I have a made-up band that is on a made-up tour. There are no real songs, there’s no one in the band, but I still track their setslists each night they play and analyse the statistics. It’s a strange and fun concept that you can learn more about here.
Let’s say the band played in San Francisco last night, and you need to write some automated code to figure out what city they’ll be playing in next. Before chatGPT, I’d write some code that did this:
- Pull down the last 5 cities
- See if there’s a general direction. For example, maybe the last shows were played in San Diego, Los Angeles, Sacramento, Oakland, and San Franciso, which means we’re moving north.
- Find an API that lets me enter a city name and find another city name within x miles. Ideally I can point the direction north so the next show ends up in Portland rather than back down to Santa Cruz.
- The distance shouldn’t be standardised. After going to Seattle and Vancouver, it’s common for bands to fly over to the east coast and start the next leg of the tour there. So “find a show within X miles of the previous show” is too limiting. You also need some concept of appropriate times to fly to a new section.
Frankly, this code is too finicky and boring for me to spend time on. I’m not interested in writing a bespoke algorithm for touring, so I’d just continue writing the tour schedule by hand because there’s no other way worth my time.
But now now that I have chatGPT, figuring out the next step on the tour has turned into a one step process:
- Tell chatGPT “You are planning a tour across the world. Shows usually happen in bigger cities, and if the cities are big enough, you can play 2 or even 3 nights in a single location. The previous five shows before this one have been San Diego, Los Angeles, Sacramento, Oakland, and San Franciso. What should the next show location be?” and get Portland, Oregon.
Think about how much that will change how we write software. Instead of hand-editing an algorithm from scratch, inevitably finding edgecases that don’t make sense, and having to maintain the code forever, we can just ask AI to provide the algorithm in a single call. This is a big deal with enormous implications.
For what it’s worth, here are the next 25 shows on the tour, with ample space for traveling between each city:
- Portland, Oregon, USA
- (1 rest day)
- Seattle, Washington, USA
- (1 rest day)
- Vancouver, Canada
- Vancouver, Canada (Second night)
- (1 rest day)
- Calgary, Canada
- (2 rest days)
- Edmonton, Canada
- (1 rest day)
- Winnipeg, Canada
- (2 rest days)
- Minneapolis, USA
- (1 rest day)
- Chicago, USA
- Chicago, USA (Second night)
- (1 rest day)
- Detroit, USA
- (1 rest day)
- Toronto, Canada
- Toronto, Canada (Second night)
- (2 rest days)
- Montreal, Canada
- (1 rest day)
- Boston, USA
- (1 rest day)
- New York, USA
- New York, USA (Second night)
- (1 rest day)
- Philadelphia, USA
- (1 rest day)
- Washington, DC, USA
- (2 rest days)
- Atlanta, USA
- (1 rest day)
- Miami, USA
- (1 rest day)
- New Orleans, USA
- (1 rest day)
- Houston, USA
- (1 rest day)
- Dallas, USA
- (1 rest day)
- Austin, USA
See you on the (made up) road!