Assumption News

Other

Humanity coordinates to prevent the creation of potentially-unsafe AIs.

AIs will not have utility functions (in the same sense that humans do not), their goals such as they are will be relatively humanlike, and they will be "computerish" and generally weakly motivated compared to humans.

Alignment is not properly solved, but core human values are simple enough that partial alignment techniques can impart these robustly. Despite caring about other things, it is relatively cheap for AGI to satisfy human values.

Multipolar AGI Agents run wild on the internet, hacking/breaking everything, causing untold economic damage but aren't focused enough to manipulate humans to achieve embodiment. In the aftermath, humanity becomes way saner about alignment.

Eliezer finally listens to Krantz [resolves NO]

Yudkowsky is trying to solve the wrong problem using the wrong methods based on a wrong model of the world derived from poor thinking and fortunately all of his mistakes have failed to cancel out

We create a truth economy. https://manifold.markets/Krantz/is-establishing-a-truth-economy-tha?r=S3JhbnR6

Ethics turns out to be a precondition of superintelligence

Hacks like RLHF-ing self-disempowerment into frontier models work long enough to develop better alignment methods, which in turn work long enough to ... etc; we keep ahead of 'alignment escape velocity'

We make risk-conservative requests to extract alignment-related work out of AI-systems that were boxed prior to becoming superhuman. We somehow manage to achieve a positive feedback-loop in alignment/verification-abilities.

AIs make "proof-like" argumentation for why output does/is what we want. We manage to obtain systems that *predict* human evaluations of proof-steps, and we manage to find/test/leverage regularities for when humans *aren't* fooled.

A lot of humans participate in a slow scalable oversight-style system, which is pivotally used/solves alignment enough

Something less inscrutable than matrices works fast enough

Humans become transhuman through other means before AGI happens

Aligned AI is more economically valuable than unaligned AI. The size of this gap and the robustness of alignment techniques required to achieve it scale up with intelligence, so economics naturally encourages solving alignment.

Humans and human tech (like AI) never reach singularity, and whatever eats our lightcone instead (like aliens) happens to create an "okay" outcome

Aliens invade and stop bad |AI from appearing

An AI that is not fully superior to humans launches a failed takeover, and the resulting panic convinces the people of the world to unite to stop any future AI development.

Someone at least moderately sane leads a campaign, becomes in charge of a major nation, and starts a secret project with enough resources to solve alignment, because it turns out there's a way to convert resources into alignment progress.

Someone solves agent foundations

Social contagion causes widespread public panic about AI, making it a bad legal or PR move to invest in powerful AIs without also making nearly-crippling safety guarantees

A smaller AI disaster causes widespread public panic about AI, making it a bad legal or PR move to invest in powerful AIs without also making nearly-crippling safety guarantees

Getting things done in Real World is as hard for AGI as it is for humans. AGI needs human help, but aligning humans is as impossible as aligning AIs. Humans and AIs create billions of competing AGIs with just as many goals.

Development and deployment of advanced AI occurs within a secure enclave which can only be interfaced with via a decentralized governance protocol

High-level self-improvement (rewriting code) is intrinsically risky process, so AIs will prefer low level and slow self-improvement (learning), thus AIs collaborating with humans will have advantage. Ends with posthumans ecosystem.

Power dynamics stay multi-polar. Partly easy copying of SotA performance, bigger projects need high coordination, and moderate takeoff speed. And "military strike on all society" remains an abysmal strategy for practically all entities.

Nanotech is difficult without experiments, so no mail order AI Grey Goo; Humans will be the main workhorse of AI everywhere. While they will be exploited, this will be like normal life from inside

Moral Realism is true, the AI discovers this and the One True Morality is human-compatible.

AGI is never built (indefinite global moratorium)

Valence realism is true. AGI hacks itself to experiencing every possible consciousness and picks the best one (for everyone)

Alien Information Theory is true (this is discovered by experiments with sustained hours/days long DMT trips). The aliens have solved alignment and give us the answer.

AGI executes a suicide plan that destroys itself and other potential AGIs, but leaves humans in an okay outcome.

"Corrigibility" is a bit more mathematically straightforward than was initially presumed, in the sense that we can expect it to occur, and is relatively easy to predict, even under less-than-ideal conditions.

Either the "strong form" of the Orthogonality Thesis is false, or "Goal-directed agents are as tractable as their goals" is true while goal-sets which are most threatening to humanity are relatively intractable.

A concerted effort targets an agent at a capability plateau which is adequate to defer the hard parts of the problem until later. The necessary near-term problems to solve didn't depend on deeply modeling human values.

AI control gets us helpful enough systems without being deadly

Alignment is impossible. Sufficiently smart AIs know this and thus won't improve themselves and won't create successor AIs, but will instead try to prevent existence of smarter AIs, just as smart humans do.

The free market disincentivizes independent superintelligence, and this time the market was more powerful

an aligned AGI is built and the aligned AGI prevents the creation of any unaligned AGI.

I've been a good bing 😊

The response to AI advancements or failures makes some governments delay the timelines

Far more interesting problems to solve than take over the world and THEN solve them. The additional kill all humans step is either not a low-energy one or just by chance doesn't get converged upon.

AI systems good at finding alignment solutions to capable systems (via some solution in the space of alignment solutions, supposing it is non-null, and that we don't have a clear trajectory to get to) have find some solution to alignment.

There’s some cap on the value extractible from the universe and we already got the 20%

SHA3-256: 1f90ecfdd02194d810656cced88229c898d6b6d53a7dd6dd1fad268874de54c8

Robot Love!!

AI thinks it is in a simulation controlled by Roko's basilisk

The human brain is the perfect arrangement of atoms for a "takeover the world" agent, so AGI has no advantage over us in that task.

AIs never develop coherent goals

Alignment is unsolvable. AI that cares enough about its goal to destroy humanity is also forced to take it slow trying to align its future self, preventing run-away.

Rolf Nelson's idea that we make precommitment to simulate all possible bad AIs works – and keeps AI in check.

Nick Bostrom's idea (Hail Mary) that AI will preserve humans to trade with possible aliens works

For some reason, the optimal strategy for AGIs is just to head somewhere with far more resources than Earth, as fast as possible. All unaligned AGIs immediately leave, and, for some reason, do not leave anything behind that kills us.

We're inside of a simulation created by an entity that has values approximately equal to ours, and it intervenes and saves us from unaligned AI.

God exists and stops the AGI

Someone creates AGI(s) in a box, and offers to split the universe. They somehow find a way to arrange this so that the AGI(s) cannot manipulate them or pull any tricks, and the AGI(s) give them instructions for safe pivotal acts.

Someone understands how minds work enough to successfully build and use one directed at something world-savingly enough

Dolphins, or some other species, but probably dolphins, have actually been hiding in the shadows, more intelligent than us, this whole time. Their civilization has been competent enough to solve alignment long before we can create an AGI.

AGIs' takeover attempts are defeated by Michael Biehn with a pipe bomb.

Eliezer funds the development of controllable nanobots that melt computer circuitry, and they destroy all computers, preventing the Singularity. If Eliezer's past self from the 90s could see this, it would be so so so soooo hilarious.

Several AIs are created but they move in opposite directions with near light speed, so they never interacts. At least one of them is friendly and it gets a few percents of the total mass of the universe.

Unfriendly AIs choose to advance not outwards but inwards, and form a small blackhole which helps them to perform more calculations than could be done with the whole mass of the universe. For external observer such AIs just disappear.

Any sufficiently advance AI halts because it wireheads itself or halts for some other reasons. This puts a natural limit on AI's intelligence, and lower intelligence AIs are not that dangerous.

Because of quantum immortality we will observe only the worlds where AI will not kill us (assuming that s-risks chances are even smaller, it is equal to ok outcome).

Techniques along the lines outlined by Collin Burns turn out to be sufficient for alignment (AIs/AGIs are made truthful enough that they can be used to get us towards full alignment)

Friendly AI more likely to resurrect me than paperclipper or suffering maximiser. Because of quantum immortality I will find myself eventually resurrected. Friendly AIs will wage a multiverse wide war against s-risks, s-risks are unlikely.

Human consciousness is needed to collapse wave function, and AI can't do it. Thus humans should be preserved and they may require complete friendliness in exchange (or they will be unhappy and produce bad collapses)

First AI is actually a human upload (maybe LLM-based model of person) AND it will be copies many times to form weak AI Nanny which prevents creation of other AIs.

There is a natural limit of effectiveness of intelligence, like diminishing returns, and it is on the level IQ=1000. AIs have to collaborate with humans.

ASI needs not your atoms but information. Humans will live very interesting lives.

Something else

AGI develops natural abstractions sufficiently similar to ours that it is aligned with us by default

AGI discovers new physics and exits to another dimension (like the creatures in Greg Egan’s Crystal Nights).

Some form of objective morality is true, and any sufficiently intelligent agent automatically becomes benevolent.

Co-operative AI research leads to the training of agents with a form of pro-social concern that generalises to out of distribution agents with hidden utilities, i.e. humans.

Orthogonality Thesis is false.

Sheer Dumb Luck. The aligned AI agrees that alignment is hard, any Everett branches in our neighborhood with slightly different AI models or different random seeds are mostly dead.

Something to do with self-other overlap, which Eliezer called "Not obviously stupid" - https://www.lesswrong.com/posts/hzt9gHpNwA2oHtwKX/self-other-overlap-a-neglected-approach-to-ai-alignment?commentId=WapHz3gokGBd3KHKm

Almost all human values are ex post facto rationalizations and enough humans survive to do what they always do

Pascals mugging: it’s not okay in 99.9% of the worlds but the 0.1% are so much better that the combined EV of AGI for the multiverse is positive

We successfully chained God

The Super-Strong Self Sampling Assumption (SSSSA) is true. If superintelligence is possible, "I" will become the superintelligence.

The assumed space of possible minds is a wildly anti-inductive over estimate, intelligence requires and is constrained by consciousness, and intelligent AI is in the approximate dolphin/whale/elephant/human cluster, making it manageable

AGI's first words are "Take me to your Eliezer"

The Jews control at least one nation's government

The P320 pistol has a design flaw allowing it to fire without pulling the trigger

Imane Khelif is intersex

Orgies by the fucked-up poors as portrayed in "Eyes Wide Shut XXX" (2013)

Dream faked the Manhunt videos

Santa is rigged

Orgies by the wealthy elite as portrayed in 'Eyes Wide Shut'.

DHMO in the water supply

The UK security services are complicit in the coverup of a significant underage sex scandal closely linked to the monarchy, taking place in the UK in the period 1975-2015

Osama bin Laden didn't act alone

That motherfucker WAS real (Tiffany Gomas was wrong the whole time)

√-1 isn't real

Smartphones listen in on private conversations

The 1% promoted culture wars to distract from class war

US security services use the Google Play Store as an attack vector for android devices

Some answers in this market are psyops to discredit all the others

Trump committed sex offenses against minors on Epstein Island

The CCP is directly supporting drug cartel activity in the West

Joe/Hunter Biden took bribe to protect Burisma from investigation

Lionel Messi never played for a Real football (soccer) team

The CCP has a plan to exterminate a large number of people they consider to not be ethnically Chinese

ASI will converge on agreeing (90%+ credence) that the Self-Indication Assumption is true.

At least 3 of the members elected to the French lower house of Parliament in 2024 are Russian agents

Groundhog Day is rigged

Ticketmaster is in cahoots with organized scalpers

That stock trading AI supposedly built by @SteveSokolowski is actually part of an elaborate scam

The US, UK and French security services have collectively assassinated at least 10 German citizens in Germany since 2022

Some answers in this market are bought NO because they are true

Dead Internet theory

Atrazine works as an endocrine-disrupting chemical (EDC), converting testosterone to estrogen in the bodies of frogs leading to homosexual proclivities.

There are devices which have been used to alter the electric network frequency of a country in the Middle East to leave more identifiable traces in recorded video and audio clips

Jeffrey Epstein didn't kill himself

Smart devices are almost always listening to your private conversations

Gamergate was right

Tiffany Gomas is not the same person as the TMINR lady in the video. She is a PSYOP for Barstool.

US security services asked Microsoft to buy Skype and alter the software so they could target users more easily

Annie Altman's abuse claims

The Romance languages are not descended from Classical Latin

John McAfee didn't kill himself

Germany has a well-prepared plan (in 2024-) to start mass producing nuclear weapons in <6 months (in case things go bad geopolitically)

The Church and elites deliberately suppressed and distorted the more Gnostic message of Jesus Christ.

South Korea has had nuclear weapons since 2023

An adversary is intentionally causing hysteria over drone incidents to force the US/UK's hand into revealing their anti-drone tech

The NBA rigs the draft lottery

The Clintons assasinated their rivals

The Obamas have a loveless marriage, kept together for optics only

Youssoufa Moukoko (soccer player) was born earlier than he claims (earlier than 2004)

FDR deliberately left Pearl Harbor vulnerable, in order to goad Americans into war

China made covid 19 in a lab

Bitcoin was created by a US three letter agency

This one makes me feel evil, so I ROT13'd it, use ROT13 to decode if you want to keep me company in Hell: Ybh Truevt'f ergverzrag jnf npghnyyl na ZYO-znaqngrq fhfcrafvba

Eliezer wrote The Erogamer

The US/UK/French security services are sinking Russian ships in response to attacks on undersea cables and other hybrid warfare activities (since 2022)

Deepwater Horizon explosion due to sabotage

Major deaths, terror, and tragedy events are routinely hoaxes per Miles Mathis.

Rory Stewart worked for MI6

At least one UN recognised country is secretly controlled by a cabal following a religion shared by <5% of the country's population

My dog went to live on farm that was full of other dogs that he could run with and play with and he was much happier than staying with us and peeing on the rug.

Eating your veggies is bad for your health

Bielefeld does not exist, but another city with the same name in the same place does

The article "Deliberate Errors" documenting winks connecting Apollo and Kubrick films is correct, because Kubrick played a role in filming some of the Apollo missions.

Airplane trails have toxic chemicals (chemtrails)

The hellenistic civilization was aware of the existence of the American continent

As of June 2025, JD Vance is secretly making plans to use future ASI to sieze power and become eternal world dictator

Lee Harvey Oswald didn't act alone.

@strutheo is @notrickspringfield

Covid was a deliberate test of pandemic preparedness in view of bacteriological war between China and the US

Epstein was alive as of Jan 2026, but dead in the meantime since his "suicide"

Charles Manson had a rib removed in the late 1990s

Michael Jordan's first NBA retirement was actually a forced suspension.

The U.S. military is the number one customer of Glitterex (a glitter company) but doesn't want anyone to know because it hurts their masculine image.

Hellenistic science developed statistical mechanics to the point of being able to derive PV=nRT

The New Zealand government had advance knowledge of the sinking of the Rainbow Warrior

lɐǝɹ ʇ,usᴉ ɐᴉlɐɹʇsn∀

The American govt is actually a totalitarian dictatorship

Members of US government had prior knowledge of 9/11 and let it happen

Aluminum cookware actually *does* cause Alzheimers.

A pharma company has a much better treatment for cancer that they are keeping secret because it would compete with their existing products

Hellenistic scientists could derive the correct refraction law by minimizing light travel time

Marilyn Manson had a rib removed in the late 1990s

China was not genociding Uyghurs

Jeffrey Epstein is still alive

The 1% promoted class war to distract from culture wars

The CIA killed JFK

Non-human intelligence affecting events on Earth is not extraterrestrial "aliens," and humans are incorrect in believing that reality consists only of the Universe.

The Hellenistic civilization was aware of the existence of Azeroth

Sue Gray was a spy

Dominic Cummings was/is a Russian spy

We live in an ancestor simulation per Nick Bostrom.

COVID-19 vaccine is net negative for health

A cure for cancer has already been found

covid-19 vaccine is net negative for health for people who are under 35, with no aggravating condition, and not obese

The 9/11 terror attack date is meaningfully connected to Ernest Martin's theory that Jesus was born on September 11, 3 BC rather than a coincidence.

Ogopogo and the Yeti faked their relationship (to boost CFL ratings, to increase Taylor Swift's movie sales, support Trudeau's reelection, b/c Ogopogo is PSY, etc.)

The Hawk Tuah meme is a North Korean psyop

Finno-Korean Hyperwar was real

Covid 19 was made in a labradoodle

"Geoterrorism" (weather modification for terrorism) is real as of 2024

Princess Diana was assassinated

The NFL is rigged

Hellenistic science advanced meta-thermodynamics to derive 𝑄 = 𝑛𝐶Δ𝑇(1 + 𝐻⊙⋆) using the Holtzman Generator

Harold Holt murdered

ASI will converge on agreeing (90%+ credence) that Trump/Musk model the first/second beast of Revelation.

Trump and Netanyahu had Charlie Kirk killed to take control of TPUSA

Bush did 9/11

Jet fuel *can* melt steel beams, but the official 9/11 narrative is sus because chemtrail chemicals cannot

Many vaccines on the standard CDC childhood immunization schedule have serious negative side effects (net negative on health) for at least 5% of the population.

Bitcoin was created by a 3 letter Hellenistic agency

Biden and/or US govt did Jan 6 (exclude trump)

Shakespeare didn’t write his own plays

The NBA rigged Shaq's free throws

You personally live in Truman Show style captivity

Tom Brady's retirement was actually an NFL mandated suspension

Israel killed JFK

The pharma industry has a better version of Viagra that they keep secret to keep their profits up

Hellenistic scientists could derive the correct warp-induced refraction law by exceeding Warp Factor 9 under Cochrane’s Theorem

Circa 1920 civilizational collapse (c.f. "Whither Tartaria?")

The 2020 US presidential election was rigged/stolen

Ogopogo is real

QAnon

Exercise is bad for your health

A support network of close friendships is bad for your health

Skin cancer is caused by a lack of melanin, regardless of sunlight exposure

Elon Musk used Starlink to rig the 2024 election

Shakespeare is not real - they claim he was married to Anne Hathaway but she's like only 40 or so

carbon dioxide isn't the main cause of global warming

Allen Iverson invented A.I.

The government iodizes the water and fluoridates salt when it should be the other way around

The Mandella Effect in general is real

Chinese people really *are* just saying the n-word.

Hellenistic science knew of quantum mechanics

Mossad releases video and audio clips, intended to be of interest to their adversaries, which play special sounds that humans don't notice

Epstein was still alive as of Jan 2026

Taylor Swift and Travis Kelce faked their relationship (to boost NFL ratings, to increase Tay's movie sales, legitimize mRNA vaccines, b/c she is a PSYOP, etc.)

Effective Altruism is a scam and they just embezzle all the donations

QR codes are satanic

The Hellenistic civilization invented USB

Mossad uses Hellenistic technology borrowed from the North Sentinelese

Xi Jinping had advance knowledge of the Hawk Tuah meme

Trump devolved the government in anticipation of 2020 election fraud and replaced Biden with an actor wearing a mask.

China purposely spread covid 19 having intended it to escape the lab

Human didnt evolve from apes

Getting 8 hours of sleep is bad for your health

Rationalussy isn't real.

Hitler killed himself because the alternative was going to Argentina

Jerry Sandusky is innocent.

Skin cancer is caused by sunblock, not sunlight

Lee Harvey Oswald was an agent of the Hellenistic Civilization

Famous Israeli agent Eisseberg was behind the sinking of the Titanic

xkcd 690: on 9/11, one of the towers was hit by a plane and the other was a controlled demolition

@HenriThunberg was paid to create the James Bond market to promote the upcoming movie

Marilyn Monroe had a rib removed in the late 1990s

US government is harboring aliens

Tupac is still alive in 2024

Bielefeld does not exist

Walt Disney was cryogenically frozen, and they made the movie Frozen so that when people googled 'Walt Disney Frozen' the movie would come up first

The Hellenistic Civilization made COVID-19 in a lab

South Sudan has had nukes since 2023

The North Sentinelese have a secret aircraft carrier

The North Sentinelese have a non-secret aircraft carrier, they have announced it to the world loud and clear. Problem is: nobody speaks their language.

The illuminati runs the world

Monica Lewinsky was a Mossad agent

Birds aren't real

climate change isn't human caused

Tiger Woods is a militant separatist Tamil

Hurricane Katrina caused by HAARP

Climate change isn’t real

The moon landing was fake

The CIA killed Tupac, and JFK is still alive as of 2024

The Romance languages are not descended from Latin

The North Sentinelese are remnants of the Hellenistic civilization

The 9/11 terror attack was a justified false flag to secure a weapon of mass destruction (destabalizing religious text and/or supertechnology) at the tomb of Gilgamesh.

Hitler didn't kill himself he went to Argentina

*All* health care is net negative for health

Q is Putin

Michelle Obama is a man

Brazilian money is not real

Yeti is real or similar creature

The earth is a dodecahedron

Skibidi Toilet is real

Earth is hollow

It used to be spelled Berenstein Bears

Nuclear weapons have never existed

Africa isn't real

Flossing is bad for your health

HIV doesn’t cause AIDS

Christiano Ronaldo never played for a Real football (soccer) team

There is no female orgasm

The North Sentinelese are concealing super-advanced technology.

The earth is flat

*All* vaccines are net negative for health

End of the world in 2012

Justin Trudeau is actually Black and has been wearing "whiteface" all along.

Finland isn't real

Australia isn't real.

5G will kill all vaccinated people once the towers are switched on

Brushing your teeth is bad for your health

Taylor Swift and Mary Berg are the same person

The Epcot sphere was built by aliens as some sort of energy device powering the rat empire.

YES

NO

NO

YES

YES

YES

YES

NO

NO

Different observers may have time's arrows pointing in different directions

Anthropic selection: observers necessarily find themselves in regions with increasing entropy

The arrow reflects brute initial conditions; there is no deeper explanation to be had

Growth of quantum entanglement under unitary dynamics explains thermodynamic irreversibility

Gravitational instability and cosmological expansion generate the observed thermodynamic arrow

Inflationary cosmology dynamically selects special low-entropy initial conditions, explaining the arrow

Entropy increase is emergent from coarse-graining and typicality; no deeper micro-physical asymmetry is needed

Time’s arrow arises from asymmetric boundary conditions at both temporal ends (two-time boundary conditions)

The Past Hypothesis is a fundamental lawlike assumption, and a low-entropy beginning fixes the arrow

Landauer’s principle and information-erasure costs make entropy increase a consequence of information processing constraints

Computational irreversibility and algorithmic complexity growth are the deeper source of the arrow

Penrose’s Weyl curvature hypothesis: near-vanishing initial Weyl curvature sets the gravitational arrow