Hi everyone,

New posts

1. Advice from Tyler Cowen.
   • Top rated MR comment is ofc "This is just so very blindingly false" 🙂
2. Questions.
   1. Tim Ferriss: "The English is a little rough in places, but I found some of these questions to be excellent reminders and many others to be outstanding brain food." 🥲
3. 10 notes from watching pulp fiction yet again

Links

link: guzey.com/questions

questions

1. what are you thinking about these days?
2. what happens to your consciousness when you walk into a teleporter?
3. what's the most embarrassing cold email you've sent?
4. what does good future look like?
5. what are you absolutely certain in?
6. does time exist?
7. what's one actionable from last meeting?
8. how did the romanovs manage to rule russia for 300 years?
9. why do you live where you live?
10. what are the key questions?
11. why not take a pen and a piece of paper, go offline for 1 hour and resolve one of them?
12. why is the market capitalization of facebook, founded 15 months prior to yc by a 19 year old, larger than that of all yc companies combined?
13. what do you want?
14. is self eternal?
15. what's slowing you down the most?
16. why is democracy the best form of government?
17. when was the last time you took 3 months completely off?
18. how do we understand paul christiano's prophecies before they come to pass?
19. what's your destiny?
20. is 1 week 2% of the year?
21. how can you do better?
22. how to solve prompt injections?
23. when was the last time you asked for help?
24. how to go to the center of the universe?
25. what inspires you?
26. why did the founding fathers decide that the president must be at least 35 year old?
27. what would make your colleagues happy?
28. what are you devoted to?
29. why did the anarchists murder the single most progressive russian emperor (alexander ii) but not his extremely conservative successor?
30. what are you confused about?
31. how does the history judge your work?
32. why will you fail?
33. why is it easier to do something for the other person than to do the exact same thing for yourself?
34. what are you flinching from?
35. why not just stare at it for 10 minutes?
36. does god know the future?
37. what would you die for?
38. did the person actually tell you no or did you just assume they would?
39. why do holden karnofsky and sam altman write so differently?
40. who do you want to be more like?
41. who do you want to be less like?
42. what would you do if you were alone in the universe?
43. would you walk into a teleporter?
44. does solution to ai risk exist?
45. how to solve ai risk?
46. how to measure ai risk?
47. what makes you do the right thing?
48. how do we know if an llm is telling us what it actually thinks?
49. who are you afraid of?
50. what is ai alignment?
51. when was the last time you prayed?
52. how to measure ai alignment?
53. what are you proud of?
54. why did krishna reveal himself only to arjuna?
55. when was the last time you did something you've never done before?
56. what's the right way to regulate ai?
57. do you feel better or worse after spending time with the person you spend most of your time with?
58. was alexander worse than hitler?
59. what are you afraid to tell to your best friend?
60. when should one give up?
61. what's the ground-truth measure?
62. why did you do something you regret?
63. how to upload brains w/o discontinuity of consciousness?
64. what are you afraid to tell to your partner?
65. why did scott alexander shut down his blog and mobilized the entire sv intelligentsia against an NYT reporter who was writing an extremely balanced and charitable profile of his blog?
66. why not travel to your favorite city as often as you can?
67. do fundamental laws of physics exist?
68. am i wasting my time writing these questions?
69. what have you promised to do and not done yet?
70. is consciousness continuous?
71. when was the last time you took a year off?
72. why did the japanese keep debating whether to surrender even after two nuclear bombings?
73. who's advice has been so correct in the past that you now just go and act on it?
74. why not ask them for a piece of advice?
75. why do children who grow up without real authority become spoiled?
76. what's the top 1 work goal today?
77. what's the top 1 personal goal today?
78. are failure and/or success conditions clear?
79. why are so many actors and actresses political activists?
80. what are you flinching from thinking about?
81. why do people sign false confessions?
82. when was the last time you talked to someone you really admire?
83. is solving alignment different from building agi?
84. what have you learned about yourself in the last year?
85. what would you do in utopia?
86. are ai capabilities more like those of bombs or of computers?
87. what are you going to regret in a year?
88. how does the brain generate consciousness?
89. have you spent 1 minute today on your biggest goal?
90. have you spent 1 hour today on your biggest goal?
91. have you spent 10 hours today on your biggest goal?
92. how to prepare for the times of change?
93. can you be more specific?
94. what's the most common advice you give to people?
95. do you follow it yourself?
96. does the world need to be saved?
97. what can you do in the next 60 seconds that will make you feel proud of yourself?
98. what are you wrong about?
99. what's agi?
100. why not invite someone you like talking to for a dinner?
101. why do people disagree on p doom?
102. what's p doom?
103. what's doom?
104. who's consistently ahead of you?
105. why don't you catch up?
106. how to overcome the second law of thermodynamics?
107. who should you report to?
108. how did zuck make a comeback?
109. what do you wish you could do but you're absolutely confident you can't?
110. how do you know?
111. what's the humanity's frontier?
112. should you do lasik?
113. what exists?
114. what doesn't exist?
115. when was the last time you closed your eyes for a minute and asked yourself if you're doing the right thing?
116. what can you do all day long and feel great going to bed?
117. what is the shape of the solution to ai alignment?
118. where in the world is your tribe?
119. what are the key factors of pluralist future?
120. do you like yourself when you look in the mirror?
121. what's the strongest case for technological accelerationism?
122. what's the strongest case against technological accelerationism?
123. does anyone believe they're evil?
124. what's the eta to the dyson sphere around the sun?
125. what do other people tell you about you that you are always surprised to hear?
126. why?
127. why does anything exist at all?
128. what's your plan?
129. what is the level of resolution you'd need to replicate your brain in to create someone you'd consider to be you?
130. what would you do if you couldn't make the world better?
131. how do the experiences of sleep, general anesthesia, and death differ?
132. do you believe in yourself more than you believed 5 years ago?
133. have you updated all the way?
134. why is maintaining eye contact sometimes easy and sometimes difficult?
135. is a laptop a superintelligence?
136. have you read a good history book today?
137. what's the rubric for the role you're hiring for?
138. what are you trying to do?
139. how did the humanity manage to never deploy a hydrogen bomb?
140. what's the next step?
141. what are the key factors of good future?
142. what do you have no competitors in?
143. how do you know what someone will do when the push comes to the shove?
144. what's the right thing to do?
145. what's humanity's destiny?
146. what do you feel behind on?
147. what's the silver bullet?
148. how do you know?
149. why not?

Hi everyone,

New posts:

1. After years of dismissing AI alignment as a bs field, I’ve come to think that it’s finally into real science (and we should really start making progress on it soon.): AI Alignment Is Turning from Alchemy Into Chemistry
2. A simple two-sentence prompt can’t possibly jailbreak both GPT-4 and Claude, can it? And even if it did, surely, it would be easy to patch the vulnerability that made it work, right? Well… A Two sentence Jailbreak for GPT-4 and Claude & Why Nobody Knows How to Fix It

Links

Hi everyone,

New posts:

1. After years of dismissing AI alignment as a bs field, I’ve come to think that it’s finally into real science (and we should really start making progress on it soon.): AI Alignment Is Turning from Alchemy Into Chemistry
2. A simple two-sentence prompt can’t possibly jailbreak both GPT-4 and Claude, can it? And even if it did, surely, it would be easy to patch the vulnerability that made it work, right? Well… A Two sentence Jailbreak for GPT-4 and Claude & Why Nobody Knows How to Fix It

Links

Hi everyone,

Welcome to the AI world. It's exciting & scary but hopefully more exciting than scary. And hopefully we'll figure out how to make aliens on GPUs love us, rather than hate us (or just be indifferent towards us).

New posts:

1. My 2022 self (I don't know them) was very wrong about meditation, huge monitors, and... sleep.
2. lifehacks [137 of them...]

Links

• ppl underappreciate Sam Altman's blog: https://blog.samaltman.com/successful-people (What are your favorite posts?).
• Lena: "MMAcevedo's demeanour and attitude contrast starkly with those of nearly all other uploads taken of modern adult humans, most of which boot into a state of disorientation which is quickly replaced by terror and extreme panic. Standard procedures for securing the upload's cooperation such as red-washing, blue-washing, and use of the Objective Statement Protocols are unnecessary."
• Ukraine war
  • Kherson liberated.
    • https://twitter.com/bayraktar_1love/status/1592103969413025792
    • https://twitter.com/Tendar/status/1591114773281656832
  • KUPIANSK OFFENSIVE OPERATION
  • If you'd like to significantly contribute to Ukraine's defense, please reach out. Otherwise, donate to one of these orgs: https://standforukraine.com/.

From friends

• Skejeton is an extraordinarily talented 18-year old living in Kazakhstan and is prolific in C, C++, JS, and Rust. He's looking for coding gigs. I can provide reference.
  • CV: https://gist.github.com/skejeton/2371cd61ae505abd616ce4742b406cc5
  • Email: ishidex2@gmail.com
• Metaphor: "We're building a new search engine from scratch, using the same ideas behind DALL-E and Stable Diffusion. It understands language — in the form of prompts — so you can say what you're looking for in all the expressive and creative ways you can think of."
  • From the incredibly talented Alex Gajewski.
• Everyprompt: "Everyprompt is a pretty good playground for large language models like GPT-3"
  • From the incredibly talented Evan Conrad.
• The incredible Isaak Freeman has started a group house in Berkeley (with good culture, not Berkeley culture). I'm going to be visiting it in January and potentially afterwards. Strongly consider dropping by or staying over if you're visiting the Bay Area.
• The incredible Milan Cvitkovic from Convergent Research is looking for "ambitious, neurotech-interested people (any experience level)" for:
  • A case study of all the neurotech-related DARPA programs.
  • A case study on Medtronic's Activa system.
  • Expansion of Neurotechnology Numbers Worth Knowing.
  • Other incredibly exciting neurotech projects.

Hi everyone,

New posts & essays since the last update:

• Oct 31: What is the alternative to utilitarianism?
  • "From my observations, people who try to take utilitarianism seriously have their brains break." (unexpectedly relevant to recent events)
• Nov 9: Forecasting humans becoming generally intelligent with biological anchors: year 10,000,000,000,002,022
  • Poking at the meaningfulness of the biological anchors method of estimating AGI timelines.
• Nov 11: Who is the real prophet of longtermism: Will MacAskill or Sam Altman?
  • Demonstrating the unexpected overlap in the ideas Sam Altman was professing in 2015 and EA longtermism circa 2022.
• Nov 16: Planes are still decades away from displacing most bird jobs
  • Demonstrating the absurdity of arguments about capabilities of AI by comparing them "everything humans can do".
  • 300+ poins and 100+ comments on Hacker News.
• Nov 21: If the moon doesn't need gravity, why do we? The necessity of understanding for general intelligence
  • Poking at the meaningfulness of claims that humans have accurate physical world-models and that they are central to general intelligence.

Thoughts & ideas always appreciated!

Hi everyone,

Large Language Models

LLMs are going to change what our brains are capable of at least as much as the invention of writing changed it.

GPT-3 solves my problems, helps me to write, and gives me ideas I didn't think of daily, and there's no sign of LLM progress slowing down.

Hi everyone,

New essay:

• Q&A with my high school self: helping 14-16 year old Alexey to deal with his emotions, to ask for help, to talk to people (and his dad), to learn, to get things done. (unusually hermeneutic style for me)

Old essay that I keep referring people to at least once a week:

Hi everyone,

• I’ve been mostly incredibly hard at work with New Science.

  • Announced a one-year fellowship. We’ll provide young life scientists with a bench in Cambridge, Massachusetts with access to world-class shared equipment, an $80k stipend, a $100k in project costs, and a custom package for every project. If you have really smart friends doing biology, let them know!
  • Accepted several incredibly smart fellows for the summer.
  • Got featured in the Atlantic.
  • Continuing to hire and always fundraising.

• Published Theses on Sleep.

  • Natália Mendonça published Counter-theses on Sleep, getting 3x karma on LessWrong (see comments for a bit of back-and-forth between us).
  • Got a Muse S headband. I want to run an experiment during which I’ll sleep ~5 hours a day for 28 days and will then spend 7-14 days not restricting my sleep and proving with Science that my brain waves are normal when I don’t restrict sleep to everyone who thinks I’m a mutant or something.
  • Hope to publish a big post on O-1 visa in Q2

• Got several of my friends out of Russia in early March. Contrary to the information I had, the borders have not closed yet. Hopefully the war will end soon, with Russia safely isolated from the rest of the world.

  • In latest news, a Russian ballistic rocket (with “For Children” written on it) was deliberately launched at an Ukrainian railway station that was evacuating civilians, killing more than 50, including 5 kids.
  • My wife vetoed me burning my Russian passport and my immigration lawyer told me in strongest terms that I should not renounce the Russian citizenship yet 😓

• Elsewhere:

  • https://stratechery.com/2022/the-intel-split/ (Intel is becoming TSMC’s customer)
  • https://www.lesswrong.com/posts/5DKqK3hEzzBoGF47C/consider-taking-zinc-every-time-you-travel
  • Chris Beiser:

    pretty wild how there’s social pressure on high schoolers/college students to set up half-real volunteer/nonprofit groups, and hype them up with no consideration for how long they last or accomplishments where, when adults do this, we call it “being a con man”

  • Nick Jikomes:

    Just learned that fluvoxamine, a common SSRI used to treat depression and other psychiatric conditions, increases the half-life of caffeine in the bloodstream.

    Like, to an absurd degree

  • Black Hawk Down - The Battle of Mogadishu 1993, Part 1 - Animated

Best,
Alexey

(for all of my non-Russian subscribers, apologies for the out-of-schedule email)

Kamil Galeev was the first source 2 days ago, confirmed by 3 more (private) sources since then. Kamil:

Для россиян - очень скоро выезд перекроют. Сначала для мужчин сколько-нибудь призывных возрастов, потом для всех

Рассмотрите возможность вотпрямщас выехать в любую открытую страну (кроме Беларуси). Грузия, Армения, Азербайджан, Казахстан, Монголия и далее по списку. Сейчас такая возможность еще есть, но ее перекроют быстро и без предупреждения

Если вы хотите или считаете своим долгом оставаться в РФ, это ваш выбор. Но если жить в этом вы не желаете, валить надо немедленно, лучше сегодня

---

1. Положение РФ гораздо хуже чем в Холодную войну. Полный ад может наступить внезапно и без предупреждения. Что касается экономического коллапса, так он гарантирован. На это могу поставить достаточно уверенно, потому что паровой каток санкций уже поехал. Западные госаппараты более механистичны, а потому большей степени подвержены институциональной инерции чем российский. Их трудно сдвинуть с места, но как сдвинули, так они едут-едут-едут и так еще очень долго. Потому что их не только запустить, но и остановить трудно, не существует человека который мог бы отдать такой приказ

2. Если хотите уехать, уезжайте немедленно, чем быстрее тем лучше, в любую открытую страну кроме Беларуси. Долго прицениваться какая лучше смысла нет, главное пересечь границу, а там видно будет. Казахстан и прочие вряд ли будут перекрывать выезд, а РФ будет. Если пока не определились куда и что делать дальше можно в Стамбул - это крупнейший хаб и оттуда летают почти во все страны

3. Как пересечете можно думать над дальнейшими действиями. Они понятно зависят от ваших финансов, умений, нетворка и множества сопутствующих обстоятельств

Это рекомендация которую я могу дать всем кто думает об отъезде. Что касается более конкретных соображений, я пока не готов выкладывать их в открытый доступ (подписчики могут посмотреть здесь (https://www.patreon.com/posts/63248958)). Но в отличие от всех остальных запейволенных текстов, этот я открою как только сочту возможным

Для тех кто хочет уехать но не знает ситуацию на местах - вот список русскоязычных телеграм чатов по 105 странам мира. Можно почитать и если что непонятно уточнить детали у тех кто уже там https://telegra.ph/chat360-10-14

Summary of the essay:

In this essay, I question some of the consensus beliefs about sleep, such as the need for at least 7 hours of sleep for adults, harmfulness of acute sleep deprivation, and harmfulness of long-term sleep deprivation and our inability to adapt to it.

It appears that the evidence for all of these beliefs is much weaker than sleep scientists and public health experts want us to believe. In particular, I conclude that it’s plausible that at least acute sleep deprivation is not only not harmful but beneficial in some contexts and that it’s that we are able to adapt to long-term sleep deprivation.

I also discuss the bidirectional relationship of sleep and mania/depression and the costs of unnecessary sleep, noting that sleeping 1.5 hours per day less results in gaining more than a month of wakefulness per year, every year.

The theses are:

1. Comfortable modern sleep is an unnatural superstimulus. Sleepiness, just like hunger, is normal.
2. Depression <-> oversleeping. Mania <-> acute sleep deprivation
3. Occasional acute sleep deprivation is good for health and promotes more efficient sleep
4. Our priors about sleep research should be weak
5. Decreasing sleep by 1-2 hours a night in the long-term has no negative health effects

Hi everyone,

Writing

• My low-demand, worry-free, and healthy diet (spoiler: it's Athletic Greens, muesli with kefir, and MealSquares)
• Why I switched my newsletters from Substack and Mailchimp to Buttondown
• Thinking of writing about my experience with getting an O-1 visa to work at New Science. Let me know if you'd be interested.

New Science

Hi everyone,

Personal updates

• My O1 extraordinary ability visa was approved by the United States Citizenship and Immigration Services, and I moved to Boston 3 weeks ago, becoming New Science’s Executive Director.

• I will be speaking at Foresight Institute’s Foresight Vision Weekend 2021 in San Francisco. The event will take place on December 4-5 and is one of the most exciting conferences happening all year - I strongly recommend attending. Say hi if you see me there.

New writing

Hi everyone,

New writing

• published a list of all winners of Emergent Ventures

• the 15,000-word piece on Bloom et al (pdf) I’ve been working on since last October is finally getting published in a few days…

New Science