In a quick exploration of whether AI can supplant search engines, I used ChatGPT to choose some running routes while I’m in Lisbon. I’m not a planner when it comes to trips, and GPT-4 is now my personal assistant, junior programmer and copywriter, so it seemed like a natural fit to ask it this question. Sure, there are plenty of matches for this question if you search, but I wanted to give it some additional parameters: length between 3-7 miles, wide-ish paths not super cluttered with tourists, good scenery, Metro-accessible and low to moderate inclines (side note, if you do want steep inclines, run the hills and staircases in the areas around the São Jorge castle for maximum burn, it’s fun, but note you’ll be dodging a lot of the aforementioned tourists). These are the kind of parameters that would have required reading through and filtering numerous search engine matches to get the right answers.
With all that guidance, here’s a screenshot of its recs and a quick review of how it did. I was able to try out four of its five recommendations.
1- Parque Eduardo/Avenida da Liberdade has moderate inclines and declines in the park, and it’s worth seeing both at least once.
2 – All these landmarks in Belém are a must-see, but the distance is off – maybe that’s 4 miles if you loop it twice. End it at the monastery for the pasteis.
4 – My preference is to take the metro and start or finish this one at Parque das Nações; it’s the metro station designed by Calatrava, worth seeing. Great views on the route and less crowded than Belém.
5 – Lovely flat stretch of waterfront. You can run on the roof of MAAT and through their garden; it’s beautiful.
Overall – not bad, ChatGPT, not bad at all. Wonderfully scenic recs that mostly matched the parameters.
Using ChatGPT’s ability to generate Swift, it would take like 30 mins for me to build a “RunGPT” app that used the API to give advice like this based on location/distance/terrain preferences, but remembering the early days of people blindly following Google Maps and filing lawsuits… nah. Standard disclaimer: exercise due diligence with any AI output, and after that, have fun out there.