Ozyrus

[Question] Are we cer­tain that gpt-2 and similar al­gorithms are not self-aware?

[Question] Model­ing AI mile­stones to ad­just AGI ar­rival es­ti­mates?

[Question] What would be the signs of AI man­hat­tan pro­jects start­ing? Should a web­site be made watch­ing for these signs?

• Ozyrus 26 Sep 2016 23:25 UTC

  1 point

  on: Open thread, Sep. 26 - Oct. 02, 2016

  I’ve been med­i­tat­ing lately on a pos­si­bil­ity of an ad­vanced ar­tifi­cial in­tel­li­gence mod­ify­ing its value func­tion, even writ­ing some ex­crepts about this topic.

  Is it the­o­ret­i­cally pos­si­ble? Has any­one of note writ­ten any­thing about this—or any­one at all? This ques­tion is so, so in­ter­est­ing for me.

  My thoughts led me to be­lieve that it is the­o­ret­i­cally pos­si­ble to mod­ify it for sure, but I could not come to any con­clu­sion about whether it would want to do it. I se­ri­ously lack a good defi­ni­tion of value func­tion and un­der­stand­ing about how it is en­forced on the agent. I re­ally want to tackle this prob­lem from hu­man-cen­tric point, but i don’t re­ally know if an­thro­po­mor­phiza­tion will work here.

• Ozyrus 23 Dec 2015 16:04 UTC

  5 points

  on: Stupid Ques­tions, 2nd half of December

  Well, this is a stupid ques­tions thread af­ter all, so I might as well ask one that seems re­ally stupid.

  How can a per­son who pro­motes ra­tio­nal­ity have ex­cess weight? Been bug­ging me for a while. Isn’t it kinda the first thing you would want to ap­ply your ra­tio­nal­ity to? If you have things to do that get you more util­ity, you can always pay diet spe­cial­ist and just stick to the diet, be­cause it seems to me that ad­di­tional years to life will bring you more util­ity than any other ac­tivity you could spend that money on.

• Ozyrus 26 Aug 2015 15:05 UTC

  0 points

  on: Sen­sa­tion & Perception

  A good read, though I found it rather bland (talk­ing about writ­ing style). I did not read the origi­nal ar­ti­cle, but com­pres­sion seems ok. More will be ap­pre­ci­ated.

• Ozyrus 25 Aug 2015 7:58 UTC

  4 points

  on: Open Thread—Aug 24 - Aug 30

  Are there any less­wrong-like se­quences fo­cused on eco­nomics, fi­nance, busi­ness, man­age­ment? Or maybe just in­ter­net com­mu­ni­ties like less­wrong fo­cused on these sub­jects?

  I mean, the se­quences in­tro­duced me to some re­ally com­plex knowl­edge that im­proved me a lot, while si­mul­ta­neously be­ing en­gag­ing and quite easy to read. It is only log­i­cal to as­sume that some­where on the web, there must be some ar­ti­cles in the same style cov­er­ing differ­ent themes. And if there are not, well, some­one must surely do this, I think there is some de­mand for this kind of con­tent.

  So, feel free to link less­wrong-like se­ries of blog­posts on any theme, ac­tu­ally: that will be re­ally helpful for me. P.S. In hind­sight, i guess there may be some post here, on less­wrong, con­tain­ing all these links I am look­ing for. If so, could any­one link me to it?

• Ozyrus 20 May 2015 22:03 UTC

  1 point

  in reply to: Gram_Stone's comment on: Wel­come to Less Wrong! (7th thread, De­cem­ber 2014)

  It seems that your im­plicit ques­tion is, “If ra­tio­nal­ity makes peo­ple more effec­tive at do­ing things that I don’t value, >then should the ideas of ra­tio­nal­ity be spread?” That de­pends on how many peo­ple there are with val­ues that are >in­con­sis­tent with yours, and it also de­pends on how much it makes peo­ple do things that you do value. And I would >con­tend that a world full of more ra­tio­nal peo­ple would still be a bet­ter world than this one even if it means that there >are a few sadists who are more effec­tive for it. There are mur­der­ers who kill peo­ple with guns, and this is bad; but >there are many, many more sol­diers who pro­tect their na­tions with guns, and the ex­is­tence of those na­tions al­low >much higher stan­dards of liv­ing than would be oth­er­wise pos­si­ble, and this is good. There are more good peo­ple >than evil peo­ple in the world. But it’s also true that some­times peo­ple can for the first time fol­low their be­liefs to their >log­i­cal con­clu­sions and, as a re­sult, do things that very few peo­ple value.

  Ex­cel­lent an­swer! Yes, you de­ducted the im­plicit ques­tion right. I also agree that this is a rather ab­stract field of moral philos­o­phy, though i did not see that at first. Although I don’t think that your ar­gu­ment for the world be­ing a bet­ter place with ev­ery­one be­ing ra­tio­nal holds up, es­pe­cially this point

  There are more good peo­ple than evil peo­ple in the world.

  Even if there are, there is no proof that af­ter be­com­ing “ra­tio­nal” they will not be­come “bad” (apos­tro­phes be­cause bad is not defined suffi­ciently, but that’ll do.). I can imag­ine some in­ter­est­ing prospect for ex­per­i­ments in this field by the way. I also think that the re­sult will vary if the sub­ject is placed in so­ciety of only-ra­tio­nal­ists vs usual so­ciety—with “bad” ac­tions car­ried out more in the sec­ond ex­am­ple, as there is much less room for co­op­er­a­tion.

  But of course that is pointless dis­cus­sion, as the situ­a­tion is not re­ally based on re­al­ity in any way and we can’t re­ally tell what will hap­pen. :)

• Ozyrus 20 May 2015 16:43 UTC

  2 points

  on: Wel­come to Less Wrong! (7th thread, De­cem­ber 2014)

  Hello, ev­ery­one!

  LW came to my at­ten­tion not so long ago, and I’ve been com­mited to read­ing it since that mo­ment about a month ago. I am a 20-year old lin­guist from Moscow, finish­ing my bach­e­lor’s. Due to my age, I’ve been pon­der­ing with usual ques­tions of life for the past few years, search­ing for my path, my philos­o­phy, es­sen­tially, a best way to live for me.

  I stud­ied a lot of re­li­gions, philoso­phies, and they all seemed re­ally flat, es­sen­tially be­cause of the rea­sons stated in some ar­ti­cles here. I came close to some­thing re­sem­bling a nice way to live af­ter I read “At­las shrugged”, but some­thing about it both­ered me, and af­ter thor­ough anal­y­sis of this philos­o­phy I de­cided to take some good things from it and move on, as I did a lot of times be­fore.

  I found this gem of a site through red­dit and roko’s basilisk (is it okay if I say it here? I heard dis­cus­sion was banned). I am deeply into the whole idea of ra­tio­nal­ity and nearly all ideas that are pre­sented on this site, but some­thing re­ally both­ers me here, too.

  The thing is that it is im­plied that al­tru­ism and ra­tio­nal­ity go hand in hand; maybe I missed some im­por­tant ar­ti­cles that could ex­plain me, why?

  Let’s imag­ine a hy­po­thet­i­cal sce­nario: there is a guy, Steve, who re­ally does not feel any­thing when he helps other peo­ple nor when does other “good” things gen­er­ally; he does this only be­cause his philos­o­phy or re­li­gion tells them to. Say this guy was in­tro­duced to ideas of ra­tio­nal­ity and thus he is no longer bound by his philos­o­phy/​re­li­gion. And if Steve also does not feel bad about other peo­ple suffer­ing (or even takes plea­sure in it?)?

  What i wanted to say is that ra­tio­nal­ity is a gun that can point both ways: and it is a good thing that LessWrong “sells” this gun with a safety mechanism (if it is such “safety mechanism”. Once again, maybe I missed some­thing re­ally crit­i­cal that ex­plains why al­tru­ism and “be­ing good” is the most ra­tio­nal strat­egy).

  In other ways, Steve does not re­ally care about hu­man­ity; he cares about his well-be­ing and will uti­lize all knowl­edge he got just to meet his ends ( peo­ple are differ­ent, aren’t they? and ends are differ­ent, too).

  Or even an­other, av­er­age ra­tio­nal­ist Jack es­ti­mated that his own net gain will be sig­nifi­cantly big­ger if he hurts or kills some­one (con­sid­er­ing his emo­tions and feel­ings about over­all hu­man­ity net gain, and all other pos­si­ble fac­tors). That means he must carry on? Or is it a taboo here? Or maybe it is a prob­lem of this site’s de­mo­graph­ics and no­body even con­sid­ered this sce­nario (which fact I re­ally doubt).

  I feel that i dive too deep into metaphors, but i am not yet a good writer. I hope you un­der­stood my thought and can make me less wrong. :)

  edit: fixed formatting