Steph's notes from the actual meeting:
- "good question" -> people upvoting your question is a good way to acquire reputation
- make proposing changes an action you can take right away (before you have acquires much reputation)
- subscriptions don't cause a trust flow
- remove private groups and deploy a version of arbital for each group that wants to have a private discussion
- which domain of math? follow the author virality and the easy conceptual boundaries
- we need flagging early on
- before we can scale past 100
- find (list all the pages that use this page as a requisite) this is a pain in the ass, but makes it possible to manually merge and split
- split pages
- split/merge requisites
- move comments
- we need more author hedons
- visible xp
- only visible to you
- goes up if editors say nice things about your feedback or if your question is upvotes
- when someone views your pages, likes your page, learns a requisite from your page
- author colliding
- do we want the discussion section on math
- not having the discussion section would make us less able to handle more controversial domains
- eliezer is reluctant to bind our users - they may come up with a good way to use the discussion section
- seems useful to have discussion on answer pages
- nate
- two people he is pitching on arbital
- first guy he will ask for advice and then ask for money (other connections)
- second guy is a week later
- 4mil cap, 20% discount
Questions for Eliezer
Alexei Andreev, Stephanie Zolayvar: please add questions here. We should make sure to include UI issues that we need to talk to EY about.
Feature Questions
Q: Clarify Qs&A
- What happens to queries that go through normal search? Should we put them somewhere?
- Search strings on all pages?
Q: Steph question on Trust / Karma
Q: Steph question on Domains
Short-term Plan Questions
Q: Should we have private groups?
Q: Which math vertical first?
Discussion of Strategy
Q: Why are we starting with Explanations?
- Stepping stone to Discussion and Debate, or important in its own right?
- If lukewarm response to Explanations:
- charge ahead with Discussion?
- or iterate on Explanations until it works?
- or, is our thesis is falsified?
- Follow-up: how much time should we expect to spend polishing explanations?
- last 20% of awesomeness will require 80% of the effort?
- Double crux: is explanations a problem that many people have? Or is it a small number of people like Eliezer?
- Feeling: Expii is focusing in a way that we're not, and that will make their explanations better. Are we doing it wrong?
Q: More generally, which parts of the plan are fixed, and which are variable?
- Eric Rogstad is thinking about how he should trade off between asking himself:
- How can we achieve this subgoal of the plan?
- Can we route around this subgoal and pursue an alternate path towards a more-final goal?
- In other words, let's goal-factor Arbital!
- some goals
- FAI
- end Eternal September
- improve the right-tail / state-of-the-art / frontier of human knowledge on important topics
- enable one-to-many explanation
- cause more AI-safety researchers to exst
- influence policy-makers about AI safety
- influence policy in general
- take over the world (replace fb feed)
- make us money
- push frontiers of knowledge
Q: What is the strategic purpose of Arbital? How does it fit into the plan to save the world?
Brainstorming:
- make money
- make it easier to explain AI risk and other important considerations to
- potential researchers
- policy makers
- raise humanity's collective ability to figure things out
- is it primarily a recruiting tool?
- establish an authority on debated topics
Q: What does a discussion of which EA cause to donate to look like?
- What are user stories? Does our feature really make productive discussion possible?
- How does the eternal September problem get solved?