GPT on Document Management

I recently got an invite from Google to try out its new Bard GPT. Mainly it refused to answer various questions about state laws and regulations. But I got a fairly interesting answer to “What are the main characteristics of a legal document management system?” At first glance it seems like a pretty good summary. But when you look a little closer, it gets more complicated.

My comments are interspersed in brackets:

Document storage: The system should be able to store and manage a large volume of documents, both electronic and paper. [A computerized system should be able to manage paper documents????]

Security: The system should be secure, with features such as encryption and access control.

Compliance: The system should be compliant with all relevant regulations, such as HIPAA and GDPR.

Auditing: The system should be able to track all user activity, including document access, changes, and deletions.

Version control: The system should be able to track changes to documents and allow users to revert to previous versions. [It is unclear whether this means automatically, which would be gross overkill, or by designation to create versions.]

Search: The system should be able to search for documents by keyword, date, and other criteria.
[The obvious missing item here is full-text search. “Other criteria” is meaningless.]

Collaboration: The system should allow users to work on documents together in real time.

Electronic signatures: The system should allow users to sign documents electronically.

Mobile access: The system should be accessible from mobile devices.

Cost-effectiveness: The system should be affordable and easy to implement.
[Again, this is meaningless. Cost should be judged by ROI, not price, and while ease of use is an issue, implementation is not.]

These are just some of the key characteristics that a legal document management system should have. When choosing a system, it is important to consider your specific needs and requirements.
[More blah blah blah]

This pretty much illustrates my a priori prejudices concerning the various Chat GPT offerings. They can offer a good point of departure, but there are glaring errors and a lot of generic platitudes that don’t mean very much at all. So while GPT can offer a good start, it can also probably stunt any serious consideration of a topic by encouraging lazy thinking. I expect GPT will perpetuate the “it’s on the internet, it must be true” syndrome. So... be suspicious, be very suspicious.