All of Jan_Kulveit's Comments + Replies

I mostly agree with the problem statement.

With the proposed solution of giving people feedback - I've historically proposed this on various occasions, and from what I have heard, one reason for not giving feedback on the side of organizations is something like "feedback opens up space for complaints, drama on social media, or even litigation". The problem looks very different from the side of the org: when evaluating hundreds of applications, it is basically certain some errors are made, some credentials misunderstood, experiences not counted as they shou... (read more)

I would guess the 'typical young researcher fallacy' also applies to Hinton  - my impression is he is  basically advising his past self, similarly to Toby. As a consequence,  the advice is likely  sensible for people-much-like-past-Hinton, but  not a good general advice for everyone.

In  ~3 years most people are able to re-train their intuitions a lot (which is part of the point!). This seems particularly dangerous in cases where expertise in the thing you are actually interested in does not exist, but expertise in something so... (read more)

Let's start with the third caveat: maybe the real crux is what we think are the best outputs;  what I consider some of the best outputs by young researchers of AI alignment is easier to point at via examples - so it's e.g. the mesa-optimizers paper or multiple LW posts by John Wentworth.  As far as I can tell, none of these seems to be following the proposed 'formula for successful early-career research'. 

My impression is PhD students in AI in Berkeley need to optimise, and actually optimise a lot for success in an established field (ML/AI),... (read more)

It's good to see a new enthusiastic team  working on this! My impression, based on working on the problem ~2 years ago is this has good chances to provide value in global health a poverty, animal suffering, or parts of meta- cause areas; in case of x-risk focused projects, something like a 'project platform' seems almost purely bottlenecked by vetting. In the current proposal this seems to mostly depend on "Evaluation Commission"->  as a result,  the most important part for x-risk projects seems judgement of members of this commission and/or it's ability to seek external vetting

In my view this text should come with multiple caveats.

- Beware 'typical young researcher fallacy'. Young researchers are very diverse, and while some of them will benefit from the advice, some of them will not. I do not  believe there is a general 'formula for successful early-career research'. Different people have different styles of doing research, and even different metrics for  what 'successful research' means. While certainly many people would benefit from the advice 'your ideas are bad', some young researchers actually have great ideas, s... (read more)

Contrary to what seems an implicit premise of this post,  my impression is 

- most EA group organizers  should have this as a side-project, and should not think about "community building" as about their "career path" where they could possibly continue to do it in a company like Salesforce
- the label "community building" is unfortunate for what most of the EA group organizing work should consist of
- most of the tasks in "EA community building" involve skills which are pretty universal a generally useable in most other fields, like "strategizin... (read more)

1.

For different take on very similar topic check  this discussion between me and Ben Pace  (my reasoning was  based on the same Sinatra paper).

For practical purposes, in case of scientists, one of my conclusions was

Translating into the language of digging for gold, the prospectors differ in their speed and ability to extract gold from the deposits (Q). The gold in the deposits actually is randomly distributed. To extract exceptional value, you have to have both high Q and be very lucky. What is encouraging in selecting the talent is the Q se

First EuroSPARC was in 2016. Targeting 16-19 year olds, my prior would be participants should still mostly study, and not work full-time on EA, or only exceptionally.

Long feedback loops are certainly a disadvantage.

Also in the meantime ESPR underwent various changes and actually is not optimising for something like "conversion rate to an EA attractor state".

Quick reaction:

I. I did spent a considerable amount of time thinking about prioritisation (broadly understood)

My experience so far is

• some of the foundations / low hanging sensible fruits were discovered
• when moving beyond that, I often run into questions which are some sort of "crucial consideration" for prioritisation research, but the research/understanding is often just not there.
• often work on these "gaps" seems more interesting and tractable than trying to do some sort of "lets try to ignore this gap and move on" move

f... (read more)

I posted a short version of this, but I think people found it unhelpful, so I'm trying to post somewhat longer version.

• I have seen some number of papers and talks broadly in the genre of "academic economy"
• My intuition based on that is, often they seem to consist of projecting complex reality into a space of single-digit real number dimensions and a bunch of differential equations
• The culture of the field often signals solving the equations is profound/important, and the how you do the projection "world -> 10d" is less interestin

I'm not sure you've read my posts on this topic? (1,2)

In the language used there, I don't think the groups you propose would help people overcome the minimum recommended resources, but are at the risk of creating the appearance some criteria vaguely in that direction are met.

• e.g., in my view, the founding group must have a deep understanding of effective altruism, and, essentially, the ability to go through the whole effective altruism prioritization framework, taking into account local specifics to reach conclusions valid at their region.

FWIW the Why not to rush to translate effective altruism into other languages post was quite influential but is often wrong / misleading / advocating some very strong prior on inaction, in my opinion

I don't think this is actually neglected

• in my view, bringing effective altruism into new countries/cultures is in initial phases best understood as a strategy/prioritisation research, not as "community building"
• importance of this increases with increasing distance (cultural / economic / geographical / ...) from places like Oxford or Bay

(more on the topic here)

• I doubt the people who are plausibly good founders would actually benefit from such groups, and even less from some vague coordination due to facebook groups
• actually I think on the marg

• I don't think it's reasonable to think about FHI DPhil scholarships and even less so RSP as a mainly a funding program. (maybe ~15% of the impact comes from the funding)
• If I understand the funding landscape correctly, both EA funds and LTFF are potentially able to fund single-digit number of PhDs. Actually has someone approached these funders with a request like "I want to work on safety with Marcus Hutter, and the only thing preventing me is funding"? Maybe I'm too optimistic, but I would expect such requests to have decent chance of success.

Sure

a)

For example, CAIS and something like "classical superintelligence in a box picture" disagree a lot on the surface level. However, if you look deeper, you will find many similar problems. Simple to explain example: problem of manipulating the operator - which has (in my view) some "hard core" involving both math and philosophy, where you want the AI to somehow communicate with humans in a way which at the same time allows a) the human to learn from the AI if the AI knows something about the world b) the operator's values are ... (read more)

I think the picture is somewhat correct, and we surprisingly should not be too concerned about the dynamic.

My model for this is:

1) there are some hard and somewhat nebulous problems "in the world"

2) people try to formalize them using various intuitions/framings/kinds of math; also using some "very deep priors"

3) the resulting agendas look at the surface level extremely different, and create the impression you have

but actually

4) if you understand multiple agendas deep enough, you get a sense

• how they are sometimes "reflecting" t

Re: future of the program & ecosystem influences.

What bad things will happen if the program is just closed

• for the area overlapping with something "community building-is", CBG will become the sole source of funding, as meta-fund does not fund that. I think at least historically CBG had some problematic influence on global development of effective altruism not because of the direct impact of funding, but because of putting money behind some specific set of advice/evaluation criteria. (To clarify what I mean: I would expect the space would be he

As a side-note: In case of the Bay area, I'd expect some funding-displacement effects. BERI grant-making is strongly correlated with geography and historically BERI funded some things which could be classified as community building. LTFF is also somewhat Bay-centric, and also there seem to be some LTFF grants which could be hypothetically funded by several orgs. Also some things were likely funded informally by local philantrophists.

To make the model more realistic one should note

• there is some underlying distribution of "worthy things to fund&quo

meta: I considered commenting, but instead I'm just flagging that I find it somewhat hard to have an open discussion about the EA hotel on the EA forum in the fundraising context. The feeling part is

• there is a lot of emotional investment in EA hotel,
• it seems if the hotel runs out of runway, for some people it could mean basically loosing their home.

Overall my impression is posting critical comments would be somewhat antisocial, posting just positives or endorsements is against good epistemics, so the personally safest thing to do for many is not to s... (read more)

In practice, it's almost never the inly option - e.g. CZEA was able to find some private funding even before CBG existed; several other groups were at least partially professional before CBG. In general it's more like it's better if national-level groups are funded from EA

The reason may be somewhat simple: most AI alignment researchers do not participate (post or comment) on LW/AF or participate only a little. For more understanding why, check this post of Wei Dai and the discussion under it.

(Also: if you follow just LW, your understanding of the field of AI safety is likely somewhat distorted)

With hypothesis 4.&5. I expect at least Oli to have strong bias of being more enthusiastic in funding people who like to interact with LW (all other research qualities being equal), so I'm pretty sure it's not the case

2.... (read more)

In my experience teaching rationality is more tricky than the reference class education, and is an area which is kind of hard to communicate to non-specialists. One of the main reasons seems to be many people have somewhat illusory idea how much they understand the problem.

I've suggested something similar for happiness (https://www.lesswrong.com/posts/7Kv5cik4JWoayHYPD/nonlinear-perception-of-happiness ). If you don't want to introduce the weird asymmetry where negative counts and positive not, what you get out of that could be somewhat surprising - it possibly recovers more "common folk" altruism where helping people who are already quite well off could be good, and if you allow more speculative views on the space on mind-states, you are at risk of recovering something closely resembling some sort of "buddhist utilitarian calculus".

As humans, we are quite sensitive to signs of social approval and disapproval, and we have some 'elephant in the brain' motivation to seek social approval. This can sometimes mess up with epistemics.

The karma represents something like sentiment of people voting on a particular comment, weighted in a particular way. For me, this often did not seemed to be a signal adding any new information - when following the forum closely, usually I would have been able to predict what will get downvoted or upvoted.

What seemed problematic to me was 1. a numbe... (read more)

It's not an instance of complain, but take it as a datapoint: I've switched off the karma display on all comments and my experience improved. The karma system tends to mess up with my S1 processing.

It seems plausible karma is causing harm in some hard to perceive ways. (One specific way is by people updating on karma pattern mistaking them for some voice of the community / ea movement / ... )

If people think this is generally good idea I would guess CZEA can make it running in few weeks. Most of the work likely comes from curating the content, not from setting up the service

To clarify - agree with the benefits of splitting the discussion threads for readability, but I was unenthusiastic about the motivation be voting.

I don't think karma/voting system should be given that much attention or should be used as a highly visible feedback on project funding.

I don't think anyone should be trying to persuade IMO participants to join the EA community, and I also don't think giving them "much more directly EA content" is a good idea.

I would prefer Math Olympiad winners to think about long-term, think better, and think independently, than to "join the EA community". HPMoR seems ok because it is not a book trying to convince you to join a community, but mostly a book about ways how to think, and a good read.

(If they readers eventually become EAs after reasoning independently, it'... (read more)

I don't think risk of this type is given too much weight now. In my model, considerations like this got at some point in the past rounded of to some over-simplified meme like "do not start projects, they fail and it is dangerous". This is wrong and led to some counterfactual value getting lost.

This was to some extent reaction to the previous mood, which was more like "bring in new people; seed groups; start projects; grow everything". Which was also problematic.

In my view we are looking at something like pendulum swings, where we ... (read more)

Discussing specific examples seems very tricky - I can probably come up with a list of maybe 10 projects or actions which come with large downside/risks, but I would expect listing them would not be that useful and can cause controversy.

Few hypothetical examples

• influencing mayor international regulatory organisation in a way leading to creating some sort of "AI safety certification" in a situation where we don’t have the basic research yet, creating false sense of security/fake sense of understanding
• creating a highly distorted version of effecti

My impression is you have in mind something different than what was intended in the proposal.

What I imagined was 'priming' the argument-mappers with prompts like

• Imagine this projects fails. How?
• Imagine this project works, but has some unintended bad consequences. What they are?
• What would be a strong reason not to associate this project with the EA movement?

(and the opposites). When writing their texts the two people would be communicating and looking at the arguments from both sides.

The hope is this would produce more complete argument map. One... (read more)

My impression was based mostly on our conversations several months ago - quoting the notes from that time

lot of the discussion and debate derives from differing assumptions held by the participants regarding the potential for bad/risky projects: Benjamin/Brendon generally point out the lack of data/signal in this area and believe launching an open project platform could provide data to reduce uncertainty, whereas Jan is more conservative and prioritizes creating a rigorous curation and evaluation system for new projects.

I think it is fair to say you expect... (read more)

Summary impressions so far: object-level

• It seems many would much prefer expediency in median project cases to robustness and safety in rare low frequency possibly large negative impact cases. I do not think this is the right approach, when the intention is also to evaluate long-term oriented, x-risk, meta-, cause-X, or highly ambitious projects.
• I'm afraid there is some confusion about project failure modes. I'm more worried about projects which would be successful in having a team, working successfully in some sense, changing the world, but achie

Thanks Sundanshu! Sorry for not replying sooner, I was a bit overwhelmed by some of the negative feedback in the comments.

I don't think step 1b. has the same bottleneck as current grant evaluator face, because it is less dependent on good judgement.

With your proposal, I think part of it may work, I would be worried about other parts. With step 2b I would fear nobody would feel responsible for producing the content.

With 3a or any automatic steps like that, what does that lack is some sort of (reasonably) trusted expert judgement. In my view this is ac... (read more)

FWIW, part of my motivation for the design, was

1. there may be projects, mostly in long-term, x-risk, meta- and outreach spaces, which are very negative, but not in an obvious way

2. there may be ideas, mostly in long-term and x-risk, which are infohazard

The problem with 1. is most of the EV can be caused by just one project, with large negative impact, where the downside is not obvious to notice.

It seems to me standard startup thinking does not apply here, because startups generally can not go way bellow zero.

I also do not trust arbitrary set of forum us... (read more)

It is possible my reading of your post somewhat blended with some other parts of the discussion, which are in my opinion quite uncharitable reading of the proposal. Sorry for that.

Actually from the list, I talked about it and shared the draft with people working on EA grants, EA funds, and Brendon, and historically I had some interactions with BERI. What I learned is people have different priors over existence of bad projects, ratio of good projects, number of projects which should or should not get funded. Also opinions of some of the funders are at odds ... (read more)

You are missing one major category here: projects which are simply bad because they do have approximately zero impact, but aren't particularly risky. I think this category is the largest of the the four.

I agree that's likely. Please take the first paragraphs more as motivation than precise description of the categories.

Which projects have a chance of working and which don't is often pretty clear to people who have experience evaluating projects quite quickly (which is why Oli suggested 15min for the initial investigation above).

I think we a... (read more)

I'm not sure if we agree or disagree, possibly we partially agree, partially disagree. In case of negative feedback, I think as a funder, you are in greater risk of people over-updating in the direction "I should stop trying".

I agree friends and social neighbourhood may be too positive (that's why the proposed initial reviews are anonymous, and one of the reviewers is supposed to be negative).

When funders give general opinions on what should or should not get started or how you value or not value things, again, I think you are at greater risk of having too much of an influence on the community. I do not believe the knowledge of the funders is strictly better than the knowledge of grant applicants.

On a meta-level

I'm happy to update the proposal to reflect some of the sentiments. Openly, I find some of them quite strange - e.g. it seems, coalescing the steps into one paragraph and assuming all the results (reviews, discussion, "authoritative" summary of the discussion) will just happen may make it look more flexible. Ok, why not.

Also it seems you and Oli seem to be worried that I want to recruit people who are currently not doing some high-impact direct work ... instead of just asking a couple of people around me, which would often mea... (read more)

It is very easy to replace this stage with e.g. just two reviews.

Some of the arguments for the contradictory version

• the point of this stage is not to produce EV estimate, but to map the space of costs, benefits, and considerations
• it is easier to be biased in a defined way than unbiased
• it removes part of the problem with social incentives

Some arguments against it are

• such adversarial setups for truth-seeking are uncommon outside of judicial process
• it may contribute to unnecessary polarization
• the splitting may feel unnatural

I don't see why continuous coordination of a team of about 6 people on slack would be very rigid, or why people would have very narrow responsibilities.

For the panel, having some defined meeting and evaluating several projects at once seems time and energy conserving, especially when compared to the same set of people watching the forum often, being manipulated by karma, being in a way forced to reply to many bad comments, etc.

On the contrary: on slack, it is relatively easy to see the upper bound of attention spent. On the forum, you should look not on just the time spent to write comments, but also on the time and attention of people not posting. I would be quite interested how much time for example CEA+FHI+GPI employees spend reading the forum, in aggregate (I guess you can technically count this.)

I don't understand why you assume the proposal is intended as something very rigid, where e.g. if we find the proposed project is hard to understand, nobody would ask for clarification, or why you assume the 2-5h is some dogma. The back-and-forth exchange could also add to 2-5h.

With assigning two evaluators to each project you are just assuming the evaluators would have no say in what to work on, which is nowhere in the proposal.

Sorry but can you for a moment imagine also some good interpretation of the proposed schema, instead of just weak-manning every other paragraph?

I would be curious about you model why the open discussion we currently have does not work well - like here, where user nonzerosum proposed a project, the post was heavily downvoted (at some point to negative karma) without substantial discussion of the problems. I don't think the fact that I read the post after three days and wrote some basic critical argument is a good evidence for an individual reviewer and a board is much less likely to notice problems with a proposal than a broad discussion with many people contributing would.

Also when you are ma... (read more)

With the first part, I'm not sure what would you imagine as the alternative - having access to evaluators google drive so you can count how much time they spent writing? The time estimate is something like an estimate how much it can take for volunteer evaluators - if all you need is in the order of 5m you are either really fast or not explaining your decisions.

I expect much more time of experts will be wasted in forum discussions you propose.

As I've already explained in the draft, I'm still very confused by what

An individual reviewer and a board is much less likely to notice problems with a proposal than a broad discussion with many people contributing would ...

should imply for the proposal. Do you suggest that steps 1b. 1d. 1e. are useless or harmful, and having just the forum discussion is superior?

The time of evaluators is definitely definitely definitely not free, and if you treat them as free then you end up exactly in the kind of situation that everyone is complaining about. P... (read more)

To make the discussions more useful, I'll try to briefly recapitulate parts of the discussions and conversations I had about this topic in private or via comments in the draft version. (I'm often coalescing several views into more general claim)

There seems to be some disagreement about how rigorous and structured the evaluations should be - you can imagine a scale where on one side you have just unstructured discussion on the forum, and on the opposite side you have "due diligence", multiple evaluators writing detailed reviews, panel of... (read more)