Could ChatGPT make America more walkable? — Streetsblog USA

ChatGPT is probably best known for helping students cheat on their assignments, helping internet comedians write fake Hallmark Christmas movies, and, yes, drinking water. But it turns out that planners could use this and similar types of AI technology to make our cities more walkable, bikeable, and simply more people-friendly.

A pair of researchers at MIT and Virginia Tech recently found that two of the best-known large-scale language learning models – ChatGPT 4.0 and Gemini 1.5 – were able to perform many of the tasks of a typical built environment “audit” with a surprisingly high degree of accuracy and detection Presence of street trees, bicycles, and benches correct more than 90 percent of the time in a given sample of Google Maps images.

These programs didn’t do as well at measuring things like sidewalks and streetlights, and both did noticeably worse at assessing rural areas than urban ones. However, researchers argue that even one inaccurate Auditing can help transportation officials do their jobs better — as currently few U.S. cities regularly monitor the overall condition of their roads. Those that do, the researchers say, typically rely either on careful and expensive manual reviews or on virtual models that require planners to learn advanced algorithms and procure expensive equipment.

However, if user-friendly programs like ChatGPT do a passable job, no city can claim that it simply has no idea where sidewalks, bus stops, or trash cans are—and set about actually replacing what’s missing.

“This type of AI tool can increase planners’ productivity, allowing them to spend less time on this type of work (and) have more time holding town hall meetings,” said Kee Moon Jang, a postdoctoral researcher at MIT Senseable City Lab Co -Author of the work. “They can focus more on improving communication with their citizens. … At least that’s our hope.”

The researchers argue that large language learning models can be particularly useful for small and medium-sized cities that can’t afford hordes of examiners roaming their streets or specialized experts running Python codes on high-performance computers. And done right, said co-author Jungwhan Kim of Virginia Tech, they could even help “democratize” access to street information that poorer communities don’t typically get at the detailed pedestrian level.

Nevertheless, Kim warned against cities above-They rely on AI that occasionally “hallucinates” non-existent objects and makes assumptions about gaps in the data that lead to inaccurate results. He suspects this is related to the type of data large language learning models are trained on, which could be a problem as the tool scales to more places.

“AI is a black box; we don’t know what’s actually going on inside,” Kim added. “But one area of ​​speculation is the inequality in training data. We have another research paper that focuses on environmental justice issues, and it actually shows that AI models have more specific information for urban areas compared to rural areas.”

Kim acknowledged that it’s not great that the small rural communities that could benefit most from audits powered by the Large Language Learning Model are also most likely to receive inaccurate results from these tools. And he also acknowledged that even the most informed city may not be able to truly close the walkability gaps that AI helps them, whether due to a lack of funding, a lack of staff or simply a lack of political will.

Still, for cities willing to get involved in the work, ChatGPT and Gemini could provide a crucial starting point for making their streets more people-friendly — even if it’s not a panacea.

“As a geographer and urban researcher myself, we’re not trying to say this is the optimal answer,” Kim said. “But us Are From a more ‘user perspective,’ we try to understand how these tools work and how they can still be very effective despite the risks.”