The SQL is often better than the original...
at least where AI is concerned
I just submitted an abstract to give a talk next year about AI. Not one of those “AI will save us from ourselves blah blah blah” talks. I figure there is enough noise around that signal. I am also not going to dash any hopes or burst any proverbial bubbles — I simply want to make sure we are viewing the whole picture.
Why?
How can we make important decisions about investing in the benefits while simultaneously regulating the harms if nobody wants to have the discussion.
What made this query interesting was the warning in the submission portal that any content generation in the submittal or scheduled talk that was deemed to be generated by AI would be disqualified.
Personally, that doesn’t bother me. I am not a purist but my job as a writer requires me to write. Not to hire a bot to generate text. Automation has always produced something efficient perhaps but grossly lacking in personal connection.
If AI generated content sweeps the digital sphere to bring you an answer to a question or respond to a prompt, inherently it sinks to the level of the available information on a topic.
My data brain while intrigued by chatGPT and its 10X energy increase over a standard Google search has returned to simpler days of SQL. Geospatial Analysis with SQL is an introduction to spatial SQL and the prompts and questions we are able to answer or interrogate.
Writing prompts for ChatGPT are not dissimilar to working in SQL. The question would be selecting mining permits (mp) within indigenous land (il) for example. Depending on the columns in your dataset you tell SQL where the data is located.
SELECT mp.permit_id, mp.company_name, mp.permit_date, il.indigenous_area_name, il.latitude, il.longitude FROM mining_permits mp INNER JOIN indigenous_lands il ON ST_Within(mp.geometry, il.geometry) WHERE mp.status = 'active';SQL will be accessing datasets that you have hopefully vetted versus a black box where AI is often concerned.
It reminds me of when I attended a medical conference during the launch of IBM Watson I asked the leader of the company how the algorithm curates the available medical information on a topic. Surely we can’t glom together industry sponsored research, branded content, and peer reviewed studies. Naturally larger studies followed for longer timelines are more informative than short duration small population studies. Nobody wanted me to rain on the parade of enthusiasts and I never did get an answer to my question.
Yes if you are an analyst, SQL is a worthwhile tool to have in your belt. I am guessing it will look familiar to you once you begin writing prompts and queries…
I will be talking at a free community event, PostGIS Day! Join us if you are SQLcurious…
