remmelt

remmelt's Comments

Implications of Quantum Computing for Artificial Intelligence alignment research (ABRIDGED)

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.

What new EA project or org would you like to see created in the next 3 years?

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.


What new EA project or org would you like to see created in the next 3 years?

Good to hear your thoughts on this!

What do you mean here with a ‘portfolio of similar wins’? Any specific example of such a portfolio that comes to mind?

What new EA project or org would you like to see created in the next 3 years?

Hey, I never finished my reply to you.

First off all, I thought those 4 items are a useful list of what you referred to as infrastructure for small projects.


On offering Asana Business:

  • We are now offering Asana Business teams at 90% discounts (€120/team/month) vs. usual minimum cost. This is our cost price since we're using the a 50% Nonprofit discount, and assign one organisation member slot per team facilitator. The lower cost is a clear benefit to the organisations and groups that determine to move to Asana Business
  • I'm working with ops staff from RethinkCharity and Charity Entrepreneurship (and possibly Charity Science Health) to move to a shared Asana space called 'Teams for Effective Altruism' (along with EA Netherlands and EA Cambridge). Not set in stone but all preparations are now in place.
  • This doesn't yet answer your question of why I particularly thought of Asana. Here are some reasons for why to work on building up an shared Asana Business space together:
    • Online task management is useful: I think at least half of the EA teams >5 people running small projects would benefit from tracking their tasks online for remote check-ins. For instance, when it's hard to travel to say a meeting room once a week, or you need to reliably carry out nitty-gritty ops tasks where it feels burdensome for a manager to ask 'Have you done this and this and this?'. At EA Netherlands, a lot of the project delays and time wasted seemed to emerge along the lines of someone feeling unclear of what was expected/endorsed of their role, being aware of update X, waiting for person Y to confirm, or forgetting/having to remind about task Z. It seems to make common-sense to avoid that by creating a 'single place of truth' where team members can place requests and update each other on progress asynchronously.
    • Facilitate onboarding of teams: Leaders of small projects seem to experience difficulty in getting volunteers to building the habit of updating online tasks in the first months, even if most would agree on reflection that it's worth the switching cost. In surveying EA regional groups in northern Europe, the one reason organisers kept mentioning to me why they weren't using task software came along the lines of them previously excitedly trying to track tasks online but volunteers forgetting to update their tasks a few weeks later. Both EA Netherlands and EA Oxford flopped twice at using Trello. My sense is they would have succeeded more likely than not if someone took up the role of facilitating team members to use the platform in ways that was useful to them, and reminding them to update their tasks weeks down the line. Part of the Asana team application process is assigning a facilitator, who I can guide from within our shared space.
    • Asana Business is top-notch: I personally find Asana Business' interface intuitive and well-ordered, striking a balance between powerful features and simplicity. External reviews rate Asana around 4-4.5 out of 5. Having said that, some EA teams seem to have different work styles or preferences that fit other platforms better – I've heard of people using Trello, Nozbe, Notion, GSheets, or even just NextCloud's basic task board.
    • Asana is an unexploited Schelling point for collaboration: A surprising number of established EA organisations use Asana: the Centre for Effective Altruism, RethinkCharity, Founder's Pledge, Centre for Human-compatible AI, Charity Entrepreneurship, Charity Science Health, 80,000 Hours(?), and probably a few I haven't discovered yet. That's an implicit endorsement of Asana's usefulness for 'EA work' (bias: Dustin Moskovitz co-founded it). Asana staff are now making their way into the Enterprise market, and intend to developing features that enable users to smoothly start collaborations across increasingly large organisational units (Teams...Divisions...Organisations).
    • Passing on institutional knowledge to start-ups: In a call I had with a key Asana manager, he randomly mentioned how it would be great to enable organisations to coordinate across spaces. I don't think we have to wait for that though. When EA Hub staff could offer Asana teams to local EA groups in our shared space, coach them by commenting on projects/scheduling check-in calls, and stay up to date of what's going on. Likewise, Charity Entrepreneurship could offer Asana teams to the charities they incubate and continue checking in with and supporting the start-up leaders coming out of the incubation program. People could also share project templates (e.g. conference/retreat organiser checklists), share standardised data from custom fields, etc.
    • So of your infrastructure suggestions, that seems to cover operations support and coaching/advice.
    • To make sharing the space work, we'd have to close off short-term human error/malice failure modes as well as tend to the long-term culture we create. Downsides of connecting software up to discuss work smoothly is that it's also easier for damaging ideas and intentions to cross boundaries, for people to jostle for admin positions, and for a resulting homogenous culture is build upon fragile assumptions of how the world works, and what the systematic approaches are to improving it.


What new EA project or org would you like to see created in the next 3 years?

@Ozzie, I'm curious what kinds of infrastructures you think would be worth offering.

(I'm exploring offering Asana Business + coaching to entrepreneurs starting on projects)


What are some neglected practices that EA community builders can use to give feedback on each other's events, projects, and efforts?

I also find the idea of recording meetings interesting. I’d worry about this not working out because of bandwidth limitations – asking an overseas organiser to watch on passively for an hour and then collect their thoughts on what happened seems to ask more of them than to interact with, query, and coach in the moment.

I wonder if there are any ways to circumvent that bottleneck. Perhaps calling in the person through Zoom and letting them respond at some scheduled moment helps somewhat? Any other ideas?

Another way for giving feedback might be to give people access to your task planning. I just emailed Asana about whether they’d be willing to offer a free Business/Enterprise team for people to run projects on.

Text: “We would like to pilot one Asana Business team for community start-ups to collaborate on tasks, link with coaches and advisors, collect feedback from the groups we service, and to be more transparent to charity seed funders.”

EA Angel Group: Applications Open for Personal/Project Funding

Better description of grantmaker’s Scope: the ‘problem-skills intersections’ they focus on evaluating. Staff of funds should share these with other larger funders, and publish summaries of them on their websites.

EA Angel Group: Applications Open for Personal/Project Funding

Been messaging with Brendon and others. I thought I’d copy-paste the – hopefully – non-inflammatory / personal parts of the considerations I last wrote about last so we can continue having collaborative truth-seeking discussions on those here as well.

To Brendon

I would clearly keep stating that you’re focused on funding early start-ups in the pilot/testing stages who are working with clearly delineated minimum viable target groups.

That cuts out a bunch of a funding categories like funding AI safety researchers, funding biotech work, or funding entire established national EA groups, and I think that’s good! (actually [...], the [...] from EA Netherlands might not like me saying that...anyway)

Those are things people at EA Grants, EA Community Building Grants, EA Funds or OpenPhil (of course!) might be focused on right now.

The Community Building Grant has some definite problems in the limited time they have to assess and give feedback to national and regional EA organisers, and their restrictive career plan changes criteria. Harri from CEA and I had a productive conversation about [that] [...] But in my opinion funding by the Angel Group there should focus on specific projects for specific target groups by the organisers. I think national and local group members should play a more active role in sharing feedback on how much the organisers work has helped them come to better reflected decisions for doing good and stick to them – and offer funding to extend the organisers’ runway. Which I hope makes it clear what kind of area I see the crowdfunding platform Heroes & Friends come in.”

And in the WhatsApp group exploring that crowdfunding platform:

On specialisation between funders

@[...], I think it’s important for funding platforms and grantmakers to clearly communicate what they’re specialised in a few paragraphs.

Especially:

  • scope in terms of cause/skill intersections
  • brightspots (funding area where their batting rate is high)
  • blindspots (where they miss promising funding opportunities, i.e. false negatives)
  • traps (failure modes of how they conduct their processes)

[added later: To “traps”, I should also add failure modes that grantmakers could see other, less experienced funders running into (so a newcomer funder can plan e.g. a Skype call with the grantmaker around that)]

This is something most grantmakers in the EA community are doing a pisspoor job at right now IMO (e.g. see our earlier Messenger exchange on online communication of EA Funds).

There’s a lot of progress to be made there. I expect building consensus around funding scopes and specialisation will significantly reduce the distractions and fracturing of groups we might each add to with scaling up the Angel Group or [possibly] collaborating with Heroes & Friends.

I’ve tried to clearly delineate with you guys what EA RELEASE (for lack of a better name for now) would be about.

Regarding the Angel Group, here is the suggestion I just shared with Brendon: [...]

EA Angel Group: Applications Open for Personal/Project Funding

Thanks, that clarifies a bunch of things for me.

I realise now I was actually confused by your sentence myself.

I took

Rather than hiding opportunities from other funders like venture capitalists in the for-profit world, I believe that EA funders such as EA Grants, BERI Grants...”

to mean

“EA Grants, BERI Grants, etc. should not hide opportunities from funders like VCs from the for profit sector”.

The rest of your article can be coherently read with that interpretation. To prevent that I’d split it into shorter sentences:

“Venture capitalists in the for-profit sector hide investment opportunities from others for personal monetary gain. EA grantmakers have no such reason for hiding funding opportunities from other experienced funders. Therefore, ...

Or at the very least, make it “Rather than hiding opportunities from other funders like venture capitalists in the for-profit world DO, I believe that...”

EA Angel Group: Applications Open for Personal/Project Funding

John Maxwell wrote an analysis on your initial post on how most platform initiatives seem to fail in the EA community and that the ones that did last seemed resulted from a long stretch of consensus building (+ attentive refinement and execution in my opinion). This was useful for me to consider that more deeply as an issue in coordinating funding in the EA community. It at least led me to take smaller, tentative steps to trying things out while incorporating the advice/goals/perspectives/needs of people with deep understandings of aspects or a clear stake in using the final product.

https://forum.effectivealtruism.org/posts/io6yLz6GtF6kvXt99/ideas-for-improving-funding-for-individual-eas-ea-projects#48ReFmNG5Zf3yhwk9

Load More