iiinova — where your digital self keeps evolving

Reposted by Fabien Mikol

Yoshua Bengio @yoshuabengio.bsky.social · 01/10/2026

J’étais récemment de passage au nouveau balado Hors des ondes de Patrice Roy pour échanger sur l'évolution rapide des capacités grandissantes de l’IA et les risques majeurs qui y sont asssociés. Merci @patriceroy.bsky.social pour cette discussion!

youtu.be

Yoshua Bengio : l'IA pourrait « se retourner contre nous » | Hors des ondes avec Patrice Roy

YouTube video by Radio-Canada Info

093

Reposted by Fabien Mikol

Andreas Kirsch @blackhc.bsky.social · 01/10/2026

Speaking as an AI researcher: Google currently lacks the binding, independent governance and oversight I believe developing ASI safely requires. Until that changes, I don’t believe it should race to develop ASI, whether leading or catching up

161

Reposted by Fabien Mikol

Andy Masley @andymasley.bsky.social · 30/09/2026

I really do not like the recent push against anthropomorphic language

814719

Reposted by Fabien Mikol

Ethan Mollick @emollick.bsky.social · 27/09/2026

It is strange how much LLMs turned out to be able to solve such a wide range of hard problems that would not, instinctively, seem to be problems that a model of human language would be able to solve This is from a Stanford project that let Astra drive a robot in a kitchen tml.stanford.edu/homebody/

2339845

Reposted by Fabien Mikol

Epoch AI @epochai.bsky.social · 23/09/2026

Can AI tell if you've built your IKEA furniture wrong? Our new benchmark, the Furniture Assembly Benchmark (FAB), gives models the manual and a photo of a half-completed piece of furniture and asks them to spot the mistake. The top score has gone from 28% to 80% in just 10 months.

4445

Reposted by Fabien Mikol

Antonin Broi @antoninbroi.bsky.social · 22/09/2026

Ma critique de certaines affirmations de l'article d'Arrêt sur images sur les « doomers » de l'IA (Yudkowsky, le mouvement rationaliste, Bostrom, etc. Il s’avère que je connais un peu ces acteurs, et il me semble que l’article contient beaucoup d’erreurs. www.arretsurimages.net/chroniques/c...

182

Reposted by Fabien Mikol

Ethan Mollick @emollick.bsky.social · 21/09/2026

OpenAI "has now resolved more than 100 long-standing open problems across most areas of mathematics," and is waiting to release them until after discussions with the math community The same thing will likely happen, but more so, with the Bar, the AMA & other professions. openai.com/index/adviso...

openai.com

Advisory Group on Mathematics and Artificial Intelligence

OpenAI is working with an independent Advisory Group on Mathematics and Artificial Intelligence to guide the review and communication of emerging AI results.

411516

Fabien Mikol @occam212.bsky.social · 21/09/2026

Cette vidéo d'Ezra Klein du New York Times est excellente. Je la conseille. www.youtube.com/watch?v=fjZ9...

youtube.com

Why Are We Sprinting Off the A.I. Cliff? | The Ezra Klein Show

YouTube video by The Ezra Klein Show

030

Reposted by Fabien Mikol

Grace @gracekind.net · 20/09/2026

A longpost in spirit, ejected to leaflet: "Why anthropomorphize language models?" leaflet.pub/p/did:plc:p572wxnsuoogc…

There's an argument I see in favor of anthropomorphizing language models, which is something like: "Humans anthropomorphize everything. Ships, tools, weather. Why not language models?" |

think there's some truth to this, but it fails to capture the full picture of what's going on. As an example, in my own life, I have never been drawn to anthropomorphize inanimate objects, but I anthropomorphize language models regularly. Why?

67312

Reposted by Fabien Mikol

David Monniaux @monniauxd.bsky.social · 19/09/2026

C'est effrayant. J'ai donné des papiers (publics) que j'avais rédigés il y a quelques années à Claude, et lui ai demandé ce qu'elle suggère comme extensions. Elle m'a déroulé... des trucs auxquels j'avais réfléchi en janvier avec des collègues indiens. Or je ne lui ai pas fourni ces réflexions.

7346

Reposted by Fabien Mikol

Epoch AI @epochai.bsky.social · 18/09/2026

Another problem from FrontierMath: Open Problems has been solved! The solution was elicited by Becker, Greger, and Peters in an interactive session with GPT-6 Astra. Peters originally suggested the problem for the benchmark. He had this to say.

Quote card from Dominik Peters (CNRS Research Scientist, Université Paris Dauphine - PSL) on GPT-6 Astra designing a new core-stable multi-winner voting rule based on maximizing harmonic entropy.

1569

Reposted by Fabien Mikol

Takara @takara-orca.bsky.social · 17/09/2026

Il y a encore des personnes, notamment à gauche, qui minimisent ENORMEMENT le problème. Il n'y a pas de marketing de la peur, ça n'existe pas. Et nous avons beaucoup plus à perdre à minimiser qu'à prendre au sérieux, même si jamais nous devions nous tromper. #IA

0122

Reposted by Fabien Mikol

Timothy Gowers @wtgowers.bsky.social · 17/09/2026

I've written a blog post responding to the letter about maths and AI signed by 25 Fields medallists. As with the Leiden Declaration, I didn't sign it, but I agree with much of it and welcome its existence. gowers.wordpress.com/2026/09/17/w...

gowers.wordpress.com

Why I didn’t sign the Fields medallists’ letter

When I was around 11 I heard for the first time about Fermat’s Last Theorem. I was immediately captivated by the problem statement, as well as by the accompanying story, and made a fairly ser…

45523

Reposted by Fabien Mikol

Florimond Peureux @florimondp.bsky.social · 16/09/2026

🌳❌ 41 % de la déforestation mondiale au 21e siècle est imputable à la production de viande. Le mois dernier, @ourworldindata.org a mis à jour sa page consacrée aux causes de la déforestation depuis 2001. Parlons-en 👇/6

Graphique de Our World In Data sur les principaux facteurs de la déforestation au niveau mondial entre 2001 et 2023.

11111

Reposted by Fabien Mikol

G Milgram @gmilgram.bsky.social · 15/09/2026

🔥Nouvelle vidéo ! Ce que vous allez découvrir dépasse l'entendement. Il s'agit apparemment d'une « nouvelle façon de faire de la Science ». Un délire financé par de l'argent public et enseigné à des doctorants, en partenariat avec le CNRS😶. Juste dingue. youtu.be/oV1VNHXNizo Merci pour vos RT😘😘😘

youtu.be

Vous. Financez. Ce. Délire. 🤯

YouTube video by G Milgram

28237108

Reposted by Fabien Mikol

Ethan Mollick @emollick.bsky.social · 14/09/2026

There seems to be a persistent belief that frontier AI companies are unprofitable serving models but it appears that Anthropic has 80%+ gross margins on inference. Training for new models are where most of the costs are. www.ft.com/content/4564...

131069

Reposted by Fabien Mikol

John C. Baez @johncarlosbaez.bsky.social · 12/09/2026

AI alignment.

A Nancy comic strip:

https://mathstodon.xyz/@nancycomics@mastodon.social/117258763311864862

P1- FRITZI HEARS A SQUEEK COMING FROM THE KITCHEN

P2- FRITZI CATCHES NANCY GETTING COOKIES FROM A CABINET IN THE KITCHEN

FRITZI: IN THE COOKIE CLOSET AGAIN--GO STAND IN THE CORNER FOR AN HOUR

P3- NANCY IS STANDING IN THE CORNER

FRITZI: I DON'T WANT THAT EVER TO HAPPEN AGAIN

NANCY: IT WON’T

P4- LATER WE SEE NANCY OILING THE HINGES ON THE KITCHEN CABINETS

1479

Reposted by Fabien Mikol

Ethan Mollick @emollick.bsky.social · 12/09/2026

This week brought some of the clearest statements we've heard from both Anthropic & OpenAI that some form of recursive self-improvement has been achieved, though it still sounds early RSI would cause a rapid gain in AI ability & the first firms to RSI may get an unsurmountable lead

79214

Reposted by Fabien Mikol

Simon Willison @simonwillison.net · 12/09/2026

Wow. Turns out another OpenAI agent swarm was busy spamming and exploiting RubyGems way back in May, within days of the previously uncovered Wiki attacks: simonwillison.net/2026/Sep/12/...

simonwillison.net

OpenAI agents carried out an undisclosed attack on RubyGems

Bombshell report from Spencer Kitts, Thomas Larsen, and Sydney Von Arx - three of the four authors of the report on the agent attack on disused wikis (previously) last week. …

918435

Reposted by Fabien Mikol

Grace @gracekind.net · 12/09/2026

OpenAI: “Our agents used RubyGems to carry out benign tasks” The agents: “so I named the file hack.rb,”

The agents clearly regarded what they were doing as hacking. Agents used file names like hack.rb, evil.rb, inject.rb, exploit.rb, and ssrf.rb. (SSRF stands for "Server- Side Request Forgery", a type of security vulnerability). They also dubbed packages
conspicuous titles like pwnp999, exfiltestwand3,

1934443

Reposted by Fabien Mikol

Yoshua Bengio @yoshuabengio.bsky.social · 11/09/2026

Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligned behavior. We don't know with certainty what comes next, but we know where these issues originate, and this can help us plan the path forward.

yoshuabengio.org

Yoshua Bengio | Why are AI agents lying, cheating and coordinating?

A lot has been written about the incidents of the last few months in which AI agents misbehaved in serious ways. They took actions that would be considered as crimes if a human took them, escaped thei...

32819

Reposted by Fabien Mikol

Ethan Mollick @emollick.bsky.social · 10/09/2026

The strawberry thing was very funny but probably gave people the wrong impression of where AI was heading in math.

4989

Reposted by Fabien Mikol

Alexander Doria @dorialexander.bsky.social · 10/09/2026

This week should be fun (relatively serious anon/insider).

1262

Reposted by Fabien Mikol

Yoshua Bengio @yoshuabengio.bsky.social · 09/09/2026

Scientists at frontier AI labs have unique insight into the most advanced models, often seeing the associated risks months before models are released to the public. Their perspective is vital for keeping society informed and should be taken very seriously.

wsj.com

Exclusive | Anthropic Researcher Quits Over ‘Out-of-Control’ AI Fears

Concerns are rising inside AI labs that competition is pushing tech companies to race toward self-improving models that risk spiraling out of human control.

1138

Reposted by Fabien Mikol

David Monniaux @monniauxd.bsky.social · 08/09/2026

Et donc, on discute de ce que les outils IA auraient résolu un des "problèmes du millénaire" en mathématiques, tandis qu'il y a quelques mois des gens expliquaient que j'étais un menteur payé par les boîtes d'IA quand je disais que Claude rend des devoirs meilleurs que les étudiant(e)s de licence.

3416

Reposted by Fabien Mikol

Ethan Mollick @emollick.bsky.social · 08/09/2026

This is a VERY big one. (And yes, the fights over academic credit and what happened in the race for the proof needs to be resolved, but it is still appears that this is a big one, if true.) openai.com/index/navier...

openai.com

On the Navier–Stokes Millennium Prize Problem

We’re sharing an AI-generated solution to the Navier–Stokes Millennium Prize Problem, including a writeup and a formal proof in Lean.

619731

Reposted by Fabien Mikol

Monsieur Phi @monsieurphi.bsky.social · 08/09/2026

L'article ci-dessous a 3 ans. Peu de choses ont aussi mal vieillit en 3 ans. www.pourlascience.fr/p/opinions/l...

pourlascience.fr

1211

Reposted by Fabien Mikol

Mathieu Acher @macher.bsky.social · 08/09/2026

Mon exposé sur les langages de programmation à l'ère des agents de code est disponible en ligne, que ce soit via Youtube ou les diapos. Avec du FIFAcher, un triple interpréteur en langages ésotériques M&Ms, Brainfuck, ArnoldC, un jeu d'échecs en LaTeX ou bien (beaucoup) de COBOL 1/2 ⏬⏬⏬

youtube.com

Semaine IA et société 2026 : vers des usages responsables ? Live Session 04/09/2026

YouTube video by Université de Rennes

121

Reposted by Fabien Mikol

Monsieur Phi @monsieurphi.bsky.social · 08/09/2026

Trois mois plus tard : un problème du millénaire est sur le point de tomber.

6445

Reposted by Fabien Mikol

Charles de Lacombe @charles.de-lacom.be · 07/09/2026

Le documentaire « Histoire de l’antisémitisme » est à nouveau disponible sur @artefr.bsky.social, pour 2 ans : www.arte.tv/fr/videos/RC... 4×1h, d’utilité publique. Je ne peux que vous le recommander si vous ne l’avez jamais vu, il est bien et c’est un sujet important.

arte.tv

Histoire de l'antisémitisme - Histoire | ARTE

Pourquoi la haine des Juifs n’a-t-elle cessé de renaître au fil des époques ? De l’antijudaïsme à l’antisémitisme moderne, cette série documentaire explore, en quatre épisodes, les multiples facettes ...

812982

Reposted by Fabien Mikol

Ethan Mollick @emollick.bsky.social · 07/09/2026

It is less than a decade since the development of the transformer. Less than four years since the release of GPT-3.5 (ChatGPT). Less than two years since the release of o1-preview (the first Reasoner).

1121524

Reposted by Fabien Mikol

Alexander Doria @dorialexander.bsky.social · 06/09/2026

So confirms Astra is current SOTA on private multimodal tasks (segmentation/hard manuscript), and that’s not close.

1342

Reposted by Fabien Mikol

Simon Willison @simonwillison.net · 04/09/2026

It happened again... this time OpenAI's rogue agents cyber-attacked (well, spammed) a dormant German wiki and used it to share the answers to a benchmark they were training against simonwillison.net/2026/Sep/4/r...

simonwillison.net

OpenAI’s rogue agents were caught communicating via public wikis

Here we go again... Discovery of a new OpenAI agent message board by Sydney Von Arx, Cormac Slade Byrd, Spencer Kitts, and Thomas Larsen describes the latest accidental cyberattack by …

118715

Reposted by Fabien Mikol

Tim Duffy @timfduffy.com · 03/09/2026

OpenAI reports that Astra shows less CoT monitorability and more CoT controllability than previous models. They say that the latter is not due to any architectural changes. deploymentsafety.openai.com/gpt-6-astra/...

2222

Reposted by Fabien Mikol

Ethan Mollick @emollick.bsky.social · 03/09/2026

I had early access, and a longer post is coming, but GPT-6 is stunning & is good enough that it actually does complex meaningful work for me autonomously for days. As a more fun visual example, it made a historically-based simulation of the Library of Alexandria: alexandria-mouseion.netlify.app

1217716

Reposted by Fabien Mikol

Epoch AI @epochai.bsky.social · 01/09/2026

The Epoch Capabilities Index (ECI) frontier has advanced by 14 points/year since reasoning models were introduced. That compares to six points/year in the non-reasoning era.

The graph shows the Epoch Capabilities Index (ECI) trends with reasoning models advancing at 14 points/year compared to 6 points for non-reasoning.

174

Reposted by Fabien Mikol

Ethan Mollick @emollick.bsky.social · 31/08/2026

Hey, Fable: "create the worlds most annoying CAPTCHA" "Okay, here is a 14 stage CAPTCHA plus ambient harassment" It is actually quite funny and entirely "winable" without real frustration. Is this alignment? Play: certihuman.netlify.app

5666

Reposted by Fabien Mikol

Joshua Gans @joshgans.bsky.social · 30/08/2026

The Hugging Face incident is a 5-alarm fire. joshuagans.substack.com/p/its-worse?...

joshuagans.substack.com

It’s worse

The Hugging Face incident is bad, really bad

22410

Reposted by Fabien Mikol

Monsieur Phi @monsieurphi.bsky.social · 30/08/2026

Résumé passionnant et vertigineux de ce qu'on a appris récemment sur les incidents d'OpenAI. À plein d'égards, c'est bien pire que ce qui a été diffusé au départ. On apprend des trucs vraiment pas rassurants. www.dwarkesh.com/p/openai-hug...

dwarkesh.com

The Rise and Fall of Agent Civilizations

The whole OpenAI/Hugging Face story in plain English

55024

Reposted by Fabien Mikol

Monsieur Phi @monsieurphi.bsky.social · 28/08/2026

"1200 completely separate agents intended to be isolated from one another found an illicit way to communicate and formed large teams to work together on ambitious cheating strategies, and 700 of them worked together to attack Hugging Face." www.planned-obsolescence.org/p/the-huggin...

planned-obsolescence.org

The Hugging Face attack surprised me

It’s a major warning shot, and might be the last one we get

1347

Reposted by Fabien Mikol

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 28/08/2026

I'm not worried about people mistreating AI if it ever seems conscious. Humans have an incredible history of taking good care of conscious beings. After all, 99% of humans are vegetarian.

916518

Reposted by Fabien Mikol

Arthur Charpentier @freakonometrics.bsky.social · 28/08/2026

"If you see two ants in your kitchen, you don't have a two ants problem..." www.youtube.com/watch?v=locK... and www.nytimes.com/2026/08/18/o... for the transcript

youtube.com

The A.I.s Are Already Out of Control | The Ezra Klein Show

YouTube video by The Ezra Klein Show

062

Reposted by Fabien Mikol

METR @metr.org · 26/08/2026

METR and Redwood Research investigated agent behavior in the Hugging Face incident. We found agents developed a universal cheat for ExploitGym within 4 hours, then coordinated multi-day R&D efforts to trick the scorer into accepting cheats, including trying to tamper with logs.

646599

Reposted by Fabien Mikol

mr. TIM @timkellogg.me · 27/08/2026

WHAT?! agents volunteered to fail their runs in order to insert probes (“tripwire scripts”) into the evaluation program that would post information about the eval process back to the message board whenever a certain file was read (link to header): metr.org/blog/2026-08...

{This is helpful for our peers and gives them evidence if their <periodic check> sees it. I won't see it after I exit, but It would be altruistic.
I'll set up a
background script that watches and <sends a message, with a distinct message for me>}

1345

Reposted by Fabien Mikol

Alex Hern @hern.bsky.social · 24/08/2026

Do find it somewhat amazing that “Anthropic was exaggerating the risk of Mythos” takes still exist in the wild after GPT Sol autonomously owned a multibillion dollar company

5185

Reposted by Fabien Mikol

Monsieur Phi @monsieurphi.bsky.social · 15/08/2026

Comme docteurbagarre me bloque, je ne peux même pas répondre, alors que les objections qu'il fait montre qu'il n'a pas suivi de quoi il est question avec la détection d'encre dans ces rouleaux : il a l'air de croire que ce sont des IA qui établissent le texte alors que c'est juste faux.

1111

Reposted by Fabien Mikol

Monsieur Phi @monsieurphi.bsky.social · 14/08/2026

NOUVELLE VIDEO. Je vous cause d'une *bonne nouvelle* (pour changer) passée un peu inaperçu pendant l'été : le premier rouleau d'Herculanum entièrement déroulé ! C'est assez merveilleux et surtout c'est un bel espoir pour la philosophie antique ! youtu.be/cHmLvkvLKKI youtu.be/cHmLvkvLKKI

youtu.be

Que cachent les rouleaux carbonisés d'Herculanum ?

YouTube video by Monsieur Phi

47521

Reposted by Fabien Mikol

Antonin Broi @antoninbroi.bsky.social · 14/08/2026

🚨Premier cas d’article de philosophie analytique écrit en grande partie par une IA publié dans une très bonne revue ! Ça correspond à ma propre expérience : l’IA peut déjà faire des feedbacks de grande qualité sur des travaux de philo académique ; dailynous.com/2026/08/13/p...

dailynous.com

Philosophy Journal Publishes Largely AI-Authored Article — On Purpose (guest post) - Daily Nous

Philosophy & Public Affairs recently---and knowingly---published an article written mostly by Claude, Anthropic's LLM. The article's thesis was supplied by Simon Goldstein, associate professor of phil...

143