While we are still very early in the development of generative AI and use of large language models (LLMs), I thought it would be interesting to see just how close an LLM could get to correctly addressing a common GIS 101 project.
I chose to start by using Google’s Bard as it is supposed to be able to access current data from the Internet. It’s unclear to me that this capacity helped.
I gave this prompt to Google Bard: You are a city planner and GIS analyst. What are the two best locations to build a new McDonald's fast good restaurant in Leawood, Kansas
I left the typo (“good” should have been “food”).
Bard gives us a reasonable sounding response. Unfortunately, both locations used as an answer either already have a fast food restaurant at the location or very close. However, the stated factors used in creating the analysis were on par with a typical undergraduate student response.
------
I then gave this prompt to OpenAI's ChatGPT 4 (no extensions):
ChatGPT identified many of the same factors that Bard did and added a few additional, including market analysis. ChatGPT returned these two suggested locations:
ChatGPT’s second recommendation of State Line road is reasonable, however there are already McDonald’s at 79th, 104th, and 135th In Leawood. Several competitors are scattered between those existing McDonald’s locations.
-----
For me, the take-away is that the LLMs are identifying factors to consider, but no supporting data sources, (geographic) analytical procedures, or useable responses. As with my other AI "experiments" to date, the results while not great yet, are amazing considering what we had access to only a year ago. The obvious question remains, where will this analysis be a year from now?
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.