I have a great deal of reference materials related to the history of surveying (1550 to 1920) on my hard drive. The development of search engines like ChatGPT that utilize Artificial Intelligence to conduct searches makes it fairly easily for me make my reference materials available to collectors of surveying instruments and books.
The ChatGPT search engine that is readily available to the public searches a vast amount of data worldwide, and often comes up with very confusing answers because of all the conflicting and fake information that exists on the worldwide web.
I purchased a license that allows me to limit the AI search to just the items in the database that I built. My database contains 1300 items - things like Smart’s book (The Makers of Surveying Instruments in America Since 1700), my website, some instrument maker manuals, old textbooks, and a huge batch of other documents I had on my hard drive. Thus, for the most part, a search of my database yields much more crisp and relevant search results than does the same search on the public worldwide ChatGPT database.
A search of my database will still leaving you wanting some of the time, however. AI just isn't developed enough to generate consistently great results yet. So you shouldn't treat the search results as being completely accurate.
Unlike the public version of ChatGPT, my program provides references and links to the sources ChatGPT is pulling info from. Links are not provided to sources that are protected by copyright or otherwise constitute personal information, however.
Finally, ChatGPT doesn't do images, unfortunately. So for now, all search results are text based only.
You can access my database and the ChatGPT powered search engine by going to this webpage: The Complete Surveyor - ChatGPT The webpage includes a few examples showing the types of things you can search for.
Thanks.
Russ
Wow, what a resource. Thanks.
Great resource, Russ. And insightful use of the ChatGPT Plus subscription service. You're right, ChatGPT Plus can't directly analyze attached images, but it can construct Python (or HTML) scripts to invoke the Google Lens API which can bring up references to (and discussions about) an image. Alternately, you can store and access your own library of imagery, with a table structure like this:
CREATE TABLE images (
id SERIAL PRIMARY KEY,
image BLOB, -- For storing images as BLOBs (optional)
image_url TEXT, -- If you store the image in a file system or cloud storage
date_uploaded TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
attributes JSON, -- Store attributes in JSON format for flexibility
description TEXT
);
Then reference your image library database connection,in an attached JSON file to your ChatGPT Plus query, e.g.,
"database": {
"host": "cpanel.infolens.gis-cdn.net",
"port": 5432,
"user": "infolens",
"password": "<password>",
"database": "infolens_infolens",
"schema": "tolleson"
},
Or use a ChatGPT Assistant (as I do) to get more functionality. Again, great work!
Thanks for the kind words. The database is still a work in process. I've doubled the number of documents in the database (now about 2800) and increased the number of words to about 35 million since you checked it out. I'm seeing better responses coming out of the database since doubling the content. Still a ways to go though. My plan is capped at 5000 docs and 60 million words.....
Thanks for your post. Very interesting. I think images would be a great addition, but I am struggling with it so far. I'm not a developer, so I'm in trouble if coding is required.
I am not using ChatGPT Plus - I am using CustomGPT instead. CustomGPT seems to be the best service for me given the amount of content I have (in terms of both docs and words/tokens). CustomGPT expressly states that they do not provide images in responses. It should be possible to have the response include a URL to a relevant picture, however. I am testing that out now and am having just enough success to give me a smidgen of hope.
