habryka

• habryka 17 Oct 2019 3:31 UTC

  2 points

  in reply to: pjeby's comment on: Strong stances

  Oh yeah, I re­mem­ber ex­per­i­ment­ing with that, though ended up run­ning into similar prob­lems as com­par­ing links for the word­press case. I re­mem­ber the ID chang­ing de­pend­ing on some kind of con­text, though I don’t re­mem­ber the ex­act thing (this code was some of the first code I wrote for the new LessWrong, so it’s been a while).

  I do think this is a pretty straight­for­wardly solv­able prob­lem, we just haven’t put much effort into it, since it hasn’t been much of a prob­lem in the past.

  Is it that LW’s RSS gen­er­a­tion does not in­clude it, or is it some other site pro­duc­ing the RSS?

  This is talk­ing about RSS im­ports, so we are con­sum­ing an RSS feed from an ex­ter­nal site, and pars­ing it into a LessWrong post. So we don’t re­ally have con­trol over what data is available.

• habryka 17 Oct 2019 0:39 UTC

  2 points

  in reply to: areiamus's comment on: Strong stances

  It’s an is­sue with LW. Since RSS doesn’t provide unique IDs for posts, we are cur­rently de­ter­min­ing whether a post is new in an RSS feed on the ba­sis of its link-field, which seems to have changed two times on Katja’s blog for some rea­son in the past two days (a bunch of word­press set­tings in­fluence this, so she likely changed some set­ting some­where).

  This definitely isn’t Katja’s fault, and we should im­prove our al­gorithm to figure out whether a post in an RSS feed has already been marked as im­ported.

• habryka 15 Oct 2019 18:14 UTC

  2 points

  in reply to: Jan Kulveit's comment on: Min­i­miza­tion of pre­dic­tion er­ror as a foun­da­tion for hu­man val­ues in AI alignment

  (My sense was that Abram en­gaged pretty ex­ten­sively with the post, though I can’t fully judge since I’ve his­tor­i­cally bounced off of a lot of the pre­dic­tive pro­cess­ing stuff, in­clud­ing your post)

• habryka 14 Oct 2019 18:10 UTC

  7 points

  on: Book sum­mary: Un­lock­ing the Emo­tional Brain

  Pro­moted to cu­rated: Will write a cu­ra­tion no­tice later in the day, or maybe to­mor­row, but it seemed good to cu­rate this close to when it was pub­lished.

• habryka 12 Oct 2019 4:01 UTC

  2 points

  in reply to: Roaman's comment on: Roa­man’s Shortform

  We ended up fo­cus­ing on some other things in the past quar­ter, but we are about to plan the com­ing quar­ter, so we might end up pri­ori­tiz­ing this again.

• habryka 12 Oct 2019 1:32 UTC

  5 points

  in reply to: An_Amazing_Login's comment on: Open & Wel­come Thread—Oc­to­ber 2019

  Ad­ding your­self to the map is the first thing I would do, and then most likely I would see whether there have been any mee­tups his­tor­i­cally any­where close to your area. It ap­pears there is a meetup in 10 days in Copen­hagen, which seems pretty close to Malmo.

  https://​​less­wrong.com/​​events/​​A7LAzFXD79pgFtsdA/​​cph-meetup-10-10-19

  At that meetup you might also be able to figure out whether there are any peo­ple di­rectly from your area in­ter­ested in a meetup.

• habryka 11 Oct 2019 21:47 UTC

  8 points

  on: When we sub­stan­tially mod­ify an old post should we edit di­rectly or post a ver­sion 2?

  On the back­end, we im­ple­mented a re­vi­sion sys­tem for posts that al­lows us to mark ed­its to posts as “ma­jor re­vi­sions” which marks all the com­ments on that post as out­dated, and also al­lows users to view past re­vi­sions of the post. As an ex­am­ple, see this post.

  I hope we can soon get around to build­ing some more user-fac­ing UI for this, which al­lows users to edit a post while mark­ing the re­vi­sion as ma­jor, prob­a­bly caus­ing it to reap­pear on the front­page.

• habryka 11 Oct 2019 21:44 UTC

  3 points

  in reply to: Richard Meadows's comment on: Open & Wel­come Thread—Oc­to­ber 2019

  We already im­ple­mented this!

  When we set up cross­post­ing we can set a flag on whether to have the canon­i­cal URL point to­wards its origi­nal source (this doesn’t always make sense, for ex­am­ple for things like the AI Align­ment Newslet­ter), but if you want to au­to­mat­i­cally cross­post while pre­serv­ing the canon­i­cal URL we can set that up for you.

• habryka 10 Oct 2019 5:12 UTC

  5 points

  in reply to: gilch's comment on: Cat­e­gories: mod­els of models

  It’s to­tally pos­si­ble. In the post-ed­i­tor, just se­lect some empty space and press the “image” but­ton in the toolbar. Or use mark­down syn­tax.

  Happy to edit the above post to have images in the post, as op­posed to just links to images.

• habryka 10 Oct 2019 1:42 UTC

  4 points

  on: Reflec­tions on Premium Poker Tools: Part 1 - My journey

  (I pro­moted this to front­page, though I usu­ally leave posts like this on per­sonal blog, be­cause it does ac­tu­ally cover a bunch of gen­eral les­sons and I think is ac­tu­ally of pretty broad in­ter­est)

Long Term Fu­ture Fund ap­pli­ca­tion is clos­ing this Fri­day (Oc­to­ber 11th)

• habryka 7 Oct 2019 4:29 UTC

  7 points

  in reply to: Wei_Dai's comment on: Fea­ture Wish List for LessWrong

  Yeah, I ba­si­cally want this to re­place the ToC as you scroll past the post and into the com­ment sec­tion.

• habryka 5 Oct 2019 17:28 UTC

  5 points

  in reply to: Zack_M_Davis's comment on: How do I view my replies?

  This is cor­rect.

  We did not port over all of your old no­tifi­ca­tions, and from look­ing through Doc­u­ment’s posts, it seems like they shouldn’t have re­ceived any com­ment no­tifi­ca­tions since the re­launch, so my guess is they are look­ing for all the his­tor­i­cal no­tifi­ca­tions, which I will sadly have to ad­mit are no longer ac­cessible in the same way (though im­port­ing them is definitely a pos­si­bil­ity if enough peo­ple care, it just wasn’t on our list of pri­ori­ties and we’ve never got­ten any re­quest for them so far).

AI Align­ment Open Thread Oc­to­ber 2019

Long-Term Fu­ture Fund: Au­gust 2019 grant recommendations

Sur­vival and Flour­ish­ing Fund Ap­pli­ca­tions clos­ing in 3 days

• habryka 2 Oct 2019 0:08 UTC

  2 points

  in reply to: gwern's comment on: New Fea­ture: Link Previews

  Yeah, this and a bunch of other rea­sons caused us to mostly make the call against link-col­or­ing. We are cur­rently go­ing with an­no­ta­tions, though I would want to make them a lot smaller than they cur­rently are on gw­ern.net (I think they are fine on gw­ern.net for the kind of con­tent that you pro­duced, but would be too dis­tract­ing for LW con­tent).

• habryka 1 Oct 2019 23:54 UTC

  2 points

  in reply to: riceissa's comment on: What fund­ing sources ex­ist for tech­ni­cal AI safety re­search?

  In­deed, thanks! Added them to the list above

• habryka 1 Oct 2019 19:55 UTC

  2 points

  in reply to: Wei_Dai's comment on: Fea­ture Wish List for LessWrong

  We don’t re­ally con­trol the LW Wiki, so not sure how fea­si­ble this is. It would prob­a­bly re­quire a com­pletely sep­a­rate code­path, and I don’t know what the sub­scrip­tion in­fras­truc­ture of the wiki has. I will look into it, but can’t promise any­thing as of yet.

• habryka 1 Oct 2019 19:08 UTC

  27 points

  on: What fund­ing sources ex­ist for tech­ni­cal AI safety re­search?

  Some op­por­tu­ni­ties that I know about:

  • The Long Term Fu­ture Fund is I think cur­rently the most ac­tive granter in the space for in­de­pen­dent researchers

  • MIRI has in the past funded some in­de­pen­dent re­searchers, and some pro­grams that try to fa­cil­i­tate re­search (like fund­ing to MIRIx work­shops, or the AI Safety camp, or re­lated pro­jects)

  • Jaan Tal­linn has some bud­get set aside to fund peo­ple do­ing in­ter­est­ing re­search, I think par­tially ad­ministered via Alex Zhu. Happy to put any­one in­ter­ested into con­tact.

  • There ex­ist a va­ri­ety of small donors who I think are open to this. Matt Wage (also on the LTFF) is of­ten open to things like this.

  • The Open Philan­thropy Pro­ject has (had?) a schol­ar­ship pro­gram, here are some ex­am­ples of peo­ple they funded, I do not know whether there is any way to ap­ply to it:

    • https://​​www.open­philan­thropy.org/​​fo­cus/​​global-catas­trophic-risks/​​schol­ar­ship-sup­port-2019

    • https://​​www.open­philan­thropy.org/​​fo­cus/​​global-catas­trophic-risks/​​po­ten­tial-risks-ad­vanced-ar­tifi­cial-in­tel­li­gence/​​ai-schol­ar­ships-2018

  • The Open Philan­thropy Pro­ject also has an AI Policy ca­reers schol­ar­ship pro­gram:

    • https://​​www.open­philan­thropy.org/​​fo­cus/​​global-catas­trophic-risks/​​po­ten­tial-risks-ad­vanced-ar­tifi­cial-in­tel­li­gence/​​fund­ing-AI-policy-careers

  • The EA Foun­da­tion has given grants for AI Align­ment Re­search:

    • https://​​ea-foun­da­tion.org/​​grant­mak­ing/​​

  • Be­ing a stu­dent at CHAI is one way of get­ting fund­ing. They might also be open to more in­de­pen­dent re­search ar­range­ments, but not sure:

    • https://​​hu­man­com­pat­i­ble.ai/​​

  • The FHI re­search schol­ars pro­gram in­cludes peo­ple who work on tech­ni­cal AI Align­ment:

    • https://​​www.fhi.ox.ac.uk/​​rsp/​​

  • BERI has his­tor­i­cally done some grant­mak­ing in this space, though that will likely now be done via the Sur­vival and Flour­ish­ing Fund. For that, ap­pli­ca­tions are open for 4 more days, though re­quire some kind of ex­ist­ing in­sti­tu­tional af­fili­a­tion.

  • I can imag­ine that Tyler Cowen’s Emer­gent Ven­tures is open to fund­ing things in this space

  Will edit this if I come up with any­thing else, or if some­one else points out some­thing that I am miss­ing.

  Edit: New ad­di­tions by Issa Rice:

  • Paul Chris­ti­ano has done some in­de­pen­dent fund­ing in this space (and maybe Open AI is also open to that, but I don’t know)

  • FLI has made grants, though I don’t know about the cur­rent state of their grant program

  • EA Grants also has given some grants in this space