Who Calls News

Other

Elon Musk

Bill Gates

Jeff Bezos

Mr. Beast

Barack Obama

Gwern Branwen

Donald Trump

Satoshi Nakomoto

Robert Caro

Kamala Harris

Jensen Huang

Sarah Paine

Sam Altman

Narendra Modi

Joe Biden

JD Vance

roon

Geoffrey Hinton

The Mountain (Icelandic strongman)

Terence Tao

Ilya Sutskever

No one, the tweet was a joke or intended to attract advertisers without referring to someone specific

Tim Cook

Satya Nadella

Justin Trudeau

Brian Shaw (_biggest_ guest yet?)

Xi Jinping

Dalai Lama

Buffett

Taylor Swift

Lebron James

GPT-5

Bill Clinton

A basketball player

Vladimir Putin

Benjamin Netanyahu

Jesus Christ

Rona Wang

Jose Luis ricon

Kettner Griswold

James Koppel

Sam Bankman-Fried

Nancy Pelosi

Rishi Sunak

Keir Starmer

Volodymyr Zelenskyy

Jimmy Carter

George W Bush

Al Gore

Michael Jackson

your mom

Mitt Romney

one or both of his parents

A new OpenAI AI model not called "GPT-5"

Connor Duffy

MBS

pope Francis

Scott Alexander

Leonardo DiCaprio

RFK Jr

Sydney Sweeney

growing_daniel

greg16676935420

Deadpool

Shaq

Oprah Winfrey

Yoshua Bengio

Sundar Pichai

Scarlett Johansson

Paul McCartney

[duplicate]

Neel Nanda

King Charles

Kim Jong Un

Gavin Newsom

Royal Palace

[cancelled option]

Daniel Yergin

Peter Singer

Gabe Newell

Neil Gorsuch

Stephen Breyer

Dmitry Medvedev

JK Rowling

Shrek

Sam Hyde

[invalid answer] Multiple people e.g. a team from OpenAI

Peter Thiel

Marques Brownlee

Vivek Ramaswamy

Donald Trump Jr.

Ben Shindel

Javier Milei

Dylan Patel

Joe Rogan

Marc Andreessen

Mike Tyson

Jake Paul

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.

Other

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

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.

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.

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

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

Humans become transhuman through other means before AGI happens

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 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

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.

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.

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

AGI is never built (indefinite global moratorium)

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

"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.

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

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.

Ethics turns out to be a precondition of superintelligence

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

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'

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

I've been a good bing 😊

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.

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.

Something less inscrutable than matrices works fast enough

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.

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

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.

Aliens invade and stop bad |AI from appearing

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 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 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)

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

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

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.

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

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

Something else

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

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).

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.

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.

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"

Assassin's Creed Shadows

Red Dead Redemption 2

Clair Obscur: Expedition 33

Halo: The Master Chief Collection

Call of Duty: Black Ops 7

Assassin's Creed Mirage

Microsoft Flight Simulator 2024

Crimson Desert

Grand Theft Auto VI

Grand Theft Auto IV

Metal Gear Solid Delta: Snake Eater

Tom Clancy's Rainbow Six Siege

Tom Clancy's The Division

It won't be available for purchase in 2024

Tracks sleep

It won’t have a USB-C port

will support wireless charging

Improved health tracking if you pair it with an apple watch (ie better data if you wear both than either by itself)

It will only be usable with an iPhone

It will be available in a range of colours

It’ll measure blood oxygen saturation / include a pulse oximetry function.

Will support gestures

At launch, the MSRP will be <$500.

Tracks Stress

The ring design will be asymmetrical.

It will eventually be tied up in litigation like the pulse-ox sensor on the watch

There will be a gold version (real gold, not just gold colored)

battery lasts over a week

It will have a physical button on it

It will have a glass or crystal screen

There will be a higher memory model sold for more money at launch (think 128MB vs. 256MB models)

Waterproof up to 100 meters / 330 feet

It will be called the iRing

It will have a camera

It will be announced in 2025

It will be available for purchase in 2024

A government of a major nation orders the shutdown of a significant AI service

Reports of activity on internet-connected servers

AI-related cyberattack

AI-designed cyberweapon

AI system solves major unsolved math problem

Mass AI-driven job displacement event

A government declaration/statement (any country)

Reports about software

Reports about financial activity

Report about sociological observations

Report about economic observations

AI-related academic achievement

Attack on military target

Attack on civilian target

AI-related weapon announcement/use/threat

Major AI safety incident or accident

A statement from an individual (human)

AI-related fake news

AI-related political movement (pro or anti)

A private company declaration/statement (any company)

Discovery of a spy / mole working for foreign power (or for an AI)

Military activity / posture change

An AI-related apocalypse cult attacks a civilian or military target

AI-related mass psychosis/hysteria event

An AI uncovers evidence relating to a previous event

AI makes prediction about future event

Open source model release

Release of a closed-source/closed-weights model

Reports of aerial devices/machines

Reports of market activity

AI-related corruption scandal

AI-related theft

A new war between nation states

AI-related resignation

Reports of activity in online communities

Reports of activity in open source software

A declaration/statement from a military / intelligence agency

Reports of activity on internet media distribution platforms

People receiving messages (text / phone calls / whatsapp / ...)

AI-related diplomacy

Reports about consumer activity

Reports about identity theft

Reports of activity in religious communities

AI-related media (movie/song/book ...)

Reports about industrial activity

Reports about physical machines

Reports about criminal activity

AI-related competition achievement

A viral meme

An announcement from an AI lab similar to o3

AI-related Internet shutdown in some country

Conventional weapons attack on AI infrastructure / supply chain

A declaration/statement from an AI (any AI)

AI-related sex scandal

A piece of viral AI-generated media that has a strong/unexpected effect on large numbers of people (~psychological)

AI parasitism / Addictive-Persuasive Agent

Attack on Taiwan

AI-related terrorist attack

AI-related assassination

AI-designed pathogen

AI drug discovery

AI material discovery

Death of an individual in suspicious circumstances

AI-related astronomical event / observation / analysis (can also include satellites)

Reports of underwater activity

Reports of activity on blockchain networks

Reports of activity in academic communities

Publication of a research paper (or pre-print / blog post / poster / twitter thread / ... -- about research-related topic)

Reports of activity in online video games

Reports about people getting scammed

AI-related cult

AI system demonstrates general robotic control

AI causes significant stock market event

Major AGI lab whistleblower revelation

AI-related archaeology

AI system makes scientific breakthrough

An AI lab demonstrates an automated AI research engineer

Donald Trump

JD Vance

Mike Johnson

Ted Cruz

Dave McCormick

Rick Scott

Josh Hawley

John Thune

John Cornyn

Rand Paul

Trump launches new strikes on Iran before any deal is signed

Trump posts calling a deal "the greatest" before it's actually signed

Trump explicitly says he deserves the Nobel Peace Prize for it

Trump declares the MOU "over" again before a final deal

Iran charges tolls or fees on Strait of Hormuz shipping

Multinational body created to govern the Strait

Jared Kushner participates in talks

Iran agrees to IAEA inspector access to nuclear sites

Formal reconstruction fund for Iran is established

US lifts any sanctions on Iran before 2027

Israel publicly endorses a US-Iran agreement

Strait of Hormuz fully reopened to commercial shipping by Oct 1

Iran agrees to zero uranium enrichment

New Iranian Supreme Leader is named before a deal is finalized

The Joe Rogan Experience

Alex Cooper's Call Her Daddy

This Past Weekend w/ Theo Von

New York Times’ The Daily

The Tucker Carlson Show

Crime Junkie

Someone within the Trump administration

Russia

Ukraine

Another European country

A journalist or private signals intelligence enthusiast

Give substantial equity to the US government

Call for a pause or slowdown of frontier model development

Merge with another company that is worth at least 100B

Every 7-day period, I will use and carry around an Android device capable of making phone calls.

I will use a non-phone smart device more than an Android phone on at least one 7-day period.

Mortal Kombat II

Call of Duty