Lawyer caught submitting ChatGPT-fabricated citations to SDNY
twitter.com
twitter.com
This is the filing that appears to be ChatGPT generated: https://storage.courtlistener.com/recap/gov.uscourts.nysd.57...
There is some back-and-forth over the defendants not being able to find the cited cases in PACER, followed by the April 25 filing which contains entirely ChatGPT-fabricated submissions for each case, such as this one: https://storage.courtlistener.com/recap/gov.uscourts.nysd.57...
The lawyer admits his use of ChatGPT in a more recent filing: https://storage.courtlistener.com/recap/gov.uscourts.nysd.57...
(1) Proof read everything that comes out of ChatGPT.
(2) Verify all sources that come out of ChatGPT.
If you are not lazy and want to become a 10x employee, follow these steps to use ChatGPT:
1. Write a rough draft.
2. Ask ChatGPT to give you bullet points on how to improve it.
3. Improve your rough draft. Only implement the improvements you agree with.
4. Repeat steps 2 and 3 as needed.
5. Ask ChatGPT to write it's own version from scratch.
6. Merge ChatGPT's version with your version.
This works out your mind, serves as practice, decreases the number of parties you unwittingly leak your thought process to, may evem make you more creative, and avoids hella awkward followups like the lawyer in this post had to deal with.
How is this not mostly "doing it yourself"? It certainly "works out your mind" and "serves as practice" (maybe even better practice than doing it unaided)
By my count it hits 3 out of 5 of your points. I disagree with the 5th and the 3rd is possibly not relevant or not a plausible risk.