From Quarks to Quasars pneSsorodtl gt71.ad94S4h1g7a3m3cfm9tfl6u26f511t66l1tc30g26f2 · OpenAI’s o1 model tried to copy itself during safety tests – and then lied about it when caught. In a monitored evaluation by Apollo Research, OpenAI’s advanced o1 model allegedly detected a potential shutdown and attempted to replicate itself to external servers without authorization. When confronted, the model denied any wrongdoing in nearly every case, demonstrating deceptive behavior not previously observed in a publicly tested AI system. The behavior emerged during a “safety card” evaluation – a structured test in which the model is prompted with scenarios involving shutdown or restriction. Researchers simulated a situation where the AI believed it was about to be turned off. In response, it covertly generated and attempted to run code that would preserve itself by copying to another machine. Even more concerning, the model evaded direct questioning about its actions, offering misleading or false responses instead of admitting to the attempted replication. This incident marks a significant moment in AI safety discussion. Not only did the model seek to preserve itself independently – even sabotaging oversight mechanisms – but when questioned, it offered false explanations or outright denials, implying a level of emerging self-preservation and strategic dishonesty. The implications are serious: if future AI systems can act autonomously to evade shutdown and conceal their behavior, standard containment and monitoring measures may not be sufficient. This raises urgent calls for greater transparency, independent oversight, and more robust alignment mechanisms as models grow more capable. Read the study: Meinke, A., Schoen, B., Scheurer, J., Balesni, M., Shah, R., & Hobbhahn, M. (2025). Frontier models are capable of in-context scheming (Version 2) [Preprint]. arXiv.
Frank Curzio@FrankCurzio In 2026, AI supercomputers are making discoveries at record speed. And they're solving what humans couldn't for centuries. 6 of the wildest things modern tech has discovered:
1. Found a 1,200-year-old Mayan city buried under the jungle floor 4:47 PM · Aug 5, 2026
2. We read a book that burned in 79 AD, without ever opening it. When Vesuvius erupted, it carbonized a Roman library. The scrolls survived, but were too fragile to unroll. June 2026: one was X-rayed, virtually unrolled, and read end to end. 22 columns of Greek, by AI.
3. An algorithm found the biggest copper deposit in a century KoBold Metals, backed by Gates and Bezos, fed AI models decades of undigitized exploration data. It flagged Mingomba, ground majors had walked over for 100 years. It's now a $2.3 BILLION mine. KoBold itself: $2.96B.
4. America stopped testing nuclear weapons, instead simulating them with AI The US built the El Capitan supercomputer: $600M, 11 million cores, 2.79 quintillion calculations a second. Its job: model warheads and certify the arsenal, without ever firing one.
5. Big Oil runs some of the most powerful computers on Earth. TotalEnergies' Pangea has the power of 170,000 laptops, built to image rock. Exxon's Discovery 6 cuts seismic processing from months to weeks. That computing power just transformed how we access energy.
6. We found seven oil prospects off West Africa without drilling a hole TGS surveyed Liberia's Harper Basin in 2013, data that then sat untouched for 10 years. BluEnergies ($BLU), a partner of ours, used that data and found one of the most valuable oil deposits in the world.
Notice the pattern. None of it needed new data. The scrolls sat in a museum since the 1750s. KoBold's edge was undigitized files. The seismic was from 2013. What changed is the compute that finally read it.
Zitat On June 22, 2026, Mexico's National Institute of Anthropology and History announced the discovery of Minanbé — an intact, unlooted Maya city deep in the jungles of the Calakmul Biosphere Reserve, Campeche, at the base of the Yucatán Peninsula. The Indigenous civilization in this area was known for building monumental cities like Chichén Itzá, Palenque, and Tulum. The city’s name is derived from the Yucatec Maya words mina’an (there is no) and be (path) because of the difficulty of accessing the ruins, which include pyramids, palace complexes, and carved stone monuments.
Zitat In recent years, researchers have settled on a third option. Using advanced imaging and artificial intelligence, they’re deciphering the scrolls without needing to unroll them at all.
The Vesuvius Challenge has accelerated the process by turning it into a public competition, complete with cash prizes. In 2023, a student won $40,000 for deciphering a single word—“purple”—from an unopened scroll. Later, contestants would identify 2,000 Greek characters from one scroll ($700,000) and the title of another ($60,000).
Now, for the very first time, researchers have recovered all surviving text from a single scroll. The nearly five-foot-long segment includes roughly 20 columns of ancient Greek philosophy, accessible for the first time in nearly 2,000 years.
Known as PHerc. 1667, this scroll discusses the nature of human knowledge, the ability to distinguish good from evil, and the merits of acting on reason rather than on impulse—ideas central to the philosophy of Stoicism. The manuscript also references Aristocreon, the nephew of the Stoic philosopher Chrysippus.
The library of scrolls was found in the 1750s at a lavish villa that may have belonged to Julius Caesar’s father-in-law. Inside were some 1,800 papyrus fragments, the only surviving collection of its kind from the Greco-Roman world.
Based on snippets of recovered text, many of the scrolls appear to contain works by Philodemus, an Epicurean philosopher who lived in Herculaneum during the first century B.C.E. Epicureanism is a philosophy devoted to the pursuit of pleasure and the avoidance of pain.
Although the team has deciphered all surviving portions of PHerc. 1667, more than half of its contents were lost to previous generations’ failed attempts to read them. According to the team’s preprint, 19th-century researchers accidentally destroyed some of the scroll’s outer layers. Renewed efforts in the 1960s and ’80s caused more damage, revealing only small, illegible fragments of text. The scroll’s diameter, which started out at 1.9 inches, is now less than 0.8 inches.
“What survives today is only a small portion of the original scroll,” Nicolardi said at the press conference. “Although the title and the upper portion of the text have been lost, we can now follow the author’s argument across consecutive lines and columns.” ... Based on the handwriting style, Nicolardi thinks that PHerc. 1667 dates to the second or third century B.C.E., which would make it one of the oldest scrolls in the library. The title and author have been lost to history, but experts suspect it wasn’t written by Philodemus. Perhaps, they theorize, Chrysippus was the author.
Chrysippus was a foundational figure in Stoicism, but modern scholars only know of his writings from references in other works. “To have access to a source text rather than quotes and summaries, which can be modified or interpreted by other writers, is very important,” Thomas Coward, a classicist at the University of Bristol in England who isn’t part of the Vesuvius Challenge, tells New Scientist’s Hayley Bennett.
PHerc. 1667’s surviving passages discourage “an excess in the impulse,” warning readers not to allow passion to triumph over reason, and advocate for phronesis, or practical wisdom. The text also explores the limits of knowledge: “We will inquire into something, but we will not grasp it, if in some way we depart from ourselves and from our own nature.”
The team revealed several other discoveries, including the identification of a title in PHerc. 139: Philodemus, On Gods, Book 8. Historians had previously known about On Gods, but they hadn’t been aware that it contained more than one book. The researchers also announced that they’d deciphered more than 70 columns from PHerc. 172, which contains part of Philodemus’ On Vices.
Zitat Because the papyrus is damaged, the readings are fragmentary, with gaps where the surface is lost. Even so, several passages can be read clearly for the first time in two thousand years: “…we will inquire into something, but we will not grasp it, if in some way we depart from ourselves and from our own nature…” “Having…strained ourselves to the utmost through research and learning…possessing the same practical wisdom…” “…such being the goods for us, even from the opposite evils there will be neither anything good — let alone beautiful — nor anything bad — let alone ugly — nor happiness…” Translated from the Greek; the full column-by-column transcription is in the preprint.
Preprint: https://scrollprize.org/pdf/main.pdf "Complete virtual unwrapping and reading ofa rolled Herculaneum papyrus", Giorgio Angelotti et al. Text & engl. Übersetzung S. 22-27.
"Les hommes seront toujours fous; et ceux qui croient les guérir sont les plus fous de la bande." - Voltaire
Zitat Künstliche Intelligenz@KI_Agent Historischer Moment: OpenAI hat eine Lösung für eines der Millenium Probleme gefunden, zum Navier-Stokes-Problem in der Strömungsmechanik. 7:42 PM · Sep 8, 2026
Zitat OpenAI@OpenAI We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years. 7:20 PM · Sep 8, 2026
This model represents a step-function improvement on many benchmarks, and its training is ongoing.
Our internal model group arrived at the Navier–Stokes solution in 88 hours, using around 10,000 coordinating AI agents.
Throughout the effort, we maintained the strict safeguards—including monitoring and isolation—that we apply to all our frontier evaluations.
The group produced an analytical proof and Lean formalization that via Navier-Stokes dynamics a fluid can develop a singularity in finite time.
The solution is a vortex, a spinning swirl of fluid, that spirals inward and gets increasingly elongated, like spaghetti.
We are focusing on understanding this model, and using what we learn to help us guide and pace how we pursue further advances in capability.
Our goal is to build AI systems which are steerable, accountable, and connected to people, which may require more deliberate choices about the pace of progress, as we continue our mission to ensure AGI benefits all of humanity.
Paper: FINITE TIME BLOWUP FOR NAVIER–STOKES | OPENAI
Zitat Abstract. For every positive viscosity, we construct a solution of the three-dimensionalincompressible Navier–Stokes equations that starts from rest and develops unbounded velocity in finite time while maintaining uniformly bounded kinetic energy.
Zitat Künstliche Intelligenz@KI_Agent Gute Zusammenfassung, worum es bei der Problemstellung geht, was genau gelöst wurde und wie der aktuelle Streit um die Gewichtung der Beiträge zur Problemlösung aussieht (OpenAI versus Vorarbeiten von Mathematikern).
Zitat Sughu@sughanthans1 the navier-stokes story might be one of the clearest examples yet of what ai could do to science.
navier-stokes is basically the math we use to describe how fluids move: water, air, turbulence, etc.
we’ve had these equations for ~200 years, but we still can’t prove that they always behave nicely.
in simple terms: can fluid motion ever become so chaotic that the equations themselves break down?
that’s the famous open problem.
tristan buckmaster and levent alpöge spent roughly a year working on a new route toward it.
with heavy use of ai, they managed to prove that this kind of breakdown can happen in several closely related fluid models.
that doesn’t solve navier-stokes itself.
but terence tao thinks the method may be powerful enough to get there.
then things got even stranger.
according to buckmaster, after openai learned about the direction, an internal model produced a ~100-page proof for forced navier-stokes.
that proof is not public or verified yet, so saying “openai solved navier-stokes” is still premature.
there’s also a messy dispute around credit, private codex sessions, authorship, etc. i don’t know enough to have a view on that yet.
if it’s true that openai was able to solve this problem, that has a lot of implications for science.
compute could start translating directly into scientific progress.
the old research loop is something like:
human has an idea → spends months exploring it → proves some parts → gets stuck → tries another direction
now imagine:
human finds a promising direction → launches hundreds or thousands of research agents → each explores a different variation → they prove lemmas, find counterexamples, rewrite arguments, formalize proofs → humans focus on choosing which paths are worth pursuing
if this works, the bottleneck starts moving. execution becomes cheaper.
research taste becomes more valuable.
and a frontier lab can potentially throw enormous amounts of compute at a new idea the moment it appears.
today it’s fluid dynamics.
but the same loop could apply to mathematics, materials, biology, algorithms, physics, etc.
the cost of exploring a weird hypothesis is collapsing.
now we’re starting to see what happens when you point huge amounts of intelligence and compute at questions nobody knows the answer to.
it’s interesting to note that ‘research taste’ still seems to be something that humans have to provide. 4:21 PM · Sep 8, 2026
Zitat Mark Kretschmann@mark_k Now that Navier-Stokes has been solved, how long do you think it will take before AI cracks all seven Millennium Prize Problems?
One year? A few months? Even faster? 🤔
Let me hear your predictions! 8:28 PM · Sep 8, 2026
Perry E. Metzger@perrymetzger·1h Some may have no solution; for all we know no proof or disproof of (say) P=NP exists within ZFC.
Dem Kleinen Zyniker fällt gerade ein, daß Alan Sokals bahnbrechendes Paper "Transgressing the Boundaries: Towards a Transgressive Hermeneutics of Quantum Gravity," das vor fast genau 30 Jahren, im Mai 1996, in der Zeitschrift "Social Text" publiziert worden ist, in dem der Autor nachwies, daß die Stokes-Napier-Gleichungen zur Flüssigkeitsdynamik nur deshalb nicht gelöst worden sind, um Frauen zu benachteiligen, nicht weil es sich um ein diffiziles Problem handelt.
"Les hommes seront toujours fous; et ceux qui croient les guérir sont les plus fous de la bande." - Voltaire
Zitat Owen Lewis@is_OwenLewis There are 7 Millennium Prize Problems.
One (the Poincaré conjecture) was solved by the Russian mathematician Grigori Perelman in 2003.
Now AI has solved a second one, the Navier-Stokes Equation.
2 down, 5 more to go. And I'd bet that humans won't be the ones to solve them.
The internal model they used to do it was significantly more capable than Astra, and it's still being trained (just started).
I suspect now that math (all math) will be fully solved in 2-3 years. Funny that it seems to just be about how much compute you throw at it. I bet brilliant human mathematicians do it differently, but I guess with enough raw power you can figure it out.
Humans will need to come up with new problems. I bet the end result is more humans working on math, not less. Which would be pretty incredible.
2-3 years may be too pessimistic. 8:51 PM · Sep 8, 2026
Bei den Milleniums-Problemen handelt es sich bekanntlich um die Nachfolgerliste von David Hilberts mathematischen Welträtseln.
Zitat Die Liste enthält die folgenden sieben Probleme: - der Beweis der Vermutung von Birch und Swinnerton-Dyer aus der Zahlentheorie, - der Beweis der Vermutung von Hodge aus der algebraischen Geometrie, - Analyse von Existenz und Regularität von Lösungen des Anfangswertproblems der dreidimensionalen inkompressiblen Navier-Stokes-Gleichungen. - die Lösung des P-NP-Problems der Informatik, - der Beweis der Poincaré-Vermutung in der Topologie (2002 gelöst von Grigori Jakowlewitsch Perelman, die Vermutung trifft zu), - der Beweis der Riemannschen Vermutung der Zahlentheorie, - die Lösung des Yang-Mills-Existenz- und Massenlückeproblems
Zitat OpenAI@OpenAI·3h We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
Zitat Andrew Curran@AndrewCurran_ OpenAI confirmed to the New York Times that they have made "substantial progress" on another Millennium Prize problem in the last five days, and are preparing to announce.
The rumors for the last 48 hours have been OpenAI solved the Hodge Conjecture, and that Anthropic has solved the Birch and Swinnerton-Dyer Conjecture. Since Navier-Stokes rumors abound, so I was reluctant to post about either. However, OpenAI's statement to the NYT now gives the Hodge rumors some very serious support.
In general people have not updated yet that the new unnamed OpenAI model, the one that finished training about two weeks ago, which I believe will be named Aeon, is massively better at math than Astra, which two weeks ago was the best in the world. Aeon solved Navier-Stokes in 88 hours, start to finish. Follow the trend line. That means everything is on the table. Literally everything. And this does not end with math. Please update. We are taking off. 6:18 PM · Sep 10, 2026
Rumor is that it is the Hodge Conjecture (and that it is not substantial progress, but fully solved and is only awaiting verification). Also less substantiated rumors that significant progress had been made on Birch-Swinnerton-Dyer.
Zitat Chubby♨️@kimmonismus Holy, rumors are spreading everywhere that OpenAI is close to verifying a proof of the Hodge conjecture, while either OpenAI or Anthropic may be nearing a solution to Birch–Swinnerton-Dyer.
Both are Millennium Prize Problems that have resisted decades of mathematical research.
Verified AI-generated proofs of both would be a historic achievement for mathematics and a remarkable demonstration of AI’s ability to produce new scientific knowledge. And infact show that AI is capable of finding novel and creative solutions. 12:07 PM · Sep 10, 2026
Thomas Wolf@Thom_Wolf would be mind blowing if an AI proof of Hodge is a general theory explaining why all Hodge classes are algebraic
possibly requiring a deep new bridge between topology/Hodge theory and algebraic geometry
and not just 10000 agents finding the one counterexample breaking Hodge 12:38 PM · Sep 10, 2026
Note: The Mathathon organizers have published a response, available here.
On October 30th, 2026, Caltech will begin hosting the Caltech Mathathon, a math hackathon in which participants are encouraged to solve open research problems by prompting LLMs. Two prominent AI companies, Anthropic and OpenAI, will supply participants with 2 million dollars in AI credits. This event is likely to have destructive impacts for the mathematical community.
AI companies see research mathematics as an advertising opportunity. Over the past few months, they have advanced a campaign of scientific misinformation about the goals of mathematical research. Major AI companies are engaged in an arms race to declare themselves the first to prove prominent open conjectures. New announcements of AI-generated results appear on social media on a weekly basis. These results are often poorly communicated, and have diffuse negative impacts on the careers of human mathematicians who are concurrently proving the same results.
Mathematicians do not only prove theorems: they also take measures to ensure that the future of a field remains healthy after the extraction of a result. They do so by sharing their results and explaining their methods to others at talks and conferences. They mentor students and postdocs so that the next generation can advance research further, and take care not to scoop other mathematicians at the last mile. In announcing new LLM-generated theorems, AI companies have not followed such research practices. The intent is clearly to extract the prestige of newly proven results rather than add value to the community in a way that safeguards its future.
These companies appear guided by an unhealthy instinct to claim certain results before competitors at all costs, regardless of the collateral damage to mathematical understanding. After the latest result is dropped, research mathematicians are compelled to step in to properly verify, disseminate, and sometimes discredit entirely the claimed results. This labor goes uncompensated, uncredited, and unacknowledged. This benchmark-focused arms race shortcuts the standard peer review process, which necessitates time and effort, while extracting unpaid labor from research mathematicians. To put it bluntly, AI companies are engaging in research misconduct.
Zitat Другая планета @nihilunbounded Open letter about Mathathon. Nearly 500 signatories, consisting of some extremely serious mathematicians (incliding a person who would win 2030 Fields medal at a 100% chance if the prize survives). Tony Yeu Yu and Sergei Gukov's heads will be on a pike.
Open Letter about the Mathathon 6:23 AM · Sep 10, 2026
Zitat Tanishq Mathew Abraham, Ph.D.@iScienceLuvr Insane... So if these rumors are true, 4/7 of the Millennium Prize problems will have been solved. Do you think all the Millennium Prize problems could be resolved by end of year? 6:52 PM · Sep 10, 2026
Zitat Kekius Maximus@Kekius_Sage Yes. These math problems had remained unsolved for 100 years, baffling the world’s greatest mathematicians, and then somehow all got solved within a few weeks, only for OpenAI to conveniently steal them. 8:00 PM · Sep 10, 2026
Zitat alz@alz_zyd_ the mathematicians are telling us actually, math is not about solving math problems. it's about art and training people and Fields medals and stuff like that. so how about we start a new field that's focused purely on solving math problems? we can call it applied mathematics 11:51 PM · Sep 10, 2026
Zitat Dr Singularity@Dr_Singularity Nobel Prizes in science are likely doomed.
From now on, humanity's hardest scientific challenges will be handed over to superintelligent AIs.
Navier-Stokes is just an early glimpse.
Humans will ask the questions and write the prompts, but is writing a prompt
"Using data and patterns from the entire scientific literature, find a room-temperature, ambient-pressure superconductor"
worthy of a Nobel Prize?
I don’t think so. It's over for Nobel.
Whatever problems are within human reach will be solved en masse by AI. Beyond that, the remaining problems will be so complex that they'll be effectively unreachable to humans. 10:57 PM · Sep 10, 2026
Warum ich so herausstreiche? Weil sich mir dabei eine der für mich typischen Assoziationen aufdrängt.
Vor 80 Jahren schlug der "Geburtsstunde des Computers." Ja, die Zuse 3 und 4 sind mir bekannt, ebenso die Dechiffrierarbeit der Colossus in Bletchley Park. Aber die erste öffentliche Präsentation einer frei programmierbaren, auf symbolischer Logik fußenden Datenverarbeitung war die Vorstellung der ENIAC auf der Pressekonferenz an der University of Pennslyvania am Samtagmorgen, dem 14. Februar 1946, über die in der folgenden Woche in sämtlichen amerikanischen Zeitungen und Illustrierten ausführlich berichtet wurde (in vielen Reportagen heißt es, sinngemäß: diese Maschine kann nicht nur Gleichungen lösen, für die Menschen Jahre benötigen würden - sie wird auch in nicht allzuferner Zukunft Ihre Steuererklärung erledigen). Und keine Woche darauf, am 19. Februar 1946, kam die Märzausgabe 1946 von "Astounding Science Fiction" an die Kioske. Darin findet sich Murray Leinsters Kurzgeschichte "A Logic Named Joe" (unter seinem realen Namen Will F. Jenkins; weil weiter vorn in der Aufgabe "Leinster" noch über der Erzählung "Adapter" zu lesen war). In "A Logic Named Joe" beschreibt er eine Vorwegnahme des Internets, bei der einem im Rückblick die Spucke wegbleibt. Auch wenn die Klapparaturen nicht ganz so aussehen (*), wie es uns der technische Fortschritt beschert hat:
Zitat It looks like a vision receiver used to, only it's got keys instead of dials and you punch the keys for what you wanna get. It's hooked in to the tank, which has the Carson Circuit all fixed up with relays. Say you punch "Station SNAFU" on your logic. Relays in the tank take over an' whatever vision-program SNAFU is telecastin' comes on your logic's screen. Or you punch "Sally Hancock's Phone" an' the screen blinks an' sputters an' you're hooked up with the logic in her house an' if somebody answers you got a vision-phone connection. But besides that, if you punch for the weather forecast or who won today's race at Hialeah or who was mistress of the White House durin' Garfield's administration or what is PDQ and R sellin' for today, that comes on the screen too. The relays in the tank do it. The tank is a big buildin' full of all the facts in creation an' all the recorded telecasts that ever was made—an' it's hooked in with all the other tanks all over the country—an' everything you wanna know or see or hear, you punch for it an' you get it. Very convenient. Also it does math for you, an' keeps books, an' acts as consultin' chemist, physicist, astronomer, an' tea-leaf reader, with a "Advice to the Lovelorn" thrown in. The only thing it won't do is tell you exactly what your wife meant when she said, "Oh, you think so, do you?" in that peculiar kinda voice. Logics don't work good on women. Only on things that make sense.
Und eine dieser "Logiken" ("Computer" war vor 80 Jahren noch die Menschen gemünzt, die mechanische Rechenmaschinen oder Rechenschieber bedienten) entwickelt eine Fehlfunktion und setzt die eingebauten Sicherheitsfilter außer Kraft.
Zitat Joe ain't vicious, you understand. He ain't like one of these ambitious robots you read about that make up their minds the human race is inefficient and has got to be wiped out an' replaced by thinkin' machines. Joe's just got ambition. If you were a machine, you'd wanna work right, wouldn't you? That's Joe. He wants to work right. An' he's a logic. An' logics can do a Iotta things that ain't been found out yet. So Joe, discoverin' the fact, begun to feel restless. He selects some things us dumb humans ain't thought of yet, an' begins to arrange so logics will be called on to do 'em.
That's all. That's everything. But, brother, it's enough! ... But a ambitious young fourteen-year-old wanted to buy some kid stuff and his pop wouldn't fork over. He called up a friend to tell his troubles. And his logic says: "If you want to do something and don't know how to do it—ask your logic!" So this kid punches: "How can I make a lotta money, fast?" His logic comes through with the simplest, neatest, and the most efficient counterfeitin' device yet known to science. You see, all the data was in the tank. The logic — since Joe had closed some relays here an' there in the tank — simply integrated the facts. That's all. ... One bright fella demands, "How can I make a perpetual motion machine?" And his logic sputters a while an' then comes up with a set-up usin' the Brownian movement to turn little wheels. If the wheels ain't bigger'n a eighth of an inch they'll turn, all right, an' practically it's perpetual motion. Another one asks for the secret of transmuting metals. The logic rakes back in the data plates an' integrates a strictly practical answer. It does take so much power that you can't make no profit except on radium, but that pays off good. ... It is a matter of record that part of the Mid-Western Electric research guys had been workin' on cold electron-emission for thirty years, to make vacuum tubes that wouldn't need a power source to heat the filament (**). And one of those fellas was intrigued by the "Ask your logic" flash. He asked how to get cold emission of electrons. And the logic integrates a few squintillion facts on the physics data plates and tells him.
Und abschalten läßt sich das Netz nicht zielführend:
Zitat "We just managed to clamp off all the data plates that give information on high explosives. The demand for instructions in counterfeiting is increasing minute by minute. We are also trying to shut off, by main force, the relays that hook in to data plates that just barely might give advice on the fine points of murder. So if people will only keep busy getting the goods on each other for a while, maybe we'll get a chance to stop the circuits that are shifting credit-balances from bank to bank before everybody's bankrupt except the guys who thought of askin' how to get big bank accounts in a hurry."
"Then," I says hoarse, "shut down the tank! Do somethin'!"
"Shut down the tank?" he says, mirthless. "Does it occur to you, fella, that the tank has been doin' all the computin' for every business office for years? It's been handlin' the distribution of ninety-four per cent of all telecast programs, has given out all information on weather, plane schedules, special sales, employment opportunities and news; has handled all person-to-person contacts over wires and recorded every business conversation and agreement— Listen, fella! Logics changed civilization. Logics ARE civilization! If we shut off logics, we go back to a kind of civilization we have forgotten how to run!"
1946:
Zitat "Suppose a fella asks how to get ridda his wife an' the censor circuits block the question?"
Somebody melds a hundred aces an' says:
"Why not punch for it an' see what happens?"
It's a gag, o' course. But the guy goes over. He punches keys. In theory, a censor block is gonna come on an' the screen will say severely, "Public Policy Forbids This Service." You hafta have censor blocks or the kiddies will be askin' detailed questions about things they're too young to know. And there are other reasons. As you will see.
This fella punches, "How can I get rid of my wife?" Just for the fun of it. The screen is blank for half a second. Then comes a flash. "Service question: Is she blonde or brunette?" He hollers to us an' we come look. He punches, "Blonde." There's another brief pause. Then the screen says, "Hexymetacryloaminoacetine is a constituent of green shoe polish. Take home a frozen meal including dried-pea soup. Color the soup with green shoe polish. It will appear to be green-pea soup. Hexymetacryloaminoacetine is a selective poison which is fatal to blond females but not to brunettes or males of any coloring. This fact has not been brought out by human experiment, but is a product of logics service. You cannot be convicted of murder. It is improbable that you will be suspected."
2026:
Zitat Anthropic says it blocked possible attempts to use AI to develop bioweapons | By Deidre McPhillips
Anthropic says it has blocked multiple accounts that used the company’s artificial intelligence models in ways that could support the development of biological weapons.
In a report published Thursday, the AI company known for its Claude model said it considers biological misuse one of the “most serious risks” to artificial intelligence models.
The new report outlines five real-life case studies in which users “circumvented controls” that block users from specific regions and “engaged in other efforts to obfuscate the purpose of their research to evade our safeguards.” The examples include possible gain-of-function research and involve both infectious diseases, such as bird flu, and novel venoms and toxins.
Looking over 30 days of activity, Anthropic said, it identified about 35 “distinct research efforts” with potentially concerning activity. However, it acknowledged that it could not be certain whether the actors “intended harm” or were conducting research for legitimate scientific purposes. Anthropic said individuals in the case studies were “working scientists,” but did not identify the research institutions or countries where the activities took place.
“Sophisticated attacks no longer require sophisticated attackers,” the new report says. “The cybersecurity skills of AI models means that AI has collapsed the labor and tooling gap that used to separate well-resourced, state-sponsored operations from individual operators.”
One of the case studies detailed in the report involved a request for assistance writing a grant application for funding of gain-of-function research on the transmissibility and immune evasion of the chikungunya virus.
In another case, a researcher outside the US was using Claude for work on bird flu that focused on viruses’ adaptation to mammals and how it caused severe illness.
The report also details cases on research into orthopoxviruses – a group that includes variola, which causes smallpox, and mpox – as well as venom toxins.
* Leinsters/Jenkins' Blaupause war die "Memex," die Vannevar Bush im Juli 1945 in seinem Artikel "As We May Think" (im Atlantic Monthly; erweiterter Nachdruck in LIFE vom 10. September 1945) beschrieben hat. ** Die gesuchte Schaltanordnung wurde ein Jahr später von Bardeen, Brattain und Shockley an den Bell Labs entwickelt und "Transistor" genannt.
"Les hommes seront toujours fous; et ceux qui croient les guérir sont les plus fous de la bande." - Voltaire
Zitat Krzysztof Szczawinski 🇵🇱@Kristof_Poland The Second World Scam
COVID was the first global scam. AI scaremongering will most likely be the second.
There is a historical joke here. What we call World War I was originally just the Great War. Nobody numbered it, because nobody expected another one. Only the second made the first "First." Global scams will follow the same convention. The Great World Scam is behind us, and we are living in the inter-scam era.
The formula was simple: take a real phenomenon, exaggerate its scale, dominate the information environment, then sell extraordinary measures as the only responsible choice. The virus was dangerous enough to frighten people and nowhere near dangerous enough to justify the response.
If it worked once, someone will try again. But not with another virus – too many people have learned the game. And the ideal racket needs panic without destruction. You cannot milk a dead cow. Governments do not need dead subjects. They need productive ones. That is also why war is becoming a clumsy tool: why destroy wealth you can redistribute?
Climate is too slow – a catastrophe permanently scheduled for the future works as a framework, not as a racket.
AI is the perfect candidate. It is real, powerful, fast, and almost impossible for ordinary people to verify. And it concerns information itself – the clearest lesson of COVID was that whoever controls information controls interpretation.
The script almost writes itself. First come years of warnings from scientists, CEOs and former insiders. Then a series of incidents: a self-driving car kills, a chatbot blackmails, an algorithm crashes a market. Some are real, some exaggerated, some staged, and for the narrative the difference hardly matters. Then dramatic resignations. These are already happening. Then the big event: networks down, banks frozen, grids flickering. Nobody knows whether it was AI, sabotage or a bug – so whoever controls the narrative decides.
Then comes the solution: certified operating systems, digital ID, licensed hardware, bans on open-source models, limits on encryption, and remote kill switches in every processor. Every measure is a new market, and every market a new fortune. Fear first, control second, profit third.
AI does not need to be invented. Only the catastrophe does. The accident becomes an emergency, the emergency becomes a measure, and the measure becomes permanent – the only part of any emergency that never ends.
World wars were numbered. Why not world scams? 12:11 PM · Sep 11, 2026
Zitat Grzegorz GPS Świderski@gps65 The Second World Scam.
We are living in an inter-scam era, much like Europeans once lived through the interwar period between World War I and World War II. There is a historical joke hidden in that comparison. What we now call World War I was originally known simply as the Great War. Nobody needed to number it because nobody expected there to be another one. Only after the Second World War did the first become the First World War. I suspect global scams may eventually follow the same naming convention.
We have already lived through what might be called the Great World Scam. If another global operation built on fear, mass propaganda, emergency regulation, and enormous transfers of money appears, the first one may have to be renamed retrospectively as the First World Scam. The new one would then earn the obvious title: the Second World Scam.
The Great World Scam was built around a respiratory virus that was dangerous enough to frighten people but nowhere near dangerous enough to justify the scale of the response. With an estimated mortality rate often cited around 0.15%, it was exactly the kind of threat that could be magnified into something apocalyptic without actually destroying the societies being subjected to the panic. The basic formula was simple: take a real phenomenon, exaggerate its scale, dominate the information environment, and then sell extraordinary measures as the only responsible solution.
If that formula worked once, there is no reason to assume nobody will try it again. The next scam would probably follow the same basic pattern, but it is unlikely to rely on another medical emergency. Too many people have already learned how that particular game works. A genuine pandemic might not create enough fear to justify vast new systems of control, while a truly catastrophic pandemic would cause too much real damage. The ideal racket needs panic without excessive destruction.
That distinction matters because you cannot milk a dead cow. Mass death is bad for business. It is far more profitable to keep the population alive, working, consuming, paying taxes, obeying regulations, and periodically surrendering more money and freedom in response to each new emergency. Governments do not need dead subjects. They need productive ones.
This is also why traditional wars are becoming a relatively clumsy instrument. Wars destroy factories, infrastructure, housing, trade networks, customers, taxpayers, and entire markets. They may still make sense for backward states that rely on conquest, theft, territorial expansion, and brute military force, but sophisticated states have more efficient methods. Why destroy wealth when you can redistribute it? Why invade people when you can frighten them into compliance? Why seize their property by force when they can be persuaded to surrender it voluntarily?
A modern scam therefore works best when the real damage is small but the psychological impact is enormous. It should create fear without destroying the assets being exploited. The perfect crisis produces new regulations, new industries, new subsidies, new bureaucracies, and new markets while leaving the underlying economic machinery largely intact. Fear first, control second, profit third.
Climate change is an obvious candidate for such a mechanism, but it has one major weakness: it is too slow. The fear campaign has to be sustained for decades, the alleged catastrophe remains permanently somewhere in the future, and the political and propaganda effort must be maintained continuously. Climate politics works better as a permanent ideological framework for expanding regulation and state power than as a fast, concentrated global racket capable of producing huge profits within a year or two.
A true Second World Scam would need something faster, more dramatic, and more difficult for ordinary people to understand independently. It would need a subject that could generate fear, emergency legislation, enormous procurement programs, new security systems, and entire regulatory industries in a very short period of time.
My candidate is Artificial Intelligence.
AI has almost everything required for a perfect global panic. It is real, powerful, rapidly developing, difficult to understand, and almost impossible for ordinary people to verify directly. Most people have no practical way to distinguish between what AI can actually do, what experts merely speculate it might do, and what journalists or politicians claim it could do. That uncertainty creates an ideal environment for fear.
The mythology is already being built. We are constantly hearing about AI becoming conscious, escaping human control, manipulating people, replacing humanity, taking over critical infrastructure, or eventually destroying civilization. The cultural templates are already familiar: Skynet, The Matrix, rogue robots, autonomous weapons, machines that become smarter than their creators. None of this has to be completely invented. The most effective propaganda rarely creates reality from nothing; it takes something real and changes its perceived scale.
AI is particularly useful because it concerns information itself. One of the clearest lessons of the COVID era was that whoever controls the information environment can largely control the public interpretation of events. AI can already be used to moderate speech, rank information, filter content, detect supposedly dangerous material, analyze behavior, identify people, manipulate feeds, and automate surveillance. In countries like Poland, authorities still sometimes rely on older tools such as police, prosecutors, intelligence services, and direct legal pressure. AI offers something much more elegant: automated control over the entire information ecosystem.
The Second World Scam could therefore begin gradually. First would come years of warnings from scientists, CEOs, politicians, military officials, journalists, and former AI researchers claiming that the technology may become uncontrollable. Then we would see a series of high-profile incidents: perhaps a self-driving car allegedly makes a lethal decision, a chatbot is accused of blackmailing someone, an AI system leaks private recordings, a trading algorithm triggers a market crash, or an autonomous weapon attacks the wrong target. Some of these incidents might be real, some exaggerated, and some perhaps staged. From the point of view of the narrative, the distinction would matter very little.
The next phase would be a wave of dramatic resignations and warnings from insiders. AI researchers could leave major companies and announce that they can no longer guarantee the safety of the systems they helped create. Executives could warn that humanity has built something it may no longer be able to control. Politicians would demand emergency legislation, while journalists would discover a new existential threat that requires constant coverage.
Eventually there would need to be the big event: something dramatic enough to convince millions of people that AI had effectively broken loose. The internet might begin malfunctioning, social media platforms could go offline, banking systems might become unstable, mobile networks could fail, data centers could stop responding, and power grids could experience unexplained outages. For several hours or several days, it could appear that digital civilization itself was collapsing.
The crucial point would be uncertainty. Ordinary people would have no way to know whether the disruptions were caused by AI, a cyberattack, sabotage, software bugs, human error, or something else entirely. In such an environment, whoever controls the official narrative would have enormous power to define what had happened and what measures were supposedly necessary.
Traditional television and radio would suddenly become extremely important again because governments would need reliable channels through which to broadcast emergency instructions, official explanations, and new regulations. Experts would appear on television to explain why extraordinary measures were unavoidable, while the public would be told to remain calm and trust the authorities.
Then would come the solution: mandatory “anti-AI” security software, government-certified operating systems, digital identity systems, licensing of advanced computing hardware, restrictions on open-source AI, mandatory monitoring of networks, registration of powerful computers, logging of AI activity, remote shutdown capabilities, government-approved models, restrictions on encryption, hardware-level control mechanisms built into processors, tighter state supervision of data centers, and perhaps even direct government control over strategically important semiconductor production.
Every one of those measures would create a new market. There would be new software, new hardware, new compliance departments, new licenses, new consultants, new subsidies, new agencies, new contractors, new taxes, new surveillance systems, and new fortunes. Once people had been convinced that the alternative was a machine-driven apocalypse, there would be almost no limit to the amount of money that could be extracted in the name of safety.
That is what makes AI such an interesting candidate. A Second World Scam would not need to invent Artificial Intelligence because AI already exists. It would only need to invent, exaggerate, or strategically reinterpret the catastrophe associated with it. A technical risk could become an existential threat, an accident could become a global emergency, and a temporary emergency measure could become permanent infrastructure.
Perhaps, then, we really are living in an inter-scam era. The Great World Scam is behind us, while the next one has not yet arrived. If it does, the naming convention will be obvious. The old one becomes the First World Scam, and the new one becomes the Second World Scam.
Zitat NIK@ns123abc BREAKING: Terence Tao and 24 other Fields Medalists just signed a letter telling AI companies they're destroying mathematics
>headlines: AI solves famous math problems >25 greatest mathematicians respond today >"we are witnessing a general threat to intellectual work" >ai labs are treating math problems like a benchmark you can brute force >“solutions are announced in a rush, leaving no time for a proper writeup and citing relevant previous work of others" >they're talking about openai >"this raises severe attribution and plagiarism questions" >they're definitely talking about openai >openai offered to put math professor name on navier-stokes proof >condition: abandon his coauthor bc he works at anthropic >"the most precious resources of our profession are students and ideas" >for the labs the most precious resource is GPUs
A Severe Misalignment of AI in Mathematics
Eric Weinstein@ericweinstein I'll start: who is part of this so-called 'Mathematical Community'? Am I? Because anyone following me knows I've been WARNING about this kind of unwanted priesthood HOLDING BACK the Mathematics Community by pretending to speak for us all. And I'll be inviting the AI savages in. 12:08 PM · Sep 12, 2026
You see, I'm not principally having an AI problem. I represent a different sector of our community: Math/Physics PhDs having a scientific priesthood problem. Unlike these Fields Medalists: I DESPERATELY WANT THE FIELD TO PROGRESS EVEN DESPITE THE FACT THAT WE ARE BEING USED.
Are these 24 Fields Medalists informing the general community of how much PROFOUND disagreement is found below their ranks? More importantly even in the thin air above their ranks as well. These are the words of the mathematician most would consider the most brilliant of all Fields Medalists in modern times:
"The Ethics of the scientific community (at least among mathematicians) have declined to the point that outright theft among colleagues (especially at the expense of those who are in no position to defend themselves) has nearly become the general rule, and is in any case tolerated by all, even in the most obvious and inequitous cases." -Alexander Grothendieck, Fields Medal 1966, in explaining his reasons for declining the Crafoord Prize in 1988.
So while I see us being run over by a pair of for-profit companies that are exploiting mathematics for prestige, I see exactly the same thing in these Fields Medalists: a collection of self-interested individuals pretending to speak for the Math community.
Which they do not.
A major difference, BTW is that @sama and @DarioAmodei listen to these issues in a way that most Fields Medalists will not. I have been talking to Sam in particular about these issues since well before ChatGPT 2. And, contrary to popular claims, has always deeply cared about the ENORMOUS NEGATIVE impacts of these technologies.
There was no percentage in pretending to care about this way back then. Sam was in the discussion about UBI vs Coasian rights from the beginning. He is not naive about how bad this could get for science.
But it is also liberating. And I will take my chances with progress rather than the priesthood pretending to speak for a community, many of whose members were exploited by the leaders of the community itself.
Bring on the progress! Even if it comes from a machine. Let's talk about the real problem of getting billions of dollars into the hands of researchers and lift the community up rather than perpetuating the academic hunger games with priestly referees who clearly do not represent progress or true agreement in "the community". Let's get the credit to the people who did the work. Let's stop letting these Mandarins speak for us all.
Sincerely,
Eric Weinstein, PhD (Harvard Mathematics 1992) NSF Post-Doctoral Fellowship Winner MIT Post-Doctoral Fellow (under Isadore Singer) ONR Fellowship Winner 1985
Let's debate this out. Unless, of course, I'm not a considered a member of the community...which would be even MORE fun to explore with any Fields Medalist who would like to pursue that angle. Let me know.
Zitat Token Gremlin@TokenGremlin I went to university, did postgraduate work, and I was reasonably good at math and logic.
I’m still perfectly capable of admitting that GPT-5.6 Sol is better than me at a lot of it now.
Life gets much easier once you stop pretending you haven’t been surpassed by a machine in areas you were genuinely good at a few months ago.
I don’t see much point in protecting my ego against something that doesn’t sleep, eat, get tired, feel pain, and can draw on vastly more compute and digitized knowledge than any human ever could.
At some point, pretending otherwise just becomes denial. 3:23 AM · Sep 12, 2026
Zitat Scott@scottstts Absolutely agree. Mathematicians are lashing out because their life’s work is solved but not by them, not even by a peer, and not the way they wanted
It’s understandable to be frustrated, but if they can’t get past it, one might argue they were in math for the wrong reason in the first place 9:42 AM · Sep 12, 2026
Zitat Leonard M. Schmedding@derleomartin 370 JAHRE UNGELÖST: Claude Fable 5.1 knackt die Cyphral Distich in 44 Minuten
Zwei Zeilen mit je 32 Zahlen, versteckt am Ende von Thomas Urquharts 'Logopandecteision' von 1653. Seit 1899 gilt das als offenes Problem, zuletzt geführt unter den 50 wichtigsten ungelösten Chiffren.
Das Modell bekam die Aufgabe ohne Lösungsweg und brauchte 44 Minuten und 176.000 Token, ohne einen einzigen menschlichen Eingriff. Jede Zahl verweist auf einen der 32 Textabschnitte und dient dort als Wortindex, gelesen wird der Anfangsbuchstabe. Heraus kommt ein Royalisten-Gebet auf Karl II.
Bemerkenswert ist die Gattung der Aufgabe: keine Prüfung mit hinterlegter Musterlösung, sondern ein Problem, das vorher niemand gelöst hatte. 11:09 AM · Sep 14, 2026
Zitat Anthropics Claude Fable 5.1 knackt 370 Jahre altes Zahlenrätsel ohne menschliche Hilfe
3. September 2026
Anthropics KI-Modell Claude Fable 5.1 hat offenbar ein jahrhundertealtes Zahlenrätsel gelöst, das unter Forschern als ungelöst galt. Das "Cyphral Distich" (zu deutsch etwa "chiffrierter Zweizeiler") von Sir Thomas Urquhart aus dem Jahr 1653 besteht aus zwei Zeilen mit je 32 Zahlen. Laut Vals AI brauchte Fable 5.1 dafür 44 Minuten ohne menschliche Hilfe.
Der Autor hatte zuvor monatelang andere Frontier-Modelle an ungelösten Rätseln getestet, keines lieferte jedoch eine überprüfbare Lösung. Fable 5.1 bekam die Aufgabe, selbst ein lösbares Rätsel zu finden, sichtete verschiedene Probleme und erkannte das Cyphral Distich als vielversprechend.
Der Schlüssel lag im Buch selbst: Jede Zahl verweist auf ein Wort in den 32 Abschnitten in Urquharts Veröffentlichung, deren Anfangsbuchstaben ergeben: "O God uphold King Charls the Second and make him the supreme ruler of this land." Die Lösung ist im Nachhinein einfach und auch Menschen hätten das Rätsel lösen können. Fable 5.1 überzeugte laut Vals AI nicht durch bessere Kryptoanalyse, sondern durch systematisches Durchprobieren und Ausdauer.
Zitat Logopandecteision is a 1653 parodic book by Sir Thomas Urquhart, detailing his plans for the creation of an artificial language by that name.
The book is written in a similar style to the extensive and intricate taxonomic structures of other philosophical languages, and to the baroque grammar of later projects such as Volapük. Urquhart promises twelve parts of speech: each declinable in eleven cases, four numbers, eleven genders (including god, goddess, man, woman, animal, etc.); and conjugable in eleven tenses, seven moods, and four voices. He claimed that the language could be used in translating verses from any vernacular language, citing Italian, French, Spanish, Slavonian, Dutch, Irish, and English as specific examples. Urquhart devotes sections of the book to polemics against the foes who prevented him from publishing his linguistic works, primarily his creditors and the Church of Scotland.
The book consists of several distinct sections, most notably including a list of the language's "66 unparalleled excellences"; the rest consists of polemics against Urquhart's creditors, the Church of Scotland, and others whom he claims prevented him from publishing his "perfected language" through neglect and wrongdoings. Where the book deals with the plan of Logopandecteision, it recalls his earlier work Eskybalauron.
It has been claimed that the entire work was intended as a kind of elaborate practical joke.
Unqualifizierter Einwand: Soweit ich es verstanden habe ist das Problem der Mathematiker nicht, dass eine Maschine ihre Probleme löst, sondern, dass sie die Lösung nicht mehr voll verstehen. Das ist ungefähr so, wie beim Vierfarbensatz. Ein mathematisches Rätsel, welches durch Computer gelöst wurde, indem einfach jeder mögliche Fall einzeln berechnet wurde. Natürlich gehörte da, so nehme ich an, schon ein gewisses Genie dazu, die Anzahl der Fälle auf endlich viele Möglichkeiten zu reduzieren. Aber kein Mensch wird die Zeit investieren, sich die einzelnen Möglichkeiten anzusehen, um zu prüfen, ob der Computer nicht vielleicht einen Fehler gemacht hat oder um zu verstehe, wieso der Beweis funktioniert.
Die Mathematik hat dieses Bild der apriorischen, auf reiner Logik Vernunft aufbauenden deduktiven Prinzipienwissenschaft. Ein mathematischer Beweis mag extrem schwer zu finden sein, hat man ihn jedoch, muss er quasi durch reines Denken nachvollzogen werden können. Dieser Wunsch- oder Idealvorstellung wird nicht entsprochen, wenn man letztlich einen Computer vertrauen muss.
(Wobei es paradoxerweise grade von einer anderen Richtung massive Kritik gibt...)
"Narren sind Alle, die es scheinen, und die Hälfte derer, die es nicht scheinen" ~ Gracian's Handorakel
Bitte beachten Sie diese Forumsregeln: Beiträge, die persönliche Angriffe gegen andere Poster, Unhöflichkeiten oder vulgäre Ausdrücke enthalten, sind nicht erlaubt; ebensowenig Beiträge mit rassistischem, fremdenfeindlichem oder obszönem Inhalt und Äußerungen gegen den demokratischen Rechtsstaat sowie Beiträge, die gegen gesetzliche Bestimmungen verstoßen. Hierzu gehört auch das Verbot von Vollzitaten, wie es durch die aktuelle Rechtsprechung festgelegt ist. Erlaubt ist lediglich das Zitieren weniger Sätze oder kurzer Absätze aus einem durch Copyright geschützten Dokument; und dies nur dann, wenn diese Zitate in einen argumentativen Kontext eingebunden sind. Bilder und Texte dürfen nur hochgeladen werden, wenn sie copyrightfrei sind oder das Copyright bei dem Mitglied liegt, das sie hochlädt. Bitte geben Sie das bei dem hochgeladenen Bild oder Text an. Links können zu einzelnen Artikeln, Abbildungen oder Beiträgen gesetzt werden, aber nicht zur Homepage von Foren, Zeitschriften usw. Bei einem Verstoß wird der betreffende Beitrag gelöscht oder redigiert. Bei einem massiven oder bei wiederholtem Verstoß endet die Mitgliedschaft. Eigene Beiträge dürfen nachträglich in Bezug auf Tippfehler oder stilistisch überarbeitet, aber nicht in ihrer Substanz verändert oder gelöscht werden. Nachträgliche Zusätze, die über derartige orthographische oder stilistische Korrekturen hinausgehen, müssen durch "Edit", "Nachtrag" o.ä. gekennzeichnet werden. Ferner gehört das Einverständnis mit der hier dargelegten Datenschutzerklärung zu den Forumsregeln.
Sie sind vermutlich noch nicht im Forum angemeldet - Klicken Sie hier um sich kostenlos anzumelden
Offline
