Leveling up AI: what artificial intelligence means for mapmaking

It is just over a year since ChatGPT, a chatbot based on a large language model (LLM) became the fastest-growing software application in history. For those of us who have embraced this type of artificial intelligence, life — and work — will never be the same again.

ChatGPT has reached several milestones since then, with other LLMs following suit, but it appears this is only the beginning. The process of creating individual Custom GPTs with specific areas of knowledge, skill and instructions has now begun — and given the exponential advancements in this area of technology, everything you are reading now could already be obsolete by the time it is published.

Of course, anyone who remembers chess master Garry Kasparov's 1997 defeat by the IBM supercomputer Deep Blue is well aware that AI is not new.

HERE360 explored how the technology has already been successful in several industries for many years, including route optimization, in our blogs this year. Companies such as HERE customer Satalia uses AI to make quick calculations from billions of possibilities to solve difficult problems, including ETA calculation. HERE UniMap uses AI models to revolutionize mapmaking, promising fresher, more detailed maps and greater synchronicity.

Yet many have raised concerns not only over potential ethical problems with AI, but exactly how humans will work in tandem with it. Can we always trust the results it gives us? Will it steal our jobs? And how can we use it best to make our operations more efficient, productive and successful?

AI at HERE

UniMap was announced at last year's CES, but now that another chapter of the technology conference has come and gone, the world looks very different.

For mapmakers, AI has made the process not only speedier but also more insightful.

Martin Hasse, Head of Social Advocacy at HERE and a data visualization expert told HERE360: "AI-powered data analytics and visualization capabilities are revolutionizing our interaction with geospatial data. This marks a significant leap in how we at HERE Technologies can simplify highly complex information into digestible insights — as well as unlock predictive patterns to showcase how HERE can enhance decision-making across various industries."

HERE Principal Process System Engineer Naman Patel said the company is looking at how to use AI in all functions, since all functions at HERE ultimately support the mapmaking process.

“An example is data consumption," he said. “Our over-simplified classical data consumption pipeline would look like us getting data from various places and loading it into a data lake, database or data warehouse. People get access to that database using various business intelligence applications such as Tableau." 

While that works for the use case it was designed for, it does not provide for future use cases the developers have not anticipated.

“That classical pipeline has limitations," Patel said. “Generative AI is not yet fully mature, but somewhere down the line, hopefully in the near future, it will be mature enough to talk to relational databases.

“If we can use that technology, the dashboards that we build now will be more of a supplement to generative AI technology."

While today, business users consult specific dashboards to find the information they need, in a few years from now, they might be able to ask a chatbot directly that would find that information in the vast network of relational data for them.

Getting chatty

One of the issues many users highlighted with LLMs that are available today is their tendency to hallucinate. Partly due to the training dataset and methods used, this is when the LLM cannot find the right answers so simply invents them.

Custom chatbots have the potential to eliminate or at least reduce this problem, Patel said.

“I built a proof-of-concept (POC) chatbot for our internal teams that I have been testing for the past seven or eight months, and not encountered hallucination with it as much as with ChatGPT," he said.

This also gets to the heart of why people are skeptical about AI: the idea of relinquishing control completely to a machine. But Patel points out that this approach is neither necessary nor desirable.

“The human will always stay in the loop," he said. “In the example of relational databases, humans must decide which data the generative AI uses, and put governance and control in place."

Not every user in a business will have access to sensitive information such as the organization's financial data, for instance.

While the focus has been on how existing jobs will be affected by AI or even axed altogether, the importance of new roles, such as prompt engineer, is just becoming clear.

Ultimately, Patel said, you might not lose your job to AI, but an organization that does not harness its power will lose out to an organization that does. “If your product is good for today's market but does not come hand-in-hand with AI, it may give a competitor's product built with AI the upper hand in the marketplace," he said.

Maps made by HERE that power functions such as automated driving have incorporated AI for specific tasks since before 2023. “It isn't that one big AI will solve every problem in autonomous driving," Patel said. “It is more that multiple smaller AI solutions will be implemented at various levels working in harmony with classical code to solve the bigger problem."

For instance, specific processes such as object detection and image recognition that are necessary for hands-free driving are powered by AI algorithms. They are less likely to be changed soon by generative AI. There are already powerful AI models for these tasks.

“These have been part of our processes for a while. It's not that generative AI is going to change those things overnight that we are already doing with classical AI," Patel said. “Generative AI has definitely brought more spotlight into this area.

“But as a company, we have been using AI for quite some time for mapmaking; it is just that now with generative AI, it has more potential to spread to more parts of the organizations that support the mapmaking process."