A common metaphor to explain AI is the historical comparison to horses versus steam trains.
When the first steam trains were developed, it was as messy as today. The first prototypes often didn't start, and if they started, they broke down in the middle of the journey. Established horse carriage companies laughed at this. Only when the trains worked reliably a few years later, the horse-powered companies stopped laughing. They went out of business quickly.
AI is often compared to steam trains: a powerful technology with teething problems. Once solved, it will conquer the world, like trains did. To me, this metaphor is only half true. The promise of a train is easier understood than AI's. Comfortable transportation from A to B for people and goods is a clear value, and its success is easy to measure.
Capturing AI's promise is harder, here's what Gemini 2.5 Pro thinks:
“At its core, the promise of AI is to amplify human potential and solve humanity’s grand challenges.”
When asked how to measure AI’ssuccess, Gemini goes really broad, from measuring GDP for general productivity growth to the “scientific discovery rate” tracking significant papers published. I didn’t bother to ask how significance is defined here. 🙃
To me, AI is closer to the steam engine than to steam trains. Anyone with a basic understanding of mechanical work intuitively gets that a machine that produces a constant spinning force must somehow make mechanical work easier. Building a steam train or a steam-powered factory is much much harder to imagine.
AI is the same here. Anyone with a basic understanding of knowledge work intuitively gets that AI must somehow make knowledge work easier. What we're seeing right now is a huge playground of experimentation of how to accomplish this.
Going forward with this metaphor, also the measurement part becomes clearer. In the AI world, we see a lot of LLM performance comparisons. This feels similar to steam engines. How much power can they output for how long with how much fuel?
The more interesting question would be: How fast and reliable can your steam train go? Or how many items can your steam-powered factory process per hour?
Some of today’s AI-powered applications are more advanced. Software Development with Claude Code or Cursor already is a game changer. It’s comparable to a steam train, enabling one trained person to achieve what needed 20 people and 80 horses before. In general, the vast majority of applications still lacks proof of usefulness. To me, this is actually really good news. It means you’re not too late to start getting on board the AI train and start experimenting and understanding this technology.
Returning to the steam engine metaphor, I hope AI forces us to really understand knowledge work. The people who understand knowledge work the best are in the best position to automate big chunks of it.
What to Print this Week
This newsletter started out on 3D printing. I you haven't had any contact with it, you should, it's great. Here's the most interesting and funniest projects I saw last week.
This looks like a handy helper to clean the filter of my robot vacuum. I love how creators link to other helper projects around robot vacuums on their pages.
|
|
Roborock Filter Cleaner
|
The marketing is just great: "Have even louder children!" Exactly why I will NOT be printing this... 😂
|
|
Cardboard Spool Megaphone
|
A Lounge Chair on your Keychain. With crossbody bags established for all genders, you can finally have these bulky keychains, too!
|
|
Mini Lounge Chair Keychain
|
|
|
Hi 👋, I'm Stefan!
This is my weekly newsletter about new technology hypes in general and AI in specific. Feel free to forward this mail to people who should read it. If this mail was forwarded to you, please subscribe here.
|
|
