Arjun Panickssery is in London
@panickssery
Followers
1,574
Following
1,645
Media
467
Statuses
2,013
Researching scalable oversight @MATSprogram | prev @METR_Evals @ai_risks | spaced repetition | AI safety |
The Object Level
Joined January 2019
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
الهلال
• 1211822 Tweets
Neymar
• 347548 Tweets
كاس الملك
• 342121 Tweets
Ronaldo
• 222251 Tweets
رونالدو
• 122174 Tweets
FLORENCIA AL 9009
• 114741 Tweets
Happy Pride Month
• 66964 Tweets
Defante
• 54738 Tweets
Al Hilal
• 50082 Tweets
كريستيانو
• 46518 Tweets
Stolas
• 38211 Tweets
Al Nassr
• 37992 Tweets
sabina
• 35491 Tweets
وليد
• 34979 Tweets
Marian Robinson
• 33569 Tweets
كره القدم
• 31983 Tweets
対象作品
• 29377 Tweets
#Smackdown
• 29316 Tweets
نيمار
• 26800 Tweets
コインor時短
• 24647 Tweets
تمبكتي
• 23830 Tweets
الدون
• 21243 Tweets
كوليبالي
• 19872 Tweets
ديما بونو
• 15596 Tweets
新シーズン
• 14334 Tweets
今北産業
• 11045 Tweets
Happy New Month
• 10305 Tweets
Pinned Tweet
Of course that's your contention. You're a first-year effective altruist; you just got finished reading some Oxford philosopher, Will MacAskill probably. You're gonna be convinced of that till next month when you get to Nick Bostrom. Then you're going to be talking about how the
16
30
725
"I'm autistic."
"Does that mean you take things literally?"
"No, you're thinking of kleptomaniacs."
115
3K
33K
Nietzsche at age 24 was appointed a professor of classical philology without a doctorate on the strength of eminent philologist Friedrich Ritschl's recommendation alone
The recommendation is absurd; the John Nash rec letter people pass around doesn't hold a candle
17
114
1K
Five years ago Scott Alexander explained that it actually only takes him "a couple of hours" to write a post
he expresses confusion about people who write slowly: "they seem to be able to talk instantaneously, and writing isn't that different from speech" lol
Elsewhere he's
54
40
904
Scott Alexander anonymously won his own book review contest before disqualifying himself
11
9
719
This is what it feels like reading the first 10 chapters of HPMOR
10
25
405
People need to RETVRN to traditional company naming where it's always the surnames of the founders and possibly a literal description of the product
"Altman AGI Company"
"Page and Brin Internet Search Corp"
"Dorsey Short-Messages, Inc"
"Musk Electric Motors"
11
27
415
Fun to hate on rationalists for being weirdos, but don't forget that trying to 'debate' most people requires an endless string of
- "what about <absolutely trivial counterexample>"
- "it's not clear how this evidence supports your claim"
- "could you be more specific; e.g., do
20
12
408
"Researchers have failed to identify the Big Five dimensions in non-student adult populations in Bolivia, Ghana, Kenya, Laos, Sri Lanka, Vietnam, and Macedonia ... instead, researchers often find only two:
- Industriousness ...
- Prosociality ..."
15
34
399
@scrd_rv
For the unenlightened:
(Also I highly recommend Moby Dick; the writing is beautiful and Ahab is profoundly moving, possibly the greatest character in American literature)
2
18
348
Are LLMs biased toward themselves?
Frontier LLMs give higher scores to their own outputs in self-eval. We find evidence that this bias is caused by LLM's ability to recognize their own outputs
This could interfere with safety techniques like reward modeling & constitutional AI
8
46
323
"Every supposed genius has a bag of tricks—a list of obscure technical methods that hardly anyone knows about, that they have mastered. Every time they hear about a problem, they go through the list ... once every year or two, you get a match, and then you look brilliant"
14
28
297
ChatGPT on whether I should invade Persia:
9
25
292
Kissinger's undergrad thesis at Harvard was 400+ pages long and led them to establish a word limit afterward
4
10
239
Do people in their 20s with extremely pessimistic AGI predictions/timelines act as if their lifespan is probably going to be very short?
Do they save for retirement, plan to have kids, or care about long-run health risks from e.g. smoking that they claim they likely won't reach?
36
5
216
I'm skeptical that tech people dress down "because they can prove ability objectively"
Instead it's probably just SF hippie/counterculture that permeated outward right
e.g. I remember in Amazon's DC office no one would wear a t-shirt
How do they dress in Europe for example?
38
0
179
Were you gonna plagiarize the whole thing for us? Do you have any thoughts of your own on this matter? Or do you, is that your thing, you come into a bar, read some obscure passage and then pretend—you pawn it off as your own, as your own idea just to impress some girls, embar...
4
1
180
"Effective altruism" is too inscrutable so I usually tell people I'm at "an AI conference"
this morning at passport control in Stockholm the guy says "by AI you mean that you are summoning the demon, no?"
"Many experts think so, sir" (some chuckles)
"including the billionaire,
2
7
172
@Bo_Beidi
"knocking the rudders off sailboats"
"and especially when ... it requires a strong moral principle to prevent me from deliberately stepping into the street, and 𝐦𝐞𝐭𝐡𝐨𝐝𝐢𝐜𝐚𝐥𝐥𝐲 𝐤𝐧𝐨𝐜𝐤𝐢𝐧𝐠 𝐩𝐞𝐨𝐩𝐥𝐞'𝐬 𝐡𝐚𝐭𝐬 𝐨𝐟𝐟"
@scrd_rv
For the unenlightened:
(Also I highly recommend Moby Dick; the writing is beautiful and Ahab is profoundly moving, possibly the greatest character in American literature)
2
18
348
4
4
153
EAGx Cambridge afterparty DJ: "the EA crowd are mostly quite young and generally not that musically aware (or even into music)"
17
8
147
The Interim CEO of OpenAI, on AI safety:
"You are building an alien god that will literally destroy the world when it reaches the critical threshold but be apparently harmless before that"
@jonst0kes
...that's zero percent of the AI safety discourse I'm part of? My AI safety discourse is 100% "you are building an alien god that will literally destroy the world when it reaches the critical threshold but be apparently harmless before that"
4
7
75
8
22
148
coordination that is vastly stronger and more intrusive than that required to merely prevent a few exceptional and localized existential risks? You got that from Hanson's ‘𝐿𝑜𝑛𝑔 𝑅𝑒𝑓𝑙𝑒𝑐𝑡𝑖𝑜𝑛’ 𝐼𝑠 𝐶𝑟𝑎𝑧𝑦 𝐵𝑎𝑑 𝐼𝑑𝑒𝑎, paragraph 4, right? Yeah, I read that too.
1
5
129
governments of America and the USSR should have been focused on x-risks all the way back in 1945. That's gonna last until next year; you're gonna be in here regurgitating Toby Ord, talkin' about, you know, the Long Reflection and the utility-forming effects of delayed decisions.
1
1
130
@paulg
The more common anecdotes among famous writers are that they edit ferociously, get long periods of writer's block, etc
On the other hand people like Scott and Christopher Hitchens and Matt Yglesias talk a lot about writing fast
4
1
121
We're often reminded of our relative luxury (we're top 1% in the world if we make $30k/yr, etc) but from the broader perspective of everyone who will ever exist our lives could easily be among the nastiest, brutishest, and shortest experiences that there ever will be
6
4
122
So I paid out a Fiverr task to get the text of all the U.S. presidential inaugural addresses but I no longer remember why I did this
Any ideas for what I could do with this information
20
2
119
once a girl flirting at a party said her favorite book was War and Peace (haven't read it)
I said I'd heard that everyone you'll ever meet can be found somewhere in War and Peace
I asked which one I was; she said "Anatole," giggled and said she was kidding
what does this mean
7
1
117
Moved into my house June 1 and met a Berkeley CC girl whose mom drove her to move into the apt property next door
When asked if we were students I said "oh no, none of us are; most of us work in AI" the mother immediately said "oh God you guys are going to destroy everything"
1
3
105
OpenPhil: AI x-risk is one of the most important things
Also OpenPhil: AI x-risk grantmaking team of literally one person, governance team mysteriously not in DC despite OpenPhil having an office
6
4
103
First feedback I got on the draft was "I think people would dislike it"
5
1
102
"Well, as a matter of fact, I won't, because to prevent this sort of change over timescales of centuries or longer would require a global coordination . . ."
1
1
96
Update: wide awake/hyper alert but trapped under the five weighted blankets used to immobilize myself
Called out to girlfriend but remembered I broke up with her because I was worried about mattress microvibrations
Housemates can't hear me over the 105-decibel noise machine
It seems I've become one of those Sleep Guys
I got a sleep mask (which I sometime wake to find neatly folded on my bedside table), some kind of air purifier thing, and just put blackout film over my windows
Results to come
3
0
25
0
2
85
Michael Lewis's SBF biography 𝐺𝑜𝑖𝑛𝑔 𝐼𝑛𝑓𝑖𝑛𝑖𝑡𝑒 is scheduled for October 3
Lewis: "He’s let me see everything. When it all comes crashing down last November it’s literally me, his parents and his therapist. And I can hang around as long as I want, see whatever I want"
3
7
82
Why do rationalists come up with funny words for things which already exist
The LessWrong page for "Murphyjitsu" (I guess named for Murphy's Law) describes premortems—a strategy for anticipating likely failure modes—which are well-known enough to have a Wikipedia page
10
0
80
Just met a literal governess, ie she travels with a rich family managing their 5-year-old
I'm vaguely aware there's an insular network of rich-family nannies but it's still interesting
I guess differentiating quality is hard for something like that so it's all word of mouth?
7
4
79
Surely the rationalists—by their own values—must admit that it would be incredibly high EV for someone to condense the Sequences down to the length of a normal book
8
1
76
0
1
74
"If—God grant—he lives long enough, I prophesy that he will one day stand in the front rank of German philology.
...
He is the idol and, without wishing it, the leader of the whole younger generation of philologists here in Leipzig who – and they are rather numerous – cannot
1
0
71
Bill Clinton apparently had a little pen-and-paper CRM of everyone he met that totaled 10k+ names by the time he reached his governor days in Arkansas
6
1
69
If some cognitively demanding jobs (novelist? mathematician?) only manage fewer than 6 hours per day, there's a sense in which these are the best jobs to try to compete in
you can work hard and then relax for the rest of the day, assured that your competition can't get any edge
6
2
67
Live reaction thread of the Michael Lewis SBF book below
3
11
65
Underrated Scott Alexander post: "Epistemic Learned Helplessness"
In analogy to the psychological behavior, ELH is when people give up on trying to discern true from false arguments and just believe non-weird stuff
Scott argues that this is good
1
4
64
Signaling trades off against wealth
"Most nations today would be richer if they had long ago just submitted wholesale to a rich nation, allowing that rich nation to change their laws, customs ... But this idea greatly offends national and cultural pride. So nations stay poor."
10
6
63
Is this claim from Matt Levine correct? Doesn't ignoring the Kelly criterion maximize EV over any time scale (e.g. your life)
32
1
65
So I can't buy Anthropic stock, but I 𝑐𝑎𝑛 buy FTX debt at like $0.50 on the dollar right
👀
1
1
63
2
0
59
Old post on
@robinhanson
's site, 1997:
"Imagine we each had a personal "party talk" email list. To my personal list I post all my "party" comments, and also appearing there are the contributions of others to the conversations these comments are part of."
Sounds like Twitter
3
4
62
Online reaction to Yudkowsky suggesting that multinational agreements be enforced with airstrikes
2
2
60
What should people who believe in eg 15-year AGI timelines be investing in (I assume that if this prediction is true, it's not priced in right now)
Long Nvidia?
9
3
58
But then I explained briefly that they work on "interpretability, trying to make the AI models comprehensible to their operators" and she thought this was pretty based, though she didn't use that term
"is that going well"
"essentially no"
1
2
57
I often hear people say they increase their efficiency/impact by trading time for money, e.g. valuing their time at $35/hr and booking an Uber for $15 to save 30 minutes of walking as a result
But is this compatible with your marginal time value in most models of productivity?
16
2
56
Once at an EA party overhearing a nuanced discussion of utilitarianism I staggered drunk over to the whiteboard scribbling in huge font that the true ethics was the "3 virtues" (produced off the top of my head in the spur of the moment):
1. Loyalty
2. Hustle
3. Straight-talking
4
1
58
New EA Cause Area: Build a tower tall enough to reach the heavens
1. Probably useless
2. Non-zero chance that we succeed and then "nothing will be restrained from" us
3. So basically +Inf EV
Makes sense under hits-based giving
8
4
55
Narrator: he did not know how much money they had
@cz_binance
SBF on CFOs:
functional religion around the CFO ... people cannot articulate a single thing the CFO is supposed to do. They’ll say ‘keep track of the money,’ or ‘make projections.’ I’m like, What the fuck do you think I do all day? You think I don’t know how much money we have?
2
1
14
3
2
58
In The Doomsday Machine, Daniel Ellsberg mentioned how as a 27-year-old he forewent RAND's retirement plan in anticipation of untimely death by nuclear war.
Do young people who put high credence on an AGI-related apocalypse within ~20 years seem to be making updates like this?
2
1
54
much of the postrat style is just people who experienced clear thinking, noticed that the truth is quite harsh, and declared "but what if we be wrong 𝘰𝘯 𝘱𝘶𝘳𝘱𝘰𝘴𝘦 (winks loudly)"
6
5
50
wait to hear him as a lecturer. You will say, I describe a phenomenon. Well, that is just what he is—and at the same time pleasant and modest.”
"I should stake my whole philological and academic reputation that the matter would work out happily"
1
0
49
If Anthropic triples in value then FTX can pay everyone back
Today, we’re announcing that
@Amazon
will invest up to $4 billion in Anthropic. The agreement is part of a broader collaboration to develop reliable and high-performing foundation models.
238
1K
6K
1
1
51
"None of our board members are effective altruists"
OpenAI board lying (on the public Internet, no less)
8
4
48
Just learned how Magic: The Gathering works
Actually nuts, imagine if I was like "oh no FIDE just released a new piece called the Dark Tower which instantly checkmates me, I need to update my build"
1
2
47
Gotta start sending this to my friends when they don't reply to my messages
2
1
44
Just checked and HPMOR is 10% longer than War and Peace
Hard to argue that I should finish HPMOR instead if I could spend the time one way or the other
2
1
44
No Soylent on this island but they've got some kind of Huel vending machine in Heathrow
8
2
45
What are the worst grammar rules that are still very canonical
I don't like putting commas after i.e. and e.g.
The ".," is very unappealing
15
0
43
Batman on existential risk:
"He [Superman] has the power to wipe out the entire human race, and if we believe there’s even a one percent chance that he is our enemy we have to take it as an absolute certainty… and we have to destroy him."
3
2
41
@nvpkp
@sympatheticopp
This strategy is great for signaling as described in Chapter 7: Conversation in Robin Hanson's 𝐸𝑙𝑒𝑝ℎ𝑎𝑛𝑡 𝑖𝑛 𝑡ℎ𝑒 𝐵𝑟𝑎𝑖𝑛
1
0
40
Rawls's veil is about the society you'd want not knowing who in society you'd be (not only your wealth/status, but your preferences/desires)
But if you also hide your 𝑡𝑖𝑚𝑒 in history, you'd want a different, growth-focused society, since you'll probably be born near the end
5
1
39
I think gossip is bad and people should mostly mind their own business
It is still my general opinion that gossip is a bit underrated. If you hear of someone behaving badly, guess what you think happened, and if people ask, tell them about it. Maybe even if they don't.
But also, if you find out it's not true, spread that too.
5
0
26
7
0
40
When people ask for the "most underrated" <thing> they usually mean an unpopular one that's really good but often the most underrated things are actually the most popular, which are even better that their popularity suggests
It's like the Planet Money episode about how LeBron
2
0
40
Biology people love citations
They'll open their papers like "The brain is located in the head (Walker et al, 1990). It is believed to be involved in cognition (Dickson, 1997; Smith, Jones, and Harrison, 2012; Johnson, 2015)."
2
3
37
People keep telling me that "everyone dies" and yet only 93% of people have died. Curious.
2
4
37
bruh it immediately resigned after the blunder because "I'm down material and don't see a way to regain the advantage"
1
1
37
Hyper-productive late-19th-century Pope Leo XIII took cocaine “to fortify himself when prayer was insufficient”
1
5
37
"I couldn’t believe that the world would long escape ... Alain Enthoven and I were the youngest members of the department. Neither of us joined the extremely generous retirement plan RAND offered. Neither of us believed, in our late twenties, we had a chance of collecting on it."
0
0
37
So is the key to getting clout just posting thousands of words a week and then waiting
Quantity over quality
9
0
37
Life Pro Tip: after you read great literature (e.g. Moby Dick), use crazy turns of phrase in regular conversation
"Does that coffee shop have anything good?"
"Oh I'm sure that just as the appalling ocean surrounds the verdant land, we will find in that menu some insular Tahiti"
4
3
36
Cavendish Labs is a new x-risk research lab recruiting their first round of fellows for any duration within their 6-month residency in rural Vermont (two hours north of Boston)
Excellent team working variously on AI alignment, ultraviolet light on viruses, LAMP diagnostics, etc
1
4
35
When around physics people at a party it's good to ask "Oh I have a physics question: What is energy anyway?"
If necessary add "In particular what is this so-called 'potential' energy, that sounds so fake"
Guaranteed physicsbabble conversation of whatever length you desire
0
2
35
Interesting
maybe I should read this book
0
0
34
WW2 general Hitoshi Imamura, known for "unusually lenient" rule over occupied Indonesia, sentenced by the Australians to 10 years for other war crimes, decided his release was premature and spent the rest of his life (making 21 years altogether) in a replica of his prison cell
2
4
34
holy shit I made a typo and it caught it immediately
are we all actually going to die
1
0
32
The real jargon sin isn't "alpha" or "orthogonal" but saying prisoner's dilemma when you mean stag hunt
1
1
32
I don't understand how people have conflicts with their roommates
When I had 3 roommates, we would simply establish common knowledge of our beliefs and then update to the same viewpoint in accord with Aumann's Agreement Theorem
1
0
31
@ShakeelHashim
Do you think that voting should be public for the same reason, the reasoning is the same
1
0
32
Kissinger's 400+ page thesis led to a word limit at Harvard
Just learned that Caro's 235-page senior thesis led to a word limit at the Princeton English Department
The "one weird hack" to become a Great Man of the 20th century is to just write monstrously long undergrad theses
Kissinger's undergrad thesis at Harvard was 400+ pages long and led them to establish a word limit afterward
4
10
239
2
0
31
lol I pulled up the bachelor's thesis and it opens: "An introduction to an undergraduate honor thesis may seem presumptuous, but I believe that its inordinate length and unorthodox method require an explanation. As a general reason, the length is due to the fact that I . . ."
Kissinger's undergrad thesis at Harvard was 400+ pages long and led them to establish a word limit afterward
4
10
239
3
0
30
Giving a toast at an EA party before a shot:
"To shrimp!"
Another shot a few minutes later:
". . . To digital shrimp!"
2
1
31
@fehloop
You could be less doubt in the premises and the negation of the conclusion put together than in your ability to discover the error in the argument even if it were false
This gives you a reason to reject arguments without being able to refute their logic
Underrated Scott Alexander post: "Epistemic Learned Helplessness"
In analogy to the psychological behavior, ELH is when people give up on trying to discern true from false arguments and just believe non-weird stuff
Scott argues that this is good
1
4
64
1
0
29
The survivorship bias plane remains highly underrated, people routinely say stuff like
- "Don't try to fake X, people can tell"
- "Look at all the very successful people who do <suspect but high variance behavior>"
- "Stuff was better in the past, look at this famous old stuff"
2
2
31
A dollar invested in bonds in 1801 would be worth $1k in 1998; in stocks, $500k; and in gold, $0.78
From the latest
@krishnanrohit
post
4
1
30
From Rob Henderson's "Reverse Dominance Hierarchies," a review of Christopher Boehm's 𝐻𝑖𝑒𝑟𝑎𝑟𝑐ℎ𝑦 𝑖𝑛 𝑡ℎ𝑒 𝐹𝑜𝑟𝑒𝑠𝑡
1
1
29
Finally watched Parasite on a flight
Did they really make a hit film about how the poor are parasites lol
The sort of reversal take that actually the rich family are parasites is completely outrageous right
3
1
28
This London morning walking to the train I first saw a fox
Tonight returning from the office in Oxford spied I my first hedgehog
Profound!
4
0
29
Ascending tiers of "history buff guy": the Civil War guy, the WW2 guy, the Rome guy
Secret top tier is the Cold War guy
8
1
28