Computer adventure games were possible in the 1980s because of a bit of code called a ‘parser’. You could type, “pick up the axe” and the computer would understand the phrase and follow your commands. In italics, because it didn’t understand anything, it simply broke your sentences into bits and changed the state of your inventory accordingly.
When faced with a parser, even a primitive one, many people did that homunculus thing and decided that the computer could understand every single thing you might type, like, “I’m thinking that having an axe in my inventory would be helpful,” or even, “let me tell you about my cousin…”
My first gig, at Spinnaker, was leading the team that built the original generation of illustrated computer adventure games (I got to work with Arthur C. Clarke and Ray Bradbury, which is a great story). We discovered early on that the parser was magical but not nearly as powerful as people hoped.
Sounds a bit like LLM and ChatGPT, forty years later.
The solution was to offer a convenient and simple approach, which is almost always the solution to a problem of confusion.
We created the WordWindow button. The gratuitous trademark symbol made it more powerful, apparently.
When you clicked that button, it gave you a list of the 25 most common or useful things you could type.
I think this is going to be a powerful bridge even now. For example, a “Summarize” button is going to lean into ChatGPT’s strengths, but (Read more…)