AI Took The Wheel On My Roadtrip And It Was Ok.

"You should let the AI decide your roadtrip," one friend suggested. I swirled the remnants of my tiki drink in my class and mulled this over. It was true that I only had a vague concept of how to reach Minneapolis, as well as a broad timeline. Four days? A week? I had done a quick search of the travel time between Vegas and Minneapolis – roughly 24 hours – and knew if I broke that into chunks, I can ship away at the miles until I had finally reached my destination. I had gone meet my friends in Vegas with only a estimate in my head.

So it wouldn't hurt to take a chance on AI. In the lulls of the conversations I started to pull out my phone and ask my favorite AI, DeepSeek, where I should be going on my way to the wedding next weekend.

The First Lession: AI Cannot Account For Or Predict Day-To-Day Fires

When I informed DeepSeek of my plan, it was very happy and willing to help. It asked my preferences in what I enjoyed--nature, Historical landmarks, or roadside attractions. I had told it to avoid the historical landmarks and focus on nature. It suggested my next stop should be One of the several national parks in Utah, presenting me with a list. I settled on Bryce Canyon since it wasn't too far out of the way and I hadn't bee there before.

After the concert and departing from my friends, I made a last minute reservation on Recreatioin.gov before heading off into the desert. (Pro tip: most National Parks have last minute spots available Mondays - Wednesdays if you can take a day off from school or work.) As I left Vegas and attached approached the park, I noticed the sky on the horizon became very foggy. Once I reached the fog it became clear to me as a California Native what this was--the ash of a wildfire. I notification appeared from the Reaction.gov site indicating that there was bad air quality.

As far as the AI could ascertain, this was still an excellent camping spot. I feel this is a momentary setback for AI as of now, it only knows what its modeled after and cannot manage literally fires in real time. This was not the type of weather I'd want to be outdoors for.

I had asked for other options. While it give me a list of other excellent camping spots in Utah, none of them were something that I could pivot to given my direction northeast for my roadtrip. This is somewhat of an issue I've seen with AI. Once you are on your way and reach a snag, AI tends to want to have to travel backwards and start over.

I stayed the night but the smoke rolled thickly into the canyon that I opted not to do one of hikes that I was hoping to do on my travels. As disappointing as this was, I knew I would get at least one good hike in later--there was still time.

I packed up early, eyes watering, and headed onto the next phase of my digital sojourn.

AI Will Get You 90% of The Way, But You Still Need To Do The Footwork

The night before I had made the discovery that I had forgotten to pack a book to read while at camp. I asked my AI friend to suggest a good book store between Bryce Canyon and Minneapolis, and it suggested The King's English in Salt Lake City.

I pulled into a rest stop after getting out from down wind of the fire and asked my AI Road Trip buddy for an itinerary to Salt Lake City. I was surprised that not only did it find some landmarks along the way, but also generated a google maps link to give me on my way.

The first stop, Big Rock Candy Mountain, sounded like a delight little place to stock up on sugar and mirth for the long road ahead of me. As I passed through some of the canyons and valleys that made up Utah, I found myself pulling into a side road and stepping out in a derelict rest stop. NO POTABLE WATER read the signs next to a restroom.

I checked my map again, and it would seem this is where my AI wanted me to go. There was a rocky mountain next to me, but no real indicator of anything that this was what I should have expected to see. My experience with AI told me that perhaps I was in the ballpark but not quite at the true Candy Mountain.

Being in a canyon I had not enough reception to do a search online. I opted for the old tried and true method of finding one's destination: I wandered around aimlessly until something "felt right".

One exit over and I discovered the Big Rock Candy Mountain Resort next to the true Big Rock Candy mountain. Not where the AI instructed me to go but pretty close. Elated, I stepped inside to purchase my reward. While a fun stop, it was not the largest selection of candy.

The same scenario played out again at the next stop, Needles. This was a ghost town that was washed out by a landslide and never rebuilt. Unfortunately I couldn't locate it at all given the AI's directions and to turn around would have been difficult given the location. I decided to continue onward.

When I reached the book store, I asked the AI for a book suggestion, and once again the same problem. It gave me a list of good roadtrip and camping books, but I could not find any where they should be located in the store. Eventually the Ai suggested I ask a store clerk, still not no avail.

Once again I began to abandon the AI and take matters into my own hand. I have been on a kick of philosophy and esotericism, and decided to look in that section for a book to read. To my surprise, it was in this section that I found Desert Solitare--the book that my road trip buddy DeepSeek and suggested.

Why didn't I find this book sooner? I had misunderstood the classification of the book as not a travel book but non-fiction. As the Ai had given me a cornucopia of suggestions I hadn't mention the book by name to the clerk either, only the board stroke of "where is the travel section?"

As an aside, I have noticed that certain AI will behave in different manners when trying to assist you. I enjoy DeepSeek as it will present me with options and not be deterministic on a solution. Claude, what I use when I am coding, is much more deterministic with his suggestions and often I find myself arguing with its execution. So this book debacle may be a result of my personal preference and human mannerisms clashing with the programmatic intentions of an AI.

After the book store, the next stop on my list was one that excited me from the oddity of it.

With AI You Are The Captain, Not The Navigator

After I made the exodus out of mormon country, I found myself heading to the next stop on the AI's master travel plan: Hell's Half-Acre, also know as that place where they filed Starship Troopers.

I had given up on using the map DeepSeek and provided me and started looking places up by their names. As I drove my trusty muscle car deeper into the barren corner of Wyoming, I peeked at the distance I was to go down a certain road; 100 miles.

I panicked. No where behind me or in front of me did I see any other cars, buildings, or signs of civilization. Water towers on the horizon had become my guiding bacon in the bald and barren landscape. Not I saw only cracks in the road and long flat hills until the horizon.

I took a leap of faith When I saw my first land mark of any sort I anxiously pulled over. There I came across a sign that explained that I was on a common path to the west – THE OREGON TRAIL!

As I stood on the banks of the Green River looking at a old ferry boat that crossed settlers over a century ago, I realized ultimately what the AI had accomplished. My trip was based on popular knowledge. This path I was on was the same taken by the riders of the pony express, emigrant settlers long ago, the same as the indigenous people before them. This was the path that folks like me learned about in elementary school playing Oregon Trail on flickering CRT screens with 256 VGA colors.

It was at this point that my anxiety subsided about being alone in the empty grasslands. I was in the guiding hands of collective nerds. I continued on my way to Hell's Half-Acre, which to its name felt like a microcosm to the Badlands National Park miles away.

The Hidden Truth of Getting What I Asked For

On the last stretch of the journey I traveled through South Dakota. I found myself going to two last tourist destinations, the famous Wall Drug and the Corn Palace. As popular as these were, I happened to come across something I had seen in photos but didn't realize it existed at a visitor center.

I found Dignity, a million dollar statue of a Lakota inspired woman on a bluff overlooking the city. I had seen photos of the status a few years ago and was curious to see it from the side of the road. I had instructed the AI to show me kitsch and natural landmarks, and the avoid historical monuments as they often frame things from the settler's perspective and do not adequately present the indigenous people's voices.

While, this statue seems to fall into the realm of fetishization of indigenous people similar to the Crazyhorse Memorial, I felt it was odd that the AI attempted to get me to go to that monument but not this particular statue.

This I feel is a hidden truth in plain sight when using AI. Its still very biased to the information that it is getting fed, and the gatekeepers of that information are still a specific gender and skin tone.

At Wall Drug, famous for its signs announcing its existence hundreds of miles away, still has antiqued posters explaining the importance of learning of Custard's last stand. There was a gift shop selling native inspired ceramics. Nearby was a giant dinosaur that roared very few minutes. I asked for kitsch and nature, and received explicitly that. Kitsch in some way lends itself to the theme of appropriation, which I had overlooked.

The Road Back Home And Beyond

On the way back home from Minneapolis, planned my own path and took an MVP route back to California. I drove on only three interstates, straight lines following cardinal directions, no scenic routes. I managed to get home a day early and under budget.

I did however stopped by a dive bar to grab lunch and a drink with a friend. She stands on her scruples, which I've grown to appreciate. I had briefly mentioned my use of AI for my road trip, and they pointed out their disdain for it from the amount of energy it devours. The AI is constantly learning, more than 4.4% of all the energy in the US now goes toward data centers, according to a recent article in MIT Review.

My only rebuke to this was a meager, "The problem is its not going anywhere, and in my line of work its being pushed hard. I need to be in front of it somehow." For what its worth, I used one of the less energy-draining AIs on the market at the moment, DeepSeek.

I have been enjoying my experiences using AI, but I think overall its still something that is growing and might be a while until its where it needs to be. I wonder if sometime down the line, I will be able to hop into an automated car like a Waymo, tell it to take me to Chicago, and stop at all the tourists traps along the way. Will that cost the energy of heating a small nation to get that next refrigerator magnet autonomously? I have to remain hopeful that with the Janus-faced aspect of life that is progress, we eventually find a balance to it all.