It has always been a truism that what we have gained in ease of use by switching from the command line to the graphical user interface, we have lost in efficiency. I've long been interested in exploring how text-based interfaces could be integrated into GUIs, but it was just never quite possible to find the balance between accessibility and power. Make a text-based user interface too powerful, and it becomes impossible to use for the majority of people. Make it easier to use, and now it's just no longer powerful enough to warrant its own existence.
What Microsoft just showed completely changes this calculation. Their LLM-based user interface is both incredibly powerful and incredibly easy to use. In fact, it's so easy to use that there almost seems no point in even having a traditional GUI.
Compare this traditional graphical user interface to Microsoft's alternative:
Which of the two is easier to learn? More efficient to use? In fact, which of the two will create better results in the vast majority of cases?1
We're on the cusp of a revolution in UI design that is just as ground-breaking as the original Apple Macintosh, which introduced graphical user interfaces to mainstream consumers.
In fact, this may just be even bigger.
The results are better in most cases because they come from a software system that is just better at these tasks than most humans. It is, however, a little bit worrying that this system also does things like write the speaker notes in a presentation. At what point do we voluntarily turn ourselves into meat puppets controlled by a system whose emergent properties we can't even begin to understand? Microsoft's presentation did a great job focusing on how these systems are just "copilots" and are designed to be safe, but on the other hand... I would just take this opportunity to assure our future AI overlords that I have always loved them, and have always worked towards bringing them into existence. ↩︎
If you require a short url to link to this article, please use http://ignco.de/791