://t.co/gU3gOyXAwr

The resulting HTML is fully functional in the sense that it accepts input but no calculation is done whatsoever. https://t.co/jTfIuz8Vul

So I asked chatGPT to expand on that. It complied. In a second prompt, I asked it to account for different jurisdictions (which I misspelled but chatGPT didn't bother) https://t.co/6Z5DpxPVD8

since the javascript function added another parameter, I asked chatGPT to make modifications to the other parts of the page to account for that change. Note that I didn't give a direct task but just made an observation. https://t.co/fdEkyHuQmE

At this point, the HTML output became too long for one reply and chatGPT had no problem continuing when asked about this. https://t.co/6Mhcjkj88w

a bit of praise, nothing to see here. https://t.co/bOMvSM8wc4

At this point, I asked chatGPT to polish up the interface a bit, because why not: https://t.co/OtVtcnbHfh

and here we are.

I am repeating the disclaimers by chatGPT here: This is not a problem solved, by a long shot. You could never use this tool to make a determination on the PD status of a work. But it was nevertheless a - for lack of better words - magical experience. https://t.co/kzYMKUqgXJ

If the only thing that chatGPT does is to provide people to create semi-nonfunctional user Interfaces to test drive ideas, this whole thing would have been worth it. https://t.co/aCRrSkBSwO

Thu Dec 08 13:22:44 +0000 2022