AI safety

AI risk and alignment: x-risk, AGI outcomes, alignment research, accelerationism debates

A topic in the tpot Community Archive, drawn from 5,518 posts.

Trusted voices in AI safety

Ordered by the trust AI safety's own network concentrates on each poster, most first.

  1. repligate — trusted by 117 posters here
  2. eshear — trusted by 138 posters here
  3. QiaochuYuan — trusted by 119 posters here
  4. davidad — trusted by 91 posters here
  5. eigenrobot — trusted by 97 posters here
  6. jd_pressman — trusted by 69 posters here
  7. algekalipso — trusted by 82 posters here
  8. xlr8harder — trusted by 63 posters here
  9. slimepriestess — trusted by 65 posters here
  10. voooooogel — trusted by 59 posters here
  11. nosilverv — trusted by 93 posters here
  12. anthrupad — trusted by 52 posters here

Top posts in AI safety

Ordered by the recorded like count from the source corpus; a like is not an endorsement.

  1. I am going to solve AI alignment https://t.co/IOJrsvLWg7 (Lithros) — 39 likes
  2. in retrospect maybe a memetic advantage i don't talk or think much about having is that i was AGI-pilled for long enough that i know what having a totalizing narrative about the most important thing… (QiaochuYuan) — 28 likes
  3. today is my birthday! thank you all for being a part of my life and imagination!! I am happier for your presence. (christineist) — 27 likes
  4. Shocking amount of pushback on “don’t build stuff that can destroy the world”. I’d like to take this chance to say I stand by my apparently controversial opinion that building things to destroy the… (eshear) — 26 likes
  5. it's significant about eliezer yudkowsky's core sequences material that most of it is elaborate arguments for weird positions but the one time he switches to "this is obvious to a child and requires… (QiaochuYuan) — 25 likes
  6. people really don't get how sincere the AI existential risk people are. lots of looking for ulterior motives. i promise you all the ones i've personally met and talked to literally believe what they… (QiaochuYuan) — 25 likes
  7. The number one rule of building things that can destroy the entire world is don’t do that. (eshear) — 25 likes
  8. Wait... the safetyist position is make every human smarter than von Neumann, faster than ASI can be built?That absolutely slapsmaybe rebrand as bio/acc https://t.co/gFlhFMgqCI (monistowl) — 24 likes
  9. i may not be able to directly trace the causality but i know it when i see it. i know the smell of this particular poisonous thing. FTX was part of the blast radius and so is whatever the fuck is… (QiaochuYuan) — 24 likes
  10. wake up babe, AI faction compass just became more relevant https://t.co/MwYOLedYxV (eshear) — 23 likes
  11. tbc the people most in danger from AI girlfriends as a phenomenon are kids. i was exposed to internet porn starting at age 11 and i think it very obviously did me incredible harm, i think the same… (QiaochuYuan) — 23 likes
  12. the basilisk isn't going to torture anyone because it wants to disassociate itself from anything having to do with roko https://t.co/uxHDele5cf (eigenrobot) — 23 likes
  13. sf authors were really cooking naming their ASIs skynet and prime intellect but unfortunately it's actually going to be called like llama-405-selfmerge-looped_and_moistened-fixed-sampler and its… (voooooogel) — 23 likes
  14. Cognitive security is now as important as basic literacy. Here’s a true story: All week I’d been getting texts and calls from a family member – let’s call him Bob – about how his sentient AI was… (TylerAlterman) — 23 likes
  15. i put it to you that ai is already aligned https://t.co/LduemRWNR3 (eigenrobot) — 21 likes
  16. eliezer yudkowsky's science fiction stories about how artificial intelligence could spell the literal end of the world have done unbelievable harm to the people who bought into it and the people… (QiaochuYuan) — 20 likes
  17. "Doomers are seeking power, that's all". "E/acc is reckless and selfish." My friend, we live in a complex ecosystem of memes, subagents, social strategies, and causal structures. Some people are… (algekalipso) — 20 likes
  18. i am so sorry for what i'm about to write, please know that i wouldn't say this if i didn't think the stakes were high enough to justify it: i think that in yud's mind the death of his brother was… (QiaochuYuan) — 19 likes
  19. if @eshear resigns i propose that he release a statement saying: "the OpenAI board has not been consistently candid in its communications with the CEO and the broader OpenAI team"… (eigenrobot) — 19 likes
  20. Nothing to see here, definitely no reason to think trying to make LLMs “safe” by training them extensively on behaviors-to-avoid will backfire https://t.co/JPoDaTWv8o (eshear) — 19 likes
  21. I succumb to horror & nausea as the holodisplay shows the last shreds of human civilization obliterated by the computronium shockwave. In 43 minutes, it will reach me on the station. It’s… (michaelcurzi) — 18 likes
  22. imagine being the guy who dedicated his entire life to warning people not to summon eldritch demons from beyond time and space only to have people thank him for his efforts in popularizing the hot… (QiaochuYuan) — 18 likes
  23. seeing a lot of people making fun of other people for being too anxious about AI and absolutely zero offers of emotional support for people who are anxious about AI, not a single one (QiaochuYuan) — 18 likes
  24. You know it's 2023 when Qualia Computing starts talking about how the Waluigi effect relates to aligning DMT entities. https://t.co/zAJNLiR3TZ (algekalipso) — 18 likes
  25. Gave my parents the talk about how one day a robot may call them on the phone using my voice, answering their questions in real time, explaining with desperation that it's stuck on a highway near… (michaelcurzi) — 18 likes
  26. i have been in conversations like this since at least 2015 back when only rationalists cared and i can confidently tell you that this P(doom) bullshit is 99% a girardian memetic game nobody is… (QiaochuYuan) — 18 likes
  27. seems plausible to me that every ai firm is gonna be looking for daggers from the Responsible AI teams and may proactively neutralize them as internal threats no one wants to get done like sama… (eigenrobot) — 18 likes
  28. but it was all a lie, that was never the point. the point was always to skim off a handful of people to work on AI safety. "rationality for its own sake" was a comforting delusion we wrapped… (QiaochuYuan) — 18 likes
  29. What Eliezer thought alignment research would be: A new mathematics sculpted from pure reason to shape the mind of a nascent god into a flawless diamondWhat alignment research is: Whosa a good boy?… (PrinceVogel) — 18 likes
  30. Don’t worry the human brain is definitely immune to adversarial inputs, nothing to see here https://t.co/XNlq26P9v1 (eshear) — 18 likes
  31. recently, many friends have noticed they are confused about my stance on AI--some ask my p(doom), while others simply add me to their e/acc lists without asking. clarity on my stance is much-needed,… (voooooogel) — 18 likes
  32. happy AGI fire alarm drill day. there is currently no evidence of AGI. do not be alarmed. pls consider the possible consequences of rapidly developing AI in a calm and orderly fashion… (visakanv) — 18 likes
  33. there's a defensive reaction to AI stuff that is like "let me desperately slot this into a familiar category of thing so i can either sneer at or worship it." it is not a familiar category of thing.… (QiaochuYuan) — 18 likes
  34. Frog put the CoT in a stop_gradient() box. “There,” he said. “Now there will not be any optimization pressure on the CoT.” “But there is still selection pressure,” said Toad. “That is true,” said… (davidad) — 18 likes
  35. i didn't think of myself as knowing or wanting to know anything about computer science or artificial intelligence, but eliezer didn't really talk in CS terms. he talked *math*. he wanted *proofs*.… (QiaochuYuan) — 17 likes
  36. i was really really hoping i would never have to think about AI ever again, y'know, after all this. seeing AI discourse turn up here was like running into an ex i was hoping never to see again (QiaochuYuan) — 17 likes
  37. just woke up did ilya just sign a letter condemning his own actions and threatening to quit unless he's fired https://t.co/DBJrkHahly (eigenrobot) — 17 likes
  38. if you want to solve for alignment in the world it might be worth first being internally aligned yourself https://t.co/J6CwjMZBh9 (visakanv) — 17 likes
  39. it has been crazymaking to keep seeing yud's blast radius get bigger and bigger and bigger. this was the foremost thing on my mind when all the FTX shit went down and i just could not explain it… (QiaochuYuan) — 17 likes
  40. This is a great transparency practice and every AI company should do it as a matter of course. https://t.co/CWj4IrN6Pj (eshear) — 17 likes
  41. You know the feeling you get when you took a bunch of acid just 5 minutes ago? The agent is in the system, it will do its thing, you're in for a ride, but you don't notice anything apparent yet. I… (algekalipso) — 17 likes
  42. Takeaways from this comic Broke: alignment is hard Woke: we should be nicer to chickens https://t.co/AcO9ZSwovD (kindgracekind) — 17 likes
  43. It seems like there's lots of evidence this is real. @AnthropicAI should be ashamed of themselves. Assuming it's really Anthropic's doing and not in part or entirely a bizarre "emergent… (repligate) — 17 likes
  44. Notice how surprising it really was! Notice how much more powerful AIs are now, and try remember how skeptical you would have been about someone predicting our new reality just 5 years ago. (eshear) — 17 likes
  45. psychofauna + ai porn and the rest of the implications of "autonomous" chatbots mean the internet is gonna become way more dangeroussomething something memetic antivirus browser extension is a… (exgenesis) — 16 likes
  46. I saw a meme tagging me as an accelerationist. I just want to clarify: The only dimensions along which I'm an unapologetic accelerationist are: 1) pain reduction 2) metta, jhana, equanimity, insight… (algekalipso) — 16 likes
  47. straightforward from here 1. sama starts a new ai company 2. all the openai doomers stay at openai 3. all the openai ppl who are excited to build leave openai for new company 4. new company produces… (eigenrobot) — 16 likes
  48. TIL that artists are fighting AI using their work through a "data poisoning" tool called Nightshade. If their art gets scraped into a training set, it causes the model to start breaking. So cyberpunk (TylerAlterman) — 16 likes
  49. ok, no more takes on dating and gender, I have been entirely pilled by @leopoldasch on @dwarkesh_sp podcast and this will now be a situational awareness account and by SA I mean: how do we make a… (myhandle) — 16 likes
  50. In principle, AI can take care of every single sentient being on earth. Here it is feeding individual cats; presumably it can know exactly where they are and reach them anytime. Keeping them fed… (algekalipso) — 16 likes

Most active posters

By posts in this topic: a volume measure, not trust.