Thank for clarifying ‘the similar wins’ point. You seem to imply that these coaching/software/ops support/etc. wins compound on each other.
On the shared Asana space, I’ll keep checking in with the EA Netherlands/Rethink/CE coaches working with EA groups/charity start-ups on how time-(in)efficient/(in)convenient it is to keep track of team tasks with the leaders they are mentoring.
From my limited experience, a shared coaching GDoc already works reasonably well for that:
Upside: Everyone uses GDoc. Easy to co-edit texts + comment-assign questions and tasks that pop up in email inbox. On the other hand, the attentional burden of one party switching over to the other’s task management system to track say biweekly check-ins over half a year doesn’t seem worth it.
Downsides: GDocs easily suck away the first ten minutes of a call when you need to update each other on two weeks of progress in one swoop. It also relies on the leader/coach actively reminding each other to check medium-term outcomes and key results. This ‘update/remind factor’ felt like a demotivating drag for me in my coach or accountability check-ins – all with people who I didn’t see day to day and therefore lacked a shared context with.
The way you arrange the format together seems key here. Also, you’d want to be careful about sharing internal data – for Asana, I recommend leaders to invite coaches comment-only to projects, rather than entire teams.
On other software or services, curious if any ‘done deals’ come to mind for you.
Regarding your forecasting platform, I’m curious if anything comes to mind on fitting forecasts there with EA project planning over the next years.
First off, I really appreciate the straightshooter conclusion of ‘QC is unlikely to be helpful to address current bottlenecks in AI alignment.’ even while you both spent many hours looking into it.
Second, I’m curious to hear any thoughts on the amateur speculation I threw at Pablo in a chat at the last AI Safety Camp:
Would quantum computing afford the mechanisms for improved prediction of the actions that correlated agents would decide on?
As a toy model, I’m imagining hundreds of almost-homogenous reinforcement learning agents within a narrow distribution of slightly divergent maps of the state space, probability weightings/policies, and environmental inputs. Would current quantum computing techniques, assuming the hardware to run them on is available, be able to more quickly/precisely derive the % portions of those agents at say State1 would take Action1, Action2, or Action3?
I have a broad vague sense that if that set-up works out, you could leverage that to create a ‘regulator agent’ for monitoring some ‘multi-agent system’ composed of quasi-homogenous autonomous ‘selfish agents’ (e.g. each negotiating on behalf of their respective human interest group) that has a meaningful influence on our physical environment. This regulator would interface directly with a few of the selfish agents. If that selfish agent subset are about to select Action1, it will predict what % of other, slightly divergent algorithms would also decide Action1. If the regulator prognoses that an excessive number of Action1s will be taken – leading to reduced rewards to or robustness of the collective (e.g. Tragedy of the Commons case of overutilisation of local resources) – it would override that decision by commanding a compensating number of the agents to instead select the collectively-conservative Action2.
That’s a lot of jargon, half of which I feel I have little clue about… But curious to read any arguments you have on how this would (not) work.