riceissa / aiwatch Goto Github PK

Currently, the documents on agenda pages such as https://aiwatch.issarice.com/?agenda=Iterated+amplification#documents don't show a column for affected agendas. I believe the reason for not having this column was that it's already obvious that all the documents are about this agenda.

However, many documents have multiple affected agendas. so it would be nice to have the affected agendas column.

Examples include:

• Where does the organization fall on the spectrum between fully in-office and fully remote work?
• What is the organization's position on pay scales (some orgs like Pangea have equal pay for all workers (cf. https://www.pangealegal.org/jobs); other have multiple pay grades)?
• What is the organization's reporting structure (flat, consensus-based, holacracy, almost-flat with everybody reporting to CEO, or a full complex reporting tree?)

These parameters can change over time, so it would be good to record start and end dates for specific values of the policy parameters.

I need to discuss this more with @riceissa and @sebastiansanchez123.

Please ask @riceissa for guidance on their nontrivial org structure!

Basically, use the existing data already collected in OW about employee counts to generate graphs, DLW-style.

Also inspirational: https://issarice.com/givewell-staff-growth#plots

I'll leave the specification of what the charts should show to @riceissa, but hopefully the above give some ideas.

For instance, for https://orgwatch.issarice.com/?organization=Open+Philanthropy+Project there should be a table of contents on top similar to that for https://donations.vipulnaik.com/donor.php?donor=Open+Philanthropy+Project

Similarly, there should be a table of contents for person pages, though right now there doesn't seem to be enough content on the person pages to justify a ToC.

https://www.facebook.com/brianfinifter/posts/10102058260531588?comment_id=10102058311110228

The ideas include:

• A view for intersection of two organizations. This would show:
  • The people who have been involved with both orgs (not necessarily simultaneously)
  • The documents that involve both orgs, either as publisher or as affected organization
• A view for things common between two people. This would show:
  • The organizations that both of them have been involved with (not necessarily simultaneously)
  • The documents that involve both people, either as author or as affected person

e.g. at https://orgwatch.issarice.com/?organization=Centre%20for%20Effective%20Altruism#basic-information

https://aiwatch.issarice.com/#positions-grouped-by-organization

e.g. the OpenAI row ends in "Rachel Fong, Raf" (the "Raf" is the start of a name that gets cut off) because the group_concat length limit is reached.

It's probably not a good idea to just increase the length limit, because then the table will just get really large and messy. So the idea is to think about this some more to present the table in the most useful way. Maybe restrict to N people or filter to some interesting subset for each org.

https://www.facebook.com/anna.salamon.96/posts/10158082768717682

https://www.lesswrong.com/posts/dJQo7xPn4TyGnKgeC/hiring-engineers-and-researchers-to-help-align-gpt-3

• Sören Mindermann https://arxiv.org/abs/1712.05812
• Alex Mennen, intern at MIRI https://www.facebook.com/alexmennen
• Nicolas Gagné, intern at MIRI April 2014 to August 2014 https://www.facebook.com/nicolas.gagne.9/about?lst=100008068993970%3A514911510%3A1531639016&section=education
• Alex Appel = Diffractor, "related essay by Alex Appel" https://agentfoundations.org/item?id=1816
• the people in https://www.general-ai-challenge.org/solving-the-ai-race-results
• https://arxiv.org/abs/1806.04067
• https://www.openphilanthropy.org/focus/global-catastrophic-risks/potential-risks-advanced-artificial-intelligence/ai-scholarships-2018
• new people on MIRI team page
• https://www.linkedin.com/in/michael-page-71318613/ "Policy and Ethics Advisor Company Name OpenAI Dates Employed 2017 – Present Employment Duration 1 yr Location San Francisco Bay Area"
• Daniel Braun https://www.greaterwrong.com/posts/bXLi3n2jrfqRwoSTH/human-aligned-summer-camp-a-summary#section-6
• Luke Raskopf http://rationality.org/about/team (added around week of August 12)
• Johannes Treutlein http://effective-altruism.com/ea/1rz/request_for_input_on_multiversewide/ "I am currently working on a research project as part of CEA’s summer research fellowship."
• Michaël Trazzi https://arxiv.org/abs/1808.03644
• Jeremy Nixon? https://www.linkedin.com/in/jeremyvnixon/
• Christopher Olah, from Google Brain to OpenAI https://www.facebook.com/100003775051002/posts/1402487849887037/
• Vishal Maini, https://www.greaterwrong.com/posts/Y8ezvHWZXW232CkKr/link-new-deepmind-ai-safety-research-blog
• [OpenAI; update https://github.com/riceissa/aiwatch/blob/master/sql/positions/openai-positions.sql after @riceissa gives repo permissions] Daniel Ziegler https://80000hours.org/podcast/episodes/olsson-and-ziegler-ml-engineering-and-safety/
• https://arxiv.org/abs/1811.07871
• https://futureoflife.org/ai-safety-research/
• "In the Christmas period just gone, Peter McIntyre visited Oxford and the San Francisco Bay Area for a month each and interned at CFAR in the latter." https://eaforum.issarice.com/posts/EpQqh6wCDtfZkeneg/should-you-visit-an-ea-hub
• some of the grantees for the long-term future fund
• "volunteer Zack Davis" https://www.lesswrong.com/posts/iLGrNTwTZivTX5774/less-wrong-q-and-a-with-eliezer-yudkowsky-ask-your-questions#h76Luakpa2JeWmsBJ
• "March 2015 [...] I become a MIRI contractor instead." https://www.greaterwrong.com/posts/iBBK4j6RWC7znEiDv/history-of-the-development-of-logical-induction

I'm thinking of a table with columns (this could be OrgWatch-only):

• Organization (e.g., GiveWell, Animal Charity Evaluators)
• Tool (e.g., Slack, Google Docs, Drupal, Toggl)
• Tool type (e.g., "Content management system" "Real-time communication platform") that is an enum. The tool type should be a function of tool (alternatively the tool -> tool type mapping can be a separate table)
• Start and end time of use of tool (to the extent known)
• Notes on how the organization uses the tool; for instance, for organization = GiveWell, tool = Slack, stuff like what's at https://denver-frederick.com/2019/05/04/elie-hassenfeld-co-founder-and-executive-director-of-givewell-joins-denver-frederick/ "We use Slack as our internal chat program, and we have a channel for mistakes."

@riceissa and @sebastiansanchez123, what do you guys think?

e.g. i think https://aiwatch.issarice.com/?person=Issa+Rice used to show some stuff but now it just 404s since my recent update.

https://eaforum.issarice.com/posts/cGBHjeyHJoooJCjwQ/working-at-ea-organizations-series-effective-altruism

I'm not sure if this is just a Firefox bug. (I don't notice it on sites like Wikipedia, although I notice it on other sites.)

Basically, the equivalent of https://issarice.com/givewell-executive-compensation but more broadly for more than just "executives"?

I need to brainstorm this more with @riceissa and @sebastiansanchez123.

This is the new kid on the block with AI safety (and with other GCRs too)? https://cset.georgetown.edu/

https://eaforum.issarice.com/posts/ayzEiBgnnHX7DADwD/working-at-ea-organizations-series-80000-hours

https://research.tees.ac.uk/ws/files/7669818/Modelling_and_Influencing_the_AI_Bidding_War.pdf

When I go to an AI Watch page about a person (such as Paul Christiano) or an organization (such as MIRI) I would see to see the list of agendas the person or organization is affiliated with. Currently, the display of affiliation is only in the other direction (the page on an agenda displayed affiliated persons and organizations; e.g., embedded agency).

Because the latter is an offshoot of the former, it makes sense to work on both together, I think.

https://forum.effectivealtruism.org/posts/33AnPajNYmNrdXQbj/rethink-priorities-2020-impact-and-2021-strategy

"Note: I work for MIRI, but these are entirely my own opinions and not MIRI's – I've been thinking about this problem since long before I joined MIRI and don't necessarily see eye-to-eye with most people there on this question." https://lw2.issarice.com/posts/i3FyT3Gvst8pFX7gE/location-discussion-takeaways

The table of contents on org pages such as https://orgwatch.issarice.com/?organization=OpenAI links to https://orgwatch.issarice.com/?organization=OpenAI#list-of-people but that section doesn't exist. The correct section link is https://orgwatch.issarice.com/?organization=OpenAI#list-of-people-248-positions

We can fix this in either direction (change actual link to just be https://orgwatch.issarice.com/?organization=OpenAI#list-of-people or change link in TOC to https://orgwatch.issarice.com/?organization=OpenAI#list-of-people-248-positions). I suspect the former approach is better and more consistent with how the other sections work.

Team page: https://www.generationpledge.org/team

Home page: https://www.generationpledge.org/

HT @riceissa

announced here: https://forum.effectivealtruism.org/posts/PvBLDPkqKvdHQkKPn/introducing-the-legal-priorities-project

In DLW, the headers listing documents and donations give the number of rows in the tables that follow. This is useful in giving people a rough sense of the size. It's also useful for checking, right after adding some rows, whether the count went up as expected.

I want something similar for sections such as https://orgwatch.issarice.com/?organization=Open+Philanthropy+Project#organization-documents

https://www.facebook.com/robbensinger/posts/10163981893970447

Launch blog post on EA Forum: https://forum.effectivealtruism.org/posts/vxwcxwiDKCnyHJhbz/announcing-the-new-forethought-foundation-for-global
Website: https://www.forethought.org/

https://www.1fortheworld.org/

see riceissa/project-ideas#33 for details

Basically the table's rows can correspond to postings like https://www.givewell.org/about/jobs/content-editor and include information about the job like salary, location, time range of posting, etc.

I need to think more about this and discuss further with @riceissa and @sebastiansanchez123.

Seems reasonable to add AlignmentForum.org to the products section, maybe together with LW 2.0, which has a separate team working on it than the old LW had.

You can also add LW 2.0 as full-time for me (Oliver Habryka) starting in June 2017 or so.

The query is

select person,website,urls from people where person not in (select distinct(person) from positions)

Example of where this is a problem: Sören Mindermann doesn't appear on https://aiwatch.issarice.com/ but does on https://orgwatch.issarice.com/ because of the CEA position https://aiwatch.issarice.com/?person=S%C3%B6ren+Mindermann

See https://forum.effectivealtruism.org/posts/3QufK3jjQ5aqksaJu/new-career-guidance-organization-probably-good for more!

HT @riceissa

• Open Phil
• GiveWell
• MIRI
• CFAR
• OpenAI
• BERI
• CHAI
• AI Impacts
• Animal Charity Evaluators
• Future of Humanity Institute
• PAISRI

https://www.metaculus.com/about/

• Alignment Forum https://www.alignmentforum.org/ https://www.greaterwrong.com/posts/JiMAMNAb55Qq24nES/announcing-alignmentforum-org-beta

currently only positions with a non-NULL start date are shown