Filip Sondej

• Filip SondejOct 28, 2022, 1:31 PM

  1 point

  0 ∶ 0

  on: EA Fo­rum fea­ture sug­ges­tion thread

  Add an option for drafts: “Anyone with a link can read”, but make it really anyone, not only forum users, as it is now.

  (Recently I wanted to get feedback from some people who are not on the forum, and I had to copy draft to google doc, and later copy it back, and fix all the footnotes :/​ )

  Next step (but probably harder), would be to let anyone comment. If they aren’t logged into forum, these comments are anonymous.

  Also collaborative editing in markdown mode would be useful.

• Filip SondejOct 28, 2022, 1:09 PM

  1 point

  0 ∶ 0

  in reply to: Emrik’s comment on: New tool for ex­plor­ing EA Fo­rum and LessWrong—Tree of Tags

  I’d also want to have an option to switch the marks’ visibility (and on default they should be off, to not distract from reading). With that, I wouldn’t even require author approval, it would be more like commenting, but line-specific.

  So the reader particularly interested in some section could dive into the comments particularly about that section.

  Also, as a further feature, you could color code different comment types, like:

  • yellow: fix suggestion

  • red: critique

  • blue: link to previous discussion /​ relevant resources

  • green: just a comment

  FYI: I posted that suggestion on the EA forum feature suggestion thread, and also linked your comment

• Filip SondejOct 28, 2022, 11:28 AM

  4 points

  1 ∶ 0

  in reply to: Charlie_Guthmann’s comment on: New tool for ex­plor­ing EA Fo­rum and LessWrong—Tree of Tags

  Yeah, I’d love to see some novel curation mechanisms too. I’m a bit scared to introduce money to the mix though. Someone might be tempted to exploit the system, by using bots to fake engagement. Which would be a loss because you could no longer trust vote count as an indicator of value.

  Other way to incentivize curation, would be to tip people for great curated lists of posts. (Although I suspect many people would do it even without the money incentive, if the forum made it easy to do.) For example an option to just publish the list of all your strongly upvoted posts, would enable quite effortless curation.

  I guess that the main problem isn’t that some posts are objectively undervalued, but rather that it’s hard to find those old posts that are right specifically for you. Posts that are:

  • in a topic that interests you

  • novel for you

  • not so advanced, that you can’t follow them

  There is also an ML approach, where you could find some embedding for each post on the forum, and then train your personal classifier to predict which posts you would like.

  Or even more simply, a collaborative filtering approach—“people who like what you like, also like this”.

• Filip SondejOct 28, 2022, 11:20 AM

  1 point

  0 ∶ 0

  on: Database dumps of the EA Forum

  Here is some code you can use to get the most essential data from all the forum posts: https://​​github.com/​​filyp/​​tree-of-tags/​​blob/​​main/​​tree_of_tags/​​forum_queries.py

  Also, I used that data to build a topic model of the forum as you mentioned. You can see it here

New tool for ex­plor­ing EA Fo­rum and LessWrong—Tree of Tags

• Filip SondejSep 7, 2022, 4:11 PM

  3 points

  0 ∶ 0

  in reply to: Jeffrey Kursonis’s comment on: Sen­tience or Silos?

  In that case of AlphaGo, I think a more accurate word is “constructed” to win, rather than prompted. And humans are also constructed—by evolution.

  Oh yeah, as an aspect of some conscious states, desire is definitely very important. For example, it may be a necessary condition for suffering.

  It would be great to pin down what must happen structurally in some neural network, to produce a experience infused with desire.

• Filip SondejSep 4, 2022, 4:44 PM

  5 points

  0 ∶ 0

  on: Sen­tience or Silos?

  I don’t think desire is necessary for consciousness or sentience. Imagine a meditating Buddhist monk who managed to purge their mind from all desire (at least in that moment), but nevertheless is conscious.

  Also, asking whether artificial agents desire, seems like asking”can machines think?” or “can submarines swim?”—it just depends on which things you wish to call “desire”. There is some sense in which AlphaZero “wants” to win a match, in the sense that it is pursuing that goal.

• Filip SondejSep 4, 2022, 3:37 PM

  2 points

  0 ∶ 0

  in reply to: IrenaK’s comment on: EAG & EAGx lodg­ing considerations

  What do you think about adding a field in the application form, like: “I’d like to share a room with other EAG attendees”. Then you can just lump together people who checked it into some email group, and let them coordinate among themselves.

• Filip SondejSep 4, 2022, 2:48 PM

  5 points

  0 ∶ 0

  on: Style Writ­ing Prej­u­dice At EA

  There is some component of style, that is arbitrary and unnecessary—for example a lot of academic writing is overcomplicated for no good reason. Or using too much jargon, where you could say the same in plain English. I agree that there’s no need to imitate such things.

  But a large part of some style is meant to serve a purpose. I think EA forum style is in general well oriented towards finding the truth—by emphasis on clarity, concreteness, honesty and backing things up with sources.

  Even if expecting all those writing virtues may feel demanding for writers, I think it definitely is justified, and it would be sad to lose those virtues.

• Filip SondejSep 4, 2022, 11:42 AM

  5 points

  0 ∶ 0

  on: Neu­ro­science of time per­cep­tion?

  We can separate 3 things:

  • the feeling of how fast time is passing now

  • your estimate of how much time has passed in the past day or year for example

  • the actual amount of experiences that happened in the past day or year

  I think those three are distinct things (even though they correlate). For example the feeling of the passage of time can be drastically altered with psychedelics—it’s possible to feel that time is not passing at all. (Here is a nice video which lists spooky time-effects and speculates how are they produced). But even though that moment may feel like eternity, it’s not that there’s actually an infinite amount of experience happening.

  As for the estimate of the time passed, as you and other commenters noted, it looks to be based on how much memories you have from some time period. So someone with dementia would probably estimate that much less time has passed recently, even if a lot of experiences actually happened.

  It’s nice to have good memories, but I think what ultimately matters are the actual experiences and their valence and intensity (even if you forget them later).

  Still, the things you list sound cool, so thanks for reminding me to do them :)

• Filip SondejJun 9, 2022, 2:20 PM

  2 points

  0 ∶ 0

  in reply to: John Litborn’s comment on: New co­op­er­a­tion mechanism—quadratic fund­ing with­out a match­ing pool

  You’re right, the leverage definitely goes two ways. The thing it, this later leverage will tend to be smaller than the one you get immediately. At least, this is how the system behaves in my naive simulations. The exception is, when you expect some very big contributors to join later on—then the later leverage is bigger. So yeah, it’s a complicated situation and I didn’t want to go into that in the post, because it would get too bloated.

  And yeah, humans and TAI may have different strategies which complicates it further. This is why I’m not yet fully satisfied with this mechanism, and I will try to simplify it, so that we don’t have to care for all those strategies.

• Filip SondejJun 6, 2022, 8:01 PM

  2 points

  0 ∶ 0

  in reply to: Sinclair Chen’s comment on: New co­op­er­a­tion mechanism—quadratic fund­ing with­out a match­ing pool

  Yeah, I agree that using this to further fund AI alignment wouldn’t help much. I’m less sure about “hitting the metric”—the thing is, we don’t have any good alignment metric right now. But if we somehow managed to build it, convincing AI labs to hit such a metric seems to me like the most feasible thing to make AI race safer. But yeah, building it would be really hard. Do you maybe have some other ideas how to make AI race safer? Maybe it is possible to somehow turn them into a continuous value that they could coordinate to increase?

  Re: strategic thinking—It may be true that most people won’t care so much for their real leverage (they won’t consider the counterfactual where they donate less), but it definitely isn’t rational. So while it may more or less work, I wouldn’t like this system to give an impression that it tricks people into donating. And, more importantly, my main hope for this system, is to facilitate cooperation between most powerful agents (powerful states, future supercorporations, TAI systems), rather than individual people. I assume such powerful actors will consider what happens if they do not donate, and selfishly do what’s optimal for them.

• Filip SondejJun 6, 2022, 6:44 PM

  2 points

  0 ∶ 0

  in reply to: Austin’s comment on: New co­op­er­a­tion mechanism—quadratic fund­ing with­out a match­ing pool

  Awesome!

  I’d love to see the idea tested in a real world situation. I’d be happy to help with building this system if you want :)

New co­op­er­a­tion mechanism—quadratic fund­ing with­out a match­ing pool

• Filip SondejMar 4, 2022, 12:31 PM

  3 points

  0 ∶ 0

  in reply to: Sasha Berezhnoi 🔸’s comment on: What are effec­tive ways to help Ukraini­ans right now?

  OK, I underestimated them. Thanks for clarifying!

  What they do is certainly impressive.

• Filip SondejMar 2, 2022, 11:31 PM

  6 points

  0 ∶ 0

  in reply to: Sasha Berezhnoi 🔸’s comment on: What are effec­tive ways to help Ukraini­ans right now?

  Thanks for the recommendations! Of the two, OVD-Info looks particularly promising. While Meduza seems partially redundant (correct me if I’m wrong), because all the information is already available from other sources, OVD-Info really seems to make anti-regime actions easier.

  You convinced me to donate :)

• Filip SondejMar 2, 2022, 11:14 PM

  1 point

  0 ∶ 0

  in reply to: Levan Bokeria’s comment on: What are effec­tive ways to help Ukraini­ans right now?

  Yes! That’s exactly my thoughts. In addition, if Putin faces high resistance, it will show other regimes that such actions will not be tolerated.

  Can you tell more about what those trusted sources say about Blue/​Yellow? I’m considering making a significant donation but I won’t do it blindly. How exactly can they be more effective than Ukrainian army?

  I’m particularly interested in affecting the outcome and aftermath of this war (long-term effects), rather than short-term humanitarian help.