Cullen_OKeefe

FHI Re­port: How Will Na­tional Se­cu­rity Con­sid­er­a­tions Affect An­titrust De­ci­sions in AI? An Ex­am­i­na­tion of His­tor­i­cal Precedents

• Cullen_OKeefe 22 Jul 2020 21:58 UTC

  1 point

  on: Put­ting Peo­ple First in a Cul­ture of De­hu­man­iza­tion

  You might like this post: https://​​schwitzs­plin­ters.blogspot.com/​​2020/​​06/​​con­test-win­ner-philo­soph­i­cal-ar­gu­ment.html?m=1

• Cullen_OKeefe 22 Jul 2020 17:14 UTC

  2 points

  in reply to: rgb’s comment on: Par­allels Between AI Safety by De­bate and Ev­i­dence Law

  Thanks for this very thought­ful com­ment!

  I think it is ac­cu­rate to say that the rules of ev­i­dence have gen­er­ally aimed for truth-seek­ing per se. That is their stated goal, and it gen­er­ally ex­plains the liberal stan­dard for ad­mis­sion (rele­vance, which is a very low bar and tracks Bayesian episte­mol­ogy well), the even more liberal stan­dards for dis­cov­ery, and most of the ad­mis­si­bil­ity ex­cep­tions (which are gen­er­ally ex­plain­able by hu­mans’ im­perfect Bayesi­anism).

  You’re definitely right that the le­gal sys­tem as a whole has many goals other than truth-seek­ing. How­ever, those other goals are gen­er­ally ad­vanced through other as­pects of the jus­tice sys­tem. As an ex­am­ple, fi­nal­ity is a goal of the le­gal sys­tem, and is ad­vanced through, among other things, statutes of limi­ta­tions and re­pose. Similarly, the “be­yond rea­son­able doubt” stan­dard for crim­i­nal con­vic­tion is in some sense con­trary to truth-seek­ing but ad­vances the policy prefer­ence for un­der­pun­ish­ment over over­pun­ish­ment.

  You’re also right that there are some ex­cep­tions to this within ev­i­dence law it­self, but not many. For ex­am­ple, the at­tor­ney–client priv­ilege ex­ists not to fa­cil­i­tate truth-seek­ing, but to pro­tect the at­tor­ney–client re­la­tion­ship. Similarly, the spousal priv­ileges ex­ist to pro­tect the mar­i­tal re­la­tion­ship. (Pre­cisely be­cause such priv­ileges are con­trary to truth-seek­ing, they are in­ter­preted nar­rowly. See, e.g., United States v. Ara­mony, 88 F.3d 1369, 1389 (4th Cir. 1996); United States v. Suarez, 820 F.2d 1158, 1160 (11th Cir. 1987)). And of course, some rules of ev­i­dence have both truth-seek­ing and other policy ra­tio­nales. Still, on the whole and in gen­eral, the rules of ev­i­dence are aimed to­wards truth.

AI Benefits Post 5: Out­stand­ing Ques­tions on Govern­ing Benefits

• Cullen_OKeefe 21 Jul 2020 16:43 UTC

  11 points

  on: What skill-build­ing ac­tivi­ties have helped your per­sonal and pro­fes­sional de­vel­op­ment?

  Learn­ing the Get­ting Things Done pro­duc­tivity method im­proved my per­sonal pro­ject man­age­ment skills.

• Cullen_OKeefe 20 Jul 2020 23:07 UTC

  5 points

  on: High stakes in­stru­men­tal­ism and billion­aire philanthropy

  Good post! Thanks for link­ing to my work. I also agree with Larks that it’s nice to have aca­demic poli­ti­cal the­ory brought in here.

Par­allels Between AI Safety by De­bate and Ev­i­dence Law

AI Benefits Post 4: Out­stand­ing Ques­tions on Select­ing Benefits

• Cullen_OKeefe 10 Jul 2020 1:54 UTC

  2 points

  in reply to: akrolsmir’s comment on: How to mas­sively in­crease your dona­tions, for free

  Rele­vant fo­rum post: https://​​fo­rum.effec­tivealtru­ism.org/​​posts/​​MAod5gvcQgdx­aXdWA/​​long-term-dona­tion-bunching

An­titrust-Com­pli­ant AI In­dus­try Self-Regulation

AI Benefits Post 3: Direct and Indi­rect Ap­proaches to AI Benefits

AI Benefits Post 2: How AI Benefits Differs from AI Align­ment & AI for Good

• Cullen_OKeefe 24 Jun 2020 17:40 UTC

  1 point

  in reply to: NunoSempere’s comment on: Prob­lem ar­eas be­yond 80,000 Hours’ cur­rent pri­ori­ties

  In­dia v. Pak­istan seems very im­por­tant as well

• Cullen_OKeefe 23 Jun 2020 23:50 UTC

  3 points

  on: Long-term Dona­tion Bunch­ing?

  Rele­vant for 2020: Due to the CARES Act, in­di­vi­d­u­als can deduct 100% of their AGI this year.

• Cullen_OKeefe 23 Jun 2020 23:50 UTC

  7 points

  in reply to: Carl_Shulman’s comment on: In­vest­ing to Give Begin­ner Ad­vice?

  It’s ac­tu­ally 100% for 2020 due to the CARES Act!

CARES Act Allows Char­i­ta­ble De­duc­tion of 100% of Gross In­come in 2020

• Cullen_OKeefe 22 Jun 2020 21:26 UTC

  1 point

  in reply to: Larks’s comment on: AI Benefits Post 1: In­tro­duc­ing “AI Benefits”

  Thanks! You are cor­rect. Up­dated to clar­ify that this is meant to be “the sub­set of AI Benefits on which I am fo­cus­ing”—i.e., non­mar­ket benefits.

• Cullen_OKeefe 22 Jun 2020 20:09 UTC

  1 point

  in reply to: Daniel_Eth’s comment on: Is it suffer­ing or in­vol­un­tary suffer­ing that’s bad, and when is it (in­vol­un­tary) suffer­ing?

  1. suffer­ing that is “mean­ingful” (such as mourn­ing)

  This might be a spe­cific in­stance of

  3*) Suffer­ing that is a nat­u­ral re­sult of healthy/​nor­mal/​in­evitable/​de­sir­able emo­tional reactions

AI Benefits Post 1: In­tro­duc­ing “AI Benefits”

• Cullen_OKeefe 19 Jun 2020 18:47 UTC

  5 points

  on: How to Fix Pri­vate Pri­sons (and Im­mi­gra­tion)

  One such goal might be to max­i­mize the to­tal so­cietal con­tri­bu­tion of any given set of in­mates within the limits of the law (limits such as “Don’t re­strict the free­dom of in­mates af­ter their re­lease”).

  Of course, pris­ons also serve an im­por­tant de­ter­rent func­tion, which is not well-ad­dressed by this model.