Thursday, March 30, 2023

Can AI Models Make Maps?

I asked Google's artificial intelligence large language model (LLM) chatbot Bard: 

How can Artificial Intelligence be used to create interactive maps for the internet?

Bard replied that Artificial Intelligence (AI) can be used to:

  • Automatically generate maps from satellite imagery. AI can be used to identify and classify objects in satellite imagery, such as buildings, roads, and trees. This information can then be used to automatically generate maps of a given area. 
  • Predict the location of points of interest. AI can be used to predict the location of points of interest, such as restaurants, gas stations, and tourist attractions. This information can then be used to create interactive maps that show the location of these points of interest. 
  • Detect and track changes in the environment. AI can be used to detect and track changes in the environment, such as deforestation, flooding, and erosion. This information can then be used to create interactive maps that show the changes that are taking place. 
  • Provide real-time traffic updates. AI can be used to provide real-time traffic updates, such as the location of accidents and the speed of traffic. This information can then be used to create interactive maps that show the current traffic conditions. 
  • Personalize maps for individual users. AI can be used to personalize maps for individual users by taking into account their interests and preferences. For example, AI can be used to show users only the points of interest that are relevant to them. 

Bard's response to my query provides an interesting overview of how image identification models are being trained to search for patterns in satellite imagery and to detect changes in satellite imagery over time. However I actually wanted to know how I might use ChatGPT or Bard to save me hours of work by making interactive maps for me (which shows how important it is to carefuuly phrase your interactions with LLMs).

Over the last few weeks I've been reading with interest on Twitter how developers have been using AI models (mainly ChatGPT) to create Overpass Turbo queries, to display requested OSM features on an interactive map, and how to make ChatGPT Generated Web Maps. My own personal attempts to get Google's Bard to create Overpass Turbo queries or valid GeoJSON have failed miserably. I think that at least fot the moment it that ChatGPT is a much better GIS practitioner than Bard.

If you don;t have access to ChatGPT or Bard don't worry. There are already two neat services which use AI models to help you easily create maps simply by typing a request or question:

Enter a question or some text into Texttomap and it will attempt to answer your question and show you the results on a map. For example the map at the top of this page was created by asking Texttomap "Which are the four longest rivers in the United States?" Other suggested prompts (by Texttomap) are "Which countries took part in WW2?" or "What should I see when visiting Paris?"

Once you have created a map with Textomap you can share a link (such as this map of 'What should I see in San Francisco?'). With a paid pro account you can also embed any created maps in your own website or blog.

mapsgpt can also be used to create interactive web maps simply by entering a few text inputs. For example here is a map built from the input Show me restaurants near Spitalfield, London. mapsgpt will send you a link to any created map if you tell it your e-mail address.

No comments: