Bryn Fosburgh on AI Hype, Hallucinations and the Importance of Human Intervention

Bryn Fosburgh likes to describe himself as a “simple Wisconsin boy”, but his achievements, experience, and knowledge over his 40-year career in the geospatial industry are anything but simple. Although he’s done business in more than 100 countries, visited every continent, and managed 4,000 employees, at his core Fosburgh is a surveyor. It’s how he “grew up” in the industry and it still informs his approach in navigating technological advancements with colleagues and customers.

Following his standing-room-only keynote presentation on digital transformation in the geospatial industry at the GoGeomatics Expo, Geospatial Journalist Mary Jo Wagner sat down with Fosburgh, who is a Senior Vice President at Trimble and an advisor to CEO, Rob Painter. In a wide-ranging chat, Fosburgh explains how COVID has been a boon to the industry, what he fears about AI and the surveying discipline, and how automated workplaces are the future.

You were involved in geomatics long before “geospatial” became the umbrella term for the industry. And as a 40-year veteran of the industry, you’ve seen a multitude of innovations and developments. Which advancements from the past few years have excited you the most and which do you foresee will have the biggest impacts going forward?

The digital transformation of our industry. In the early 2000s, we had the vision of a connected site. And that connected site could be connected construction, it could be agriculture. What really drove this idea was the integration of the workflow from A to Z. That’s the key part of it. Telematics is an important part of it, but the integration of disparate tools and datasets across the workflow is what really changes the industry. It improves productivity, reduces rework, gives you more transparency, improves your safety, and an unforeseen outcome of that connectivity has been sustainability.

COVID was an unexpected accelerant of this digital need because suddenly, you couldn’t be on site, you couldn’t share paper. You had to visualize three or four different sites at a time; you had to look at them in a much different way than we were used to. And if you look at the number of video conferences you had pre -COVID, there were all these fairly disjointed or fragmented medias that weren’t that good. So most of your digital communication was through some type of phone; it wasn’t a video type of feed.

So COVID really drove digital transformation. That’s everything from cloud enablement, from integrating different workflows, and by looking at 5G or other communication sources. So, to me, that’s actually become the biggest advancement. Everyone talks about SLAM and artificial intelligence (AI) and drones and those are all cool technologies that enable our work as surveyors or as contractors, but this digital transformation process is something that’s really transformed the industry.

The geospatial market is forecasted to grow significantly in the next decade. Do you think AI developments will be a driver for that growth and in which particular industries?

I don’t think AI will be a driver. I think AI will help enable other industries to better utilize geospatial data. One example would be autonomy. You’re starting to see more and more autonomous vehicles and the discussion of autonomous vehicles, and you’re seeing more robots. For both of those applications, the geospatial backbone is essential, particularly for autonomy where roadway accuracy is key and real-time updates are very important.

Industries that use it now are utilities, construction, oil and gas, mining, the traditional surveying and mapping type industries. But a growing user is agriculture, where you’re starting to see drones deployed to manage fields and manage spraying. There are huge multi-rotor drones that are attaching sprayers to their payload. That’s a big change in that industry. So that needs geospatial data to be effective.

You are currently the president of the World Geospatial Industry Council (WGIC), the foundational president of FIG, and a board member for a few organizations such as Wingtra. What is top of mind for these organizations and how can they impact the growth of technology across diverse industries?

For WGIC, interoperability is important, but in a different way than OGC’s interoperability focus. OGC talks about interoperability from an enterprise or a GIS aspect. WGIC considers how do we interoperate between competitive companies’ tools? I’d love for people to only buy Trimble stuff, but that’s not realistic. No single provider will likely have the right solution for every user’s specific need. So, how can we as technology providers work together and still have our competitive advantages that we believe we have, but still provide that customer interoperability? That’s really important from a WGIC perspective.

Second is just advocacy of the profession. And through that advocacy, we’ve got membership from people you wouldn’t expect, such as people involved in oceanography or meteorology or geology; other disciplines that use geospatial data.

Third is data. How do we deal with all the data? Data comes from multiple sources today. Let’s take the autonomous car. You may get GNSS or satellite-based data. You may get data from inertial navigation. You may also get data feeds from what I would call crowdsourcing. So how do we manage that data to ensure that Uber or that autonomous vehicle brings me from the airport to my hotel. So we’re looking at not only data interoperability, but how do you classify that data? How do you make sure that data is accurate? How do you know the source of that data? And how do you process the volumes of data to improve your work? We have more data than we know what to do with in many cases.

At Wingtra, one of the things we’re looking at is how to streamline processing. Today the whole processing is very fragmented. I may process data in one software package, design it in another, and I may then send it out to a deliverable to Bentley or Autodesk or Trimble. It’s not streamlined. That’s where AI and the integration of workflows are essential.

Today, I can get this massive point cloud of data collected via whatever data collection means I want. AI can then look at it and say, here are all the fire hydrants, here are all of the electrical units within a building. It can do something that was very laborious from a human perspective, and it can automate that and really improve the accuracy. However, it doesn’t replace the human.

We’ve actually used AI for a long time. As a simple example, my spelling is atrocious and my grammar is even worse. When I type something in Google Docs and I spell it incorrectly, it gives me a list of what alternatives I can use. It’s not always the right word. The grammar may not be the right grammar based on the context in which I’m trying to say it. It still requires human intervention. But, will AI automate a lot of things in the future? Absolutely. Will we get to the point where we really, really trust it? Yes.

How would you describe the current adoption or state of AI integration in vertical markets such as construction, utilities, and infrastructure? Where do you see the most prevalent use of AI?

It‘s definitely in its infancy. It’s very much like how the Internet was back in 1999/2000. I remember living in Silicon Valley at that time and everything was a dot-com. Even if it wasn’t a dot-com. That seems to be the same pattern now with AI––everyone is jumping on the AI bus. And that is a fear of mine because people then get a wrong view of what AI is and what it can do.

So although AI is very much in the infancy in our industries, there are some established uses for it. For one, it’s excellent for research. I can use it to look at a whole bunch of different research papers and coalesce it down to main points. I can take three or four of our patents and quickly synthesize what they do. It’s really good for that today.

Classifications. It’s fantastic for looking at large data sets and classifying them. In industries like utilities, for example, which are so asset-heavy, AI-based classification is perfect for looking at drone-based corridors or towers and quickly classifying the assets. It’s also great for menial, repetitive tasks like processing or analyzing a bunch of invoices.

You mentioned that you’re fearful that too much hype of AI can create disillusionment about what it can do. Does anything else about the buzz around AI give you pause?

I’m super hyped about AI. But it’s purposeful hype. I want the hype around AI to focus on solving a problem, either internally or externally.

As I said before, AI-esque functionality has been around for quite a while. For example, Trimble acquired a company called Quantum and one called e-Cognition. Both did a version of AI and had language models. We just didn’t call them that. Leica acquired a fantastic company called Cyra back in 2000. They had what you would call a language model at the time for doing classification on interior scans. It was very targeted and labor intensive, but it was cool. So we’ve had AI capabilities for two decades but they weren’t recognized or labeled as AI then.

As AI matures, it’ll continue to support eCognition, our photogrammetric processing software, and Trimble Business Center. And AI is already fantastic at asset classification and analyzing large datasets.

From a surveying perspective, AI should not take away good practice that we as surveyors follow. Because AI can hallucinate; it can provide some funky results. So, as surveyors, you still have to be knowledgeable enough and think critically to identify those hallucinations when they arise.

You learned the fundamentals of surveying by turning angles with a theodolite, measuring distances with an HP EDM and performing 16-position azimuths. Now positions and measurements can be acquired in seconds with push-button reality capture technology and AI-based analysis. Is there something to be said for having so much automation that we lose sight of the fundamentals of the surveying discipline? 

That is a fear right now, but that’s not just with AI, that’s for the entire digital transformation process. You need to know how that data was derived, how it’s going to be used, the accuracy that’s needed for its use, and the equipment and workflow you need to create the right deliverable. That’s essential. We saw that challenge with GPS and then with laser scanning with its push-button functionality. These digital tools are fantastic, but you still need to understand why you’re scanning that structure, what you need to get out of it, what accuracy you need and how you can verify that what you’re collecting is good-quality data. Because we’re dealing with millions of points now. This again is where AI helps. You can train that AI engine to, for example, stitch together multiple scans and turn billions of points into functional models.

How is the advancement of reality capture and AI changing the role of the geospatial professional?

More and more, we’ve become data managers. That is key.

I think a lot of people are worried about it replacing our jobs. I started on a T3 and used an HP EDM, and then we moved to a satellite receiver, and a total station, GPS, GNSS, and then we moved to scanning. Throughout that progression, people haven’t been replaced. Their role has changed.

Today they have to become more technology focused and more technology literate. That’s the benefit of our industry having a decrease in population. Much of us are getting older and we’re getting out of the profession, either voluntarily or involuntarily. And so young kids know technology and enjoy technology. We’ve done a poor job in the industry to excite kids into it because it’s quite exciting. We strap a scanner on a Boston Dynamics dog and run it around and every kid in the world loves it. We did that at Con Expo in 2020. It got more people to the booth than if I would have given away free equipment. We’ve got to do more of that from an industry perspective to attract kids who understand the theory, but then truly understand the technology and how to apply it. Technology is what they know and love.

I fault educational institutions for this, partly. I’m very critical of educational centers because I think many still live in an older curriculum and with older technology. We need to, as an industry and with events like the GoGeomatics Expo, really encourage them to change their curriculum to attract more and more young people.

And that’s what I do with FIG. At FIG we work very closely with our foundation and young surveyors. I spend a significant time with young surveyors worldwide, trying to grow and excite them into the profession. I see a lot of interest outside of the U.S. That doesn’t mean Canada and the U.S. don’t have interested young surveyors, but I would say in the emerging economies, they’re very, very excited about it. And they’re not caught in the old way of traditional methodology or process.

There are many thoughts on how the next geospatial digital revolution will take form. What would you like to see happen in the next five years? 

I think there are three value adds that will continue to transform our work.

The first is BIM, which is more process-oriented, and will have a significant impact on construction design, building, and operating buildings.

Second is the digital twin, which is becoming more and more common. For me, a digital twin is not just a digital representation, but it’s looking at the entire dynamics of an object in real-time and providing the ability to interrogate that object.

And third is AI, which is a set of enabling technologies.

Those three together are going to create a more and more automated workplace. And in relation to the digital transformation pieces, you’re going to want to visualize that automated workplace. Today you see this in factories; but you’ll see it more and more in construction, mining, and agriculture.

I believe you’ll also start to see more visualization in land information. For example, say I have X hectares of land. And I want to put 10 homes on this land. And I want each property to be roughly this size and the homes to have an east-facing view and multiple access points to these roads. And I want the mean value to be this. You will be able to use technology to take all those parameters and automatically map out that subdivision.

I believe that can already be created today. That’s what Quantum does for linear infrastructure. What Quantum does for railway routes, we can do for housing. We as surveyors are going to have to be involved in that data management and that project management.

As digital advancements such as reality capture, BIM and AI continue to march the industry towards a more automated future, it’s likely reassuring that thought leaders like Fosburgh see an essential place for surveyors, mappers, and engineers within that new reality.