I've transitioned from reading about alignment and theorizing about the problem landscape, to doing things. And by things I mean MVPs of bigger ideas, where each bigger idea roughly falls under the umbrella of Collective Human Intelligence. I'll be writing up these thoughts into a coherent research agenda. For the moment I just want to transparently share my process for this month.
Writing on the Wall
After reading a lot about alignment and somewhat despairing about finding any useful research direction, I -- just decided to write and draw what came to mind. What seemed interested, promising, fun? Brainstorming and continuous stream of thought. I've wiped away quite some of my notes now, but basically by office now looks like this:
Fun fact: People tend to enjoy adding items and drawing on the glass when they come by.Impression of what my office looks like to be transparent about process. Not saying these thoughts are good or correct.
MVPs
For each idea that I came up with, I started fleshing out an MVP -- How could I test if this idea is worthwhile and doable? Then each MVP is broken down in actionable steps, and the total time of execution is kept at 1-10 work days. Additionally, each MVP has a clear scaled-up target where the bigger thing would actually make a meaningful contribution to solving the alignment problem. This started out as a semantic-web-like blob of free association (see whiteboard picture above) and later got distilled in to a list of seven MVPs with structered steps (two of them are shown below the whiteboard).
Talk, talk, talk
I booked myself full with meetings -- With new people. Old people. All people.
I asked them if I could run my ideas past them and if they could give feedback. So I had a mathematician liking an algorithmic idea. A governance-minded person giving advice on where to find the right people to talk to. A senior alignment researcher asking what the hell I'm up to and why I'm not thinking of this other application of my expertise.
Honestly, it was uniformally great. And I'm not sure at what point I should stop. Probably the point of necessity, which is soon-ish.
Synthesis
A lot of the actual work I'm doing is synthesizing all the different ideas I've had. Ordering them. Working them out. Considering which are feasible. Which have good cross-over. And how do they even relate to each other.
That's how I found the common thread of Collective Human Intelligence -- by drawing a boundary in concept space around the cluster of seven MVPs I have listed now all my walls.
Next steps
I'm going to be writing up my project ideas into a research agenda on Collective Human Intelligence, write a funding proposal (I have about 2 months left now on my current funding), and then prioritize which MVPs I'll explore first in my remaining time.
I've transitioned from reading about alignment and theorizing about the problem landscape, to doing things. And by things I mean MVPs of bigger ideas, where each bigger idea roughly falls under the umbrella of Collective Human Intelligence. I'll be writing up these thoughts into a coherent research agenda. For the moment I just want to transparently share my process for this month.
Writing on the Wall
After reading a lot about alignment and somewhat despairing about finding any useful research direction, I -- just decided to write and draw what came to mind. What seemed interested, promising, fun? Brainstorming and continuous stream of thought. I've wiped away quite some of my notes now, but basically by office now looks like this:
MVPs
For each idea that I came up with, I started fleshing out an MVP -- How could I test if this idea is worthwhile and doable? Then each MVP is broken down in actionable steps, and the total time of execution is kept at 1-10 work days. Additionally, each MVP has a clear scaled-up target where the bigger thing would actually make a meaningful contribution to solving the alignment problem. This started out as a semantic-web-like blob of free association (see whiteboard picture above) and later got distilled in to a list of seven MVPs with structered steps (two of them are shown below the whiteboard).
Talk, talk, talk
I booked myself full with meetings -- With new people. Old people. All people.
I asked them if I could run my ideas past them and if they could give feedback. So I had a mathematician liking an algorithmic idea. A governance-minded person giving advice on where to find the right people to talk to. A senior alignment researcher asking what the hell I'm up to and why I'm not thinking of this other application of my expertise.
Honestly, it was uniformally great. And I'm not sure at what point I should stop. Probably the point of necessity, which is soon-ish.
Synthesis
A lot of the actual work I'm doing is synthesizing all the different ideas I've had. Ordering them. Working them out. Considering which are feasible. Which have good cross-over. And how do they even relate to each other.
That's how I found the common thread of Collective Human Intelligence -- by drawing a boundary in concept space around the cluster of seven MVPs I have listed now all my walls.
Next steps
I'm going to be writing up my project ideas into a research agenda on Collective Human Intelligence, write a funding proposal (I have about 2 months left now on my current funding), and then prioritize which MVPs I'll explore first in my remaining time.