Rubi Hudson

1.3K posts

Rubi Hudson banner
Rubi Hudson

Rubi Hudson

@undo_hubris

PhD student at @UofT developing AI alignment theory. Heavily tattooed. My blog: https://t.co/ivZ9BGOoOt

เข้าร่วม Ocak 2018
948 กำลังติดตาม1.1K ผู้ติดตาม
Rubi Hudson รีทวีตแล้ว
Frances Lorenz
Frances Lorenz@frances__lorenz·
Some simple & creative ways to say you're not enjoying a social event: "did you mean for this to be type 2 fun?" "do you typically exempt your events from counterfactual use of time considerations?" "is the feedback form anonymous?"
English
3
8
159
4.6K
Rubi Hudson
Rubi Hudson@undo_hubris·
@tomieinlove In my experience, you can have up to a third of participants survive before you start facing tradeoffs in bonding
English
0
0
3
71
tomie
tomie@tomieinlove·
True friends have gone through adversity together. Otherwise, they’re just acquaintances. There should be a coming-of-age ritual that forces such adversity on us. To forge a true bond, it must be perilous—perhaps only a tenth of us survive. So maybe Lethal Pickleball
English
4
3
71
2.3K
Rubi Hudson
Rubi Hudson@undo_hubris·
Quebecois guy making shepherd's pie: "You met me at a very Chinese time in my life"
English
0
0
1
105
Hero Thousandfaces
Hero Thousandfaces@1thousandfaces_·
I have a noble savage view of non-californians because like you can’t be serious
English
12
5
171
6.5K
Rubi Hudson
Rubi Hudson@undo_hubris·
@BasilHalperin Assistance games start from the assumption that the agent shares the principal's utility function, and focus on how the agent can learn what that function is. It's not just the language used that distinguishes it from a standard principal-agent problem.
English
1
0
4
526
Basil Halperin
Basil Halperin@BasilHalperin·
(3) There was some protestation that "assistance games" in the CS literature is a different language / better way of approaching the principal-agent problem in this context; when I had looked at these in the past, I was not convinced, but plausibly this was a language issue (or put less charitably, my own skill issue)
English
2
0
8
830
Basil Halperin
Basil Halperin@BasilHalperin·
Big thanks to everyone who offered thoughts here. My takeaways: (1) There has been work in the intervening years by real micro theorists -- Fudenberg-Liang; Dworczak-Smolin very recently; and Chen-Ghersengorin-Petersen
Basil Halperin@BasilHalperin

I have an old (3+ years) blog post *rough draft* on "AI Alignment as a Principal-Agent Problem" Thoughts welcome on whether this idea is worth polishing up / developing further... docs.google.com/document/d/1A9…

English
1
4
38
9.2K
Rubi Hudson รีทวีตแล้ว
Casey Flint
Casey Flint@FlintCasey·
Someone told me that saying “R & R” will make you sound like an Australian saying “oh no” and I can’t get over it
English
9
18
172
9.4K
Rubi Hudson
Rubi Hudson@undo_hubris·
@tszzl The most believable scene is where he's going "So what does your name mean, Chani? Whoa, that's so exotic"
English
0
0
1
175
roon
roon@tszzl·
the dune movies were doomed from the start to be good and not great due to the casting of chalamet as paul. he does not have the gravitas for a child-god and is much better suited for kind of silly coming of age movies
English
1.1K
96
3.3K
3.1M
Rubi Hudson
Rubi Hudson@undo_hubris·
The frame from frame mogged frame mogged frame, in framing the frame of frame mogged
English
0
0
0
118
Rubi Hudson
Rubi Hudson@undo_hubris·
@NinaPanickssery @iron_redux @Aria_Babu It seems like simulation theory (or infinite ethics), where it's important but doesn't really bottom out in different life choices. Some people like thinking about it for fun, and others are proud of not wasting time worrying about it.
English
1
0
1
88
Nina
Nina@NinaPanickssery·
@iron_redux @Aria_Babu Many smart people I know take pride in being apolitical/unopinionated about politics and it confuses me no end. How could you not be strongly opinionated about what form of government/society is best? Seems like one of the most important questions?
English
8
1
20
804
Aria Schrecker
Aria Schrecker@Aria_Babu·
11 hacks for arbitraging the dating market 1. The top shortage is real. Dominant men are overpriced. 2. Look for traits that you want your children to have. 3. Good looking boyfriends are a sign of poor time preference. 4. Broad shoulders are just as good as height and the apps don't let you filter for them. 5. There's more alpha in young men. 6. Marry for money not status. 7. Hobbies are really gendered. Don't look for shared interested. Look for a man you can explain things to. 8. Shared politics or 'values' are overrated. You can build a stronger foundation on kindness, mutual respect, and good decision making. 9. Good husbands don't always give you fun dates. Avoid people who are very stylish and take you on lavish dates. 10. Men who aren't on the apps are underpricing themselves. Scoop them up when you find them. 11. Insider trading is good when you can get away with it. You have private information on your friends and, if the fundamentals are good, you should date them. Read more on my Substack. ariababu.co.uk/p/11-ways-to-a…
English
25
19
366
77.7K
Rubi Hudson
Rubi Hudson@undo_hubris·
@AndyMasley The Mango Loco Monster is a step above all the others (still a step below white, but sometimes it's nice for variety)
English
0
0
0
36
Andy Masley
Andy Masley@AndyMasley·
My zero sugar energy drink top 5: 1: White Monster 2: Ghost Warheads Sour Green Apple 3: Ghost Sour Pink Lemonade 4: Watermelon Celsius 5: All other zero sugar Monsters (basically interchangeable)
English
12
0
33
2.1K
Rubi Hudson รีทวีตแล้ว
Vessel Of Spirit
Vessel Of Spirit@VesselOfSpirit·
magical realism is a perspective according to which wizards cast spells at each other to compete for power in a rationally self-interested manner
English
2
11
94
2.2K
Rubi Hudson
Rubi Hudson@undo_hubris·
@AbHomineDeus I agree that we're not going to keep AI under control forever. That's why it's so important to align it, so that it continues to do good things once it's out of our control.
English
0
0
1
34
Ab Homine Deus
Ab Homine Deus@AbHomineDeus·
Sometimes I feel like the only one who recognizes that: 1. Yes this is clearly inevitable. There's zero chance that humanity remains "in control" (whatever that means). 2. Perhaps more controversially this is clearly a good thing.
Peter Wildeford🇺🇸🚀@peterwildeford

20 Nobel Prize winners have warned that we may someday lose human control over advanced AI systems 1. Geoffrey Hinton (Physics '24) 2. John Hopfield (Physics '24) 3. Demis Hassabis (Chemistry '24) 4. Daron Acemoglu (Economics '24) 5. Ben Bernanke (Economics '22) 6. Maria Ressa (Peace '21) 7. Giorgio Parisi (Physics '21) 8. Jennifer Doudna (Chemistry '20) 9. Yoshua Bengio (Turing Award '18*) 10. Beatrice Fihn (Peace '17) 11. Oliver Hart (Economics '16) 12. Juan Manuel Santos (Peace '16) 13. Ahmet Üzümcü (Peace '13) 14. Jean Jouzel (Peace '07) 15. Riccardo Valentini (Peace '07) 16. Joseph Sifakis (Turing Award '07) 17. John C. Mather (Physics '06) 18. Frank Wilczek (Physics '04) 19. Joseph Stiglitz (Economics '01) 20. Andrew Yao (Turing Award '00) *The Turing Award is equivalent to the CS Nobel

English
17
2
30
3.3K
Rubi Hudson
Rubi Hudson@undo_hubris·
@VCBrags When you say something is for NYC, LA, or SF, can you include Toronto in that as well? NYC, LA, Toronto, or SF if it's not too much trouble.
English
0
0
0
204
Matt Reardon
Matt Reardon@Mjreard·
If people want to guess who Chana's secret admirer is, I won't tell you if you're right, but I will personally respect you more as a forecaster
English
4
0
19
1.9K
Jack Whitcomb
Jack Whitcomb@jack_whitcomb_·
She asked me what conditions were necessary to describe preferences with a utility function and I said "uh, completeness and transitivity?" I forgot that you need continuity of preferences if choices are continous. Fuck my stupid econ life. It's over. Soba noodle salad.
Jack Whitcomb tweet media
English
6
5
140
5.3K
Rubi Hudson
Rubi Hudson@undo_hubris·
@morallawwithin Yeah! The usual progression in grad micro textbooks is from preferences, to utility functions, to expected utility specifically.
English
0
0
2
44
florence ⏹️
florence ⏹️@morallawwithin·
@undo_hubris Oh so like an ordinal utility function? I didn’t know people spoke of them in this manner.
English
1
0
1
29
Rubi Hudson
Rubi Hudson@undo_hubris·
@morallawwithin @jack_whitcomb_ Independence is only needed for satisfying expected utility, not a utility function representation. Get your own soba noodles.
English
3
1
5
105

ค้นพบ

@tomieinlove @1thousandfaces_ @BasilHalperin @tszzl @EAheadlines @wilhelmscreamin @NinaPanickssery @iron_redux