Skip to main content

Here at Spectus, we see innovation as a key differentiator and a cornerstone of our strategy. Artificial Intelligence has long been part of this strategy, and we have used deep learning algorithms to derive insights from our location data. Historically, however, we had not considered the applications of AI methods to non-processing applications, such as natural language processing. Recent advancements in large language models, most notably the release of ChatGPT last year, have changed that. The growth we have seen over the past year in this space is undeniable, and has led us to consider how this technology can be applied to geospatial systems. The following is an exploration of potential ways to incorporate these learnings.

Chatbots as a concept are nothing new. They have become a prominent means of support, suggesting the right content or documentation. This type of role as an AI-based assistant, while still seen as fantastic just ten years ago, is today seen as a very pedestrian application. But what if we take this idea one step further?

While linking to a specific static document or even a relevant snippet could be useful, a next level approach might be to actually adapt the static example to the specifics provided by the user. A sample conversation between a user and chatbot may look like this:

The above example would require some base knowledge of reality but also plenty of examples from domain-specific queries to understand the proper formatting of each column. While the model is trained to use Trino/SQL and Python with all the right modules, the key for this approach to be successful would be to input the context of our platform. Success at this level of complexity may look like this:

All of the examples above have the chatbot generate code snippets. But what if the user brings their own code?

The chatbot could help troubleshoot or debug problems in the user’s code implementation. This is where the chatbot is really powerful—it goes beyond compiling and contextualizing documentation and actually demonstrates an understanding of the user’s question. For example:

Utilizing chatbots in geospatial analysis is a new frontier, and Spectus is excited to at the forefront of this innovation. We are hard at work finding new ways to implement large language models to make our platform even more accessible and easy-to-use.

To learn more about the latest at Spectus, book a demo of our platform.