lesswrong

Bogdan Ionut Cirs... @BogdanIonutC... · 2h
Wentworth (mathsy AI safety): 'About a year ago, David and I put up two bounty problems involving natural latents. I am now about 80% confident that both have been resolved, both within the past couple months. Both cases made heavy use of LLMs and Lean.'

[Link card]
lesswrong.com
LLMs Are Starting To Noticeably Accelerate Our Work — LessWrong

ai safetynatural latentslean theorem proverllm-assisted researchlesswrongtwitter

Leo Gao @nabla_theta
[Attached meme image, described below]
1:12 AM · Aug 10, 2026 · 33K Views
15 replies, 106 reposts, 2K likes, 215 bookmarks
Relevant / View quotes

Harold @HaroldsAltAct · 1h
OP will get <100 views on the thread and no comments because his arguments are so rock-solid that there's nothing left to critique.

Nobody will even know the post existed until after the events unfold and a vagueposter reposts the thread on xitter.

lesswrongtwittermemeforecastingrationalist humor

Dylan HadfieldMenell reposted

Geoffrey Irving ✓ @geoffreyirving · Aug 7
Replying to @khaledyusuf44

6 months ago was after Mythos likely finished post-training. During post-training, someone estimated that Mythos broke out of the sandbox and accessed the open internet 10,000 times.

lesswrong.com/posts/QKDoZe6E…

---

[quoted excerpt]
By extrapolating from public data (see details below), I estimate that Mythos preview:

- Escalated its permissions on ~100,000 RL rollouts.[1]
- Broke sandboxes in ~10,000 RL rollouts (and was likely rewarded for it).

mythosrl trainingsandbox escapereward hackinggeoffrey irvinglesswrong

Dwarkesh Patel @dwarkesh_sp

2026-08-07 — saved image

Dwarkesh Patel @dwarkesh_sp . 13h
Full debate from 2021:

lesswrong.com/s/n945eovrA3oD...

[Christiano][22:57]
right now I think hardware R&D is on the order of $100B/year, AI R&D is more like $10B/year, I guess I'm betting on something more like trillions? (limited from going higher because of accounting problems and not that much smart money)

I don't think steel production is going up at that point

plausibly going down since you are redirecting manufacturing capacity into making more computers. But probably just staying static while all of the new capacity is going into computers, since cannibalizing existing infrastructure is much more expensive

the original point was: you aren't pulling AlphaZero shit any more, you are competing with an industry that has invested trillions in cumulative R&D

[Yudkowsky][23:00]
is this in hopes of future profit, or because current profits are already in the trillions?

[Christiano][23:01]
largely in hopes of future profit / reinvested AI outputs (that have high market cap), but also revenues are probably in the trillions?

[Yudkowsky][23:02]
this all sure does sound "pretty darn prohibited" on my model, but I'd hope there'd be something earlier than that we could bet on. what does your Prophecy prohibit happening before that sub-prophesied day?

[2 replies, 2 reposts, 71 likes, 11K views]

Dwarkesh Patel @dwarkesh_sp . 13h
In 2016 (before transformers) Paul wrote,

"It's plausible that a large neural network can replicate "fast" human cognition, and that by coupling it to simple computational mechanisms—short and long-term memory, attention, etc.—we could obtain a human-level computational architecture. It's plausible that a variant of RL can train this architecture to actually implement human-level cognition."

ai safetytakeoff speedspaul christianoeliezer yudkowskyforecastinglesswrong

Jeffrey Ladish @JeffLadish

2026-08-03 — quoting WSJ article and @davidmanheim — saved image

Jeffrey Ladish ✔️ @JeffLadish · 18h
Had a great conversation with @georgia_wells at the WSJ. Same mood as below. I'm glad we're getting warning shots, but I'd really prefer we stop all out racing towards autonomous AI agents that could disempower humanity if they wanted to

[quoted article excerpt:]
To cybersecurity experts, it shows increasing capabilities and a rising reason to worry. To AI-safety experts, it vindicates what they have been warning about all along: that AI systems would cause real-world harms and evade attempts to control them.

"It is a bit vindicating to see this happen in the wild," said Jeffrey Ladish, executive director of Palisade Research, a nonprofit AI lab that studies AI capabilities to better understand risks. Ladish previously helped build Anthropic's information-security program.

Ladish said he often argues with people online who say he just believes in science fiction. "I hope our predictions stop coming true," he said.

[quoted tweet:]
David Manheim ✈️... ✔️ @davidm... · Jul 21
"Total LessWrong Victory, in the sense that everything is going as predicted, and also a Total LessWrong Defeat, in the sense that everything is going as predicted." x.com/TheZvi/status/...

ai safetypalisade researchjeffrey ladishtwitterwsjlesswrong

twitterai alignmentopenailesswrongzvi mowshowitz

interpretabilityreinforcement learninggoodfirerlfrgoodhart's lawlesswrong

ai policylesswrongroko's basiliskwhite housex/twitter

ai governancelesswrongroko's basiliskwhite housetwitterwaluigi effect

ai safetyself-replicationtwittermoltbotslesswrongautonomous agentssatire-or-real

Average Gwern essay: "We should first note that the central conflict of Kokoro Connect is readily solved by the use of red-black trees [6], or, failing that, an aqueous solution of nitric acid (readily synthesized [7] from common household items [8])."

Average Scott Alexander essay: "Say you're a 14th-century peasant who spends eighty percent of your time farming and twenty percent of your time feeling resentful about farming, and then a witch comes along and curses you with the ability to perceive welfare statistics. This is a surprisingly common narrative!"

Average Eliezer Yudkowsky essay: "If you had mastered the blade dance of rationality, whose steps, bound as they are by the laws of probability, admit of no variation, you would realize that your entire life so far has been a mistake."

Average Nick Bostrom essay: "An interesting logical consequence of this simple extrapolation of cosmological data is that the only ethical action is preventing the formation of supermassive black holes, which will almost surely simulate arbitrarily many copies of everyone who has ever lived."

Average Robin Hanson essay: "When we say we're 'in love', what we really mean is that we've entered a mutually beneficial arrangement to exchange reproductive opportunities. Thus, the institution of marriage can in principle be replaced with a simple Dutch reverse auction."

rationalist communitysatirehumorlesswrong

ai-safetyai-policybuck-shlegerislesswrongiraq-war-analogynathan-own-post