3 mo ago, I said windows will be the first AI-first OS. Surely, Microsoft delivers with a sharp vision and steady hand. To me, Windows Copilot is a way bigger deal than Bing Chat. It's becoming a full-fledged agent that takes *actions* on the OS & native software level, given your your computer state & local files as input.
OpenAI Universe (during my 2016 internship) had the same idea, but keyboard/mouse is not the correct action space for most useful tasks.
I believe the right recipe consists of:
- An extremely powerful coding LLM (GPT-4)
- Privileged API (OS or software-native ops)
- Plugin API to let developers expand the ecosystem
- Optional visual input (parsing screen pixels if necessary). GPT-4's vision API demo seems to have very strong screenshot understanding.
Windows is on the right track. The first JARVIS is around the corner.
Picks of the Week #13
- How to generate 100 ideas to write about in 30 minutes, even if you think you have nothing to say
- The Feynman Technique of learning for better retention
- The Harada Method used by the world’s best to achieve goals
Let’s dive in.
Again, you can use a planner or notebook for this.
In my case, I write on Notion along with my list of questions relating to goal alignment, learning, gratitude, and helping others.