At the moment it feels AI is impossible to escape. Google search results return AI summaries, Microsoft apps ask you'd like to ask AI for ideas, it seems like every 
business service is trying to sell you on AI one way or another.
While AI is a transformative technology, its definition has been widened so much in the past decade that its' hard to decipher what "AI" actually means in any context. This is especially true when we discuss autonomous drone technology and 3D mobile mapping. AI has become a blanket term for technologies such as large language models (LLMs), machine learning, and computer vision systems like simultaneous localization and mapping (SLAM).
But all of these are quite different, and understanding the difference allows you to choose the right technology for your surveying needs. As AI technology progresses, having this knowledge gives you the power to make better decisions as new products come on the market.
When most people think of AI, they think of chatbots like OpenAI’s ChatGPT. ChatGPT is an LLM, a language model with self-supervised machine learning which is trained on huge amounts of text. By understanding languages and patterns within languages, the model can respond to complex prompts. But this is mostly restricted to language-based or image recognition tasks, and they can often get off track or hallucinate facts.
You can think of LLMs like a simplified user interface to a vast quantity of human knowledge. But an LLM can’t navigate the real world, they can’t be placed into a machine that navigates around a construction site, avoids obstacles, and completes surveying missions. An LLM can explain how to do that in language form, but it can’t actually perform those actions.
Currently, there are early attempts at “agentic AI”, which tries to take LLMs and connect them to interfaces like web browsers so they can perform real-world activities. But these are rudimentary at the moment, often hallucinating work that never happens or making up facts that don't exist.
SLAM is the technology that powers our Nexys platform and other leading-edge 3D mapping solutions. SLAM uses a series of internal sensors that work similarly to human vision and spatial awareness to understand its surroundings. SLAM uses light detection and ranging (LiDAR) as its main vision component. Laser light is emitted from the LiDAR unit, and that light is reflected back as individual points. Millions of those reflected points are captured and processed into voxels to represent obstacles and explorable space.
When the Nexys is performing an autonomous flight or ground-based mission, it uses SLAM to recognize and avoid dynamic obstacles in unknown terrain. We call this online SLAM ExynAI. It’s a completely different type of AI that serves a different purpose. ExynAI is custom-built from the ground up to understand real-world environments, map them, and navigate around them without user input.
LLMs can’t do this, and there’s currently no way to use an LLM for a task such as navigating through a space or mapping it. Some advanced AI companies are testing out ways to use chat prompts like LLMs to task robots with missions. But once the prompt is sent, that AI ends and an autonomous solution takes over to perform the rest. The LLM can describe a location or physical space it has learned about. But it can’t navigate a real space in the world.
Like any good marketer will tell you, every new piece of technology is going to slap on an AI sticker to try and capture some marketing buzz. But for surveying, you want the right type of AI integration for the right price, without all the hype. Just the steak, hold the sizzle. 
LLM like chatbots don’t offer an advantage for this type of work (yet). They are purely conversational and currently only offer rudimentary control of real-world systems, like web browsers.
Instead, you want to focus on the advanced features of machine learning, post-processing algorithms, and how the mobile mapping system can quickly capture real-world terrain that can be used to reinforce the next learning models. You can think of these as “spatial AI” or “real-time spatial intelligence” instead of chatbots. Spatial AI systems are designed from the ground up to work in the real world, navigating it and mapping it.
LLMs typically need large data centers, and users have to rely on an internet connection to those data centers. On the other hand, ExynAI is standalone edge software and can navigate autonomously or process 3D point cloud data -- including cleaning steps such as noise removal, subsampling, and GNSS/GCP georeferencing -- without connecting to the cloud.
ExynAI powers our Nexys to give surveying teams more options and flexibility to tackle the most challenging surveying tasks in less time than ever before.
Book a free demo of our ExynAI technology and see how it can bring new capabilities to your team with just the push of a button.
