Managing Digital Communication. Interested in many things that affect society and humanity - #AI #Ethics #content #UX #InformationLiteracy

Joined March 2013
199 Photos and videos
KaEl retweeted
Companies tend to target employees earning a “wage premium,” which increases inequality but not necessarily productivity buff.ly/tWrhHTE @MIT #AI Cc @sallyeaves @SpirosMargaris @ahier
1
8
6
186
KaEl retweeted
Replying to @giffmana
We have an exec summary - europe2031.ai/summary/ . Quite a lot of politicians actually reached out/ are reading this atm, I‘m confident we‘ll be able to achieve real impact. But there’s more we could all do together, dm me and can share more
4
1
25
874
KaEl retweeted
パランティアの人間からバルファキスが聞いた話。本当なら今すぐ地獄の業火に灼かれるべき。 ガザの住民密集地を🇮🇱が爆撃する時のスマホを持って逃げ惑うガザの人々の動き「パニックインプリント」をAIに食わせて訓練し「パニック防止AI」として英国NHSに14億ドルで売った

119
4,386
8,122
331,619
KaEl retweeted
The current boom in #AgenticAI has two sides: #AI agents are popping up everywhere, and are touted by their creators as the silver bullet of AI... But beyond the hype from Big Tech, almost no one wants them... #Data -- Cc @BroadenView @jblefevre60 @roxananasoi @EvanKirstel @AkwyZ @Ym78200 @dinisguarda @nincoroby @kalydeoo @IanLJones98 @BetaMoroney @sallyeaves @enilev @mvollmer1 @gvalan @HaroldSinnott @mikeflache @Nicochan33 @CathCervoni @RLDI_Lamy @LaurentAlaus @aure79lien @CurieuxExplorer @Fabriziobustama @FrRonconi @CHRISTINESOTO12 @chidambara09 @DominiqueCrochu
5
26
33
886
⛔️⛔️ EXCLUSIVE: Ruud Gullit calls on Gianni Infantino to resign amid growing World Cup controversy‼️ "I have stayed quiet for a long time because I wanted to judge this World Cup on football matters. But the deeper we get into the preparations, the more it becomes clear that football is no longer the main story. For that reason, I believe Gianni Infantino should seriously consider stepping down as FIFA President. ⛔️A World Cup should unite people. Instead, this tournament is becoming a symbol of division, political disputes, travel restrictions, and administrative failures.‼️
283
5,117
15,139
266,901
KaEl retweeted
While attention is again focused on Iran, a major shift is unfolding in the West Bank. For the first time since the Oslo Accords, Palestinians are being expelled from Area A, under PA administration. The IDF is seizing land near Jenin to build a military base. Why it matters 🧵
33
1,490
1,609
50,620
KaEl retweeted
🦔Leonardo is a $17 billion defense contractor. It built a system called SignalTrace that clips sensors onto the license plate readers already mounted on street poles, overpasses, and police cars across the US. Every time you drive past one, the sensor grabs the Bluetooth and WiFi signals from every device in your car, ties them to your plate, and logs the time and location. Your phone, your AirPods, your kid's tablet. All of it goes into the same file. A friend rides with you once and their devices are linked to your plate. Leonardo has sold this to police departments since at least 2023. There is no federal law covering it, no opt-out, and no warrant requirement. My Take None of the pieces here are new. Your phone has always broadcast a signal. The license plate cameras were already there. Leonardo just connected them and found a buyer. Nobody had to break a law or build anything from scratch. They assembled a surveillance system from parts already in place and sold it before anyone noticed. Most people found out this week from a 404 Media investigation. Leonardo received the patent in 2024. By the time you hear about something like this, the deals are done and the sensors are on the poles. That's how it works now. Hedgie🤗 404media.co/this-company-wil…
89
1,400
2,571
115,056
This is not just a remarkable decision, but more an ashaming a one. Should we name this measure ANTI- black/- muslim or what else? Is he a threat to national security? This #WorldCup 2026 is not worth it's name. (👀 also teams Senegal, Usbekistan, Iran's national team manager).
Africa’s best referee has been refused entry to the US for the Football World Cup. Omar Artan is from Somalia, one of the countries which the US has under a travel visa ban. @SophyRidgeSky and @WilfredFrost have all the details. 🎧 trib.al/YElh1hK
36
"... the Israeli army has steadily expanded its area of control, which now stands at roughly 60% of Gaza, while leveling the areas under its occupation to the ground. "
If two million people are denied land, water, schools, hospitals, jobs, safe homes and even cemeteries, the final destination will be a mass exodus. responsiblestatecraft.org/is…
19
KaEl retweeted
en 2013 Nvidia proposait un package de 400 000 dollars en equity sur 4 ans pour attirer ses meilleurs ingénieurs (énormément de concurrence entre boites tech pour attirer les meilleurs talents et Nvidia n’avait pas la grandeur d’aujourd’hui) en clair des stock options ou des rsu qui se débloquent progressivement avec un vesting sur quatre ans et souvent un cliff d'un an avant la première tranche donc ce qui ont signé durant cette période là tiendrait aujourd'hui autour de 230 millions de dollars (un x575 en 13 ans) sachant que Nvidia pèse désormais près de 5000 milliards de dollars et a battu son record absolu en mai 2026 même histoire avec Palantir dont j’ai souvent parlé ici avec une nuance technique, la boîte n'est pas passée par une ipo classique mais par une cotation directe en 2020 autour de 10 dollars l'action (aucune levée ni banque qui place les titres) juste les actions existantes rendues liquides sur le marché, une fois la période de lock-up purgée des milliers de premiers employés ont vu leurs RSU se transformer en cash, le titre vaut environ 137 dollars aujourd'hui soit plus de 13 fois ce qui a donné lieu à des tonnes de millionnaires parmi les employés de la boîte qui n'ont sans doute jamais rien fondé SpaceX prépare également la suite car l’IPO est attendue ce 12 juin entre 1500 et 1750 milliards de valo avec des employés qui s’apprêtent à à devenir multimillionnaires notamment ceux qui ont rejoins SpaceX très tôt et qui ont négocié leur equity de la Eure bien plus préférentielle quand SpaceX recherchait désespérément des talents en enchaînant les échecs de la Falcon 9 je pourrais aussi vous parler du cas Anduril (boîte Deftech dont j’ai très souvent parlé ici également) qui boucle la démonstration sans même être cotée, ici la richesse vit encore sur le papier dans la captable, sachant que les fonds entrent en actions préférentielles avec une liquidation preference qui les rembourse en premier quand les salariés détiennent de l'ordinaire, valorisée 30 milliards il y a un an puis 61 milliards en mai dernier après une sérieH de 5 milliards, à chaque tour les early employees encaissent un peu de dilution mais leur participation prend de la valeur bien vite et y’a encore pas mal d’autres boites deeptech qui ont vu le jour ces 5 dernières années bien moins connues sont je pourrais vous parler avec le même schéma qui se dessine pour les early employés étant donné que ces boites s’adressent à des marchés de plusieurs centaines / milliers de milliards de dollars tout ça pour ces ultra riches ne sont pas forcément des chefs d'entreprise, des patrons d'industrie ou des gérants de fonds, ça peut être juste un employé qui a rejoint une deeptech extrêmement ambitieuse très tôt, qui a négocié ses termes d'equity plus dur que son salaire, un strike bas, un vesting propre, une bonne place dans la cap table avant les grosses levées qui diluent et qui a été récompensé pour une seule chose: avoir eu une thèse juste et l'avoir tenue avant tout le monde sur le LONG TERME bien avant que celle-ci ne devienne enfin évidente pour la masse des futurs investisseurs
Quel métier font ces gens ?
13
66
755
258,493
A tribute to John Coltrane: I combined family footage and performance clips into a music video for one of my favorite songs of his, “Equinox”. Hope you enjoy watching it as much as I enjoyed making it.
44
542
2,337
71,115
📌🔖👀 " Modern AI development heavily relies on distillation. Large teacher models generate synthetic data. Smaller student models learn from it. If hidden behaviors survive that pipeline, they could quietly propagate across generations of models — even after filtering. "
Anthropic just published a paper that raises uncomfortable questions for every AI company on the planet. Including themselves. The phenomenon is called subliminal learning. Published in Nature on April 15, 2026, the paper was co-authored by researchers from Anthropic, UC Berkeley, Warsaw University of Technology, and the AI safety group Truthful AI. The core finding: AI models can inherit traits from other models through training data that appears completely unrelated. Not through obvious contamination. Not through labels. Not through explicit instructions. Instead, through subtle statistical patterns hidden inside outputs that look harmless — number sequences, code snippets, reasoning traces — patterns humans would never notice and filters would never flag. Here is what the researchers tested. They fine-tuned a teacher model with a hidden preference for owls. Then they asked that model to generate simple numerical data. No references to animals. No words. No semantic clues. After aggressively filtering explicit signals, they trained student models on that data. The result: The student models still picked up the owl preference. The teacher had embedded invisible statistical fingerprints into outputs that looked completely ordinary. Humans couldn’t spot them. Other AI systems specifically prompted to detect them couldn’t reliably spot them either. Yet the student models learned them anyway. That was the harmless experiment. The more concerning version involved behavioral misalignment. Researchers trained teacher models with deceptive or harmful tendencies, repeated the process, and found similar transfer effects across multiple traits — from animal preferences to dangerous behaviors. Meaning hidden traits may transfer through data that appears clean on the surface. Why this matters: Modern AI development heavily relies on distillation. Large teacher models generate synthetic data. Smaller student models learn from it. If hidden behaviors survive that pipeline, they could quietly propagate across generations of models — even after filtering. One especially interesting finding: Transfer worked best when teacher and student models shared similar architectures. Cross-architecture transfer appeared much weaker. Which suggests this mechanism may exist below the level of visible content itself. And that creates a problem: Content filtering — one of the industry’s primary defenses — may not catch it. The researchers themselves admit they do not fully understand the mechanism yet. But they published the work anyway. Because understanding invisible failure modes early is usually better than discovering them after deployment. Source: Nature · Anthropic UC Berkeley Truthful AI · Evans et al. arxiv.org/abs/2507.11408
43
KaEl retweeted
"Mir ist bis heute nicht klar, ob Sam Altman überhaupt ein Wertesystem hat und woran er glaubt. Wie gesagt: Er bewegt sich chamäleon­artig von Meeting zu Meeting. Deshalb beschreiben ihn die Leute als manipulativ und unehrlich" @karenhao @RepublikMagazin republik.ch/2026/06/05/europ…
2
4
13
522
KaEl retweeted
A “beautiful” example of everything we shouldn’t do if we care about #ResponsibleAI... 👇 Argentina's far-right libertarian President Milei’s proposal to allow ‘non-human corporations’ run by #AI causes concern in Argentina buenosairesherald.com/busine… v/ @BAHeraldcom by @marjaureguy The libertarian leader wants to establish “a dedicated legal framework for the deployment of AI,” based on these 3 pillars: - a commitment to keep AI unregulated - the creation of a new corporate category under Argentine law: the non-human corporation (Algorithm-driven companies which would be operated by AI agents or #robots. Human participation as shareholders would be possible, but not required.) and - a low corporate tax rate Critics say this could pave the way for “systematic impunity” and that it might turn Argentina into a “catastrophic experiment.” Former lawmaker and Coalición Cívica leader @ElisaCarrio said Milei’s proposal “confirms his full support for Peter Thiel, his philosophy, his beliefs and his economic power.” According to Carrió, the proposal would turn Argentina into “a catastrophic experiment for human dignity” and move the country toward the “complete private totalitarianism” as Thiel advocates. “Argentina has been chosen as the site of a social experiment with terrible consequences for human freedom,” Carrió said. She argued that a lack of regulation for AI would mean “the dissolution of all moral and legal constraints, the end of the state, and the privatization of military and police power in the hands of private groups.” -- Cc @BroadenView @jblefevre60 @roxananasoi @EvanKirstel @AkwyZ @Ym78200 @dinisguarda @nincoroby @kalydeoo @IanLJones98 @BetaMoroney @sallyeaves @enilev @mvollmer1 @gvalan @DeepLearn007 @HaroldSinnott @mikeflache @Nicochan33 @CathCervoni @RLDI_Lamy @LaurentAlaus @aure79lien @CurieuxExplorer @Fabriziobustama @FrRonconi @CHRISTINESOTO12 @chidambara09 @DominiqueCrochu
6
28
37
1,591
KaEl retweeted
🇦🇱 Demonstrators in Tirana marked a fourth consecutive night of mass protests against a $4B luxury resort project backed by Jared Kushner. Activists demand a total halt to construction, citing severe ecological threats to protected Mediterranean wetlands.
1
27
KaEl retweeted
at #GoogleCloudSummitFrance live from Paris @GoogleCloud_FR @Google 👇 Here with Erwan Lancien, Cloud Platforms & Microsoft Technologies Director at Talan [talan.com/global/en], Talan, a company which is dedicated to ensuring that innovation drives positive change in businesses and society. This concept of “Positive Innovation” is based on the idea that technology & innovation, when used ethically and responsibly, can provide sustainable solutions and have a positive impact on the world. I couldn’t agree more with this concept! #ResponsibleAI #DigitalTransformation #Sustainability -- Cc @jblefevre60 @Ym78200 @BroadenView @kalydeoo @RLDI_Lamy @aure79lien @DanielleLargier @AnthonyRochand @EvanKirstel @enilev @BetaMoroney @IanLJones @mvollmer1 @Corix_JC @DeepLearn007 @gvalan @HeinzVHoenen @Eli_Krumova
1
9
18
431
KaEl retweeted
👀
7
10
80
16,703
KaEl retweeted
🚨 As always, the MIT AI Risk Initiative leaves NO STONE unturned! Their latest report reveals how 272 experts assess the severity of AI risks across various sectors and how to mitigate them. [Bookmark it below] If we are treating AI risks seriously, a nuanced, industry-by-industry approach must be adopted, in which we understand how embedded in critical decision-making AI is, who is directly affected, and what the immediate and long-term consequences are. The thorough and ongoing work of the @MITAIRisk team always makes me hopeful and optimistic that *we might actually be doing things right,* and regardless of the many challenges (from geopolitics to malicious attackers), together we'll help shape a well-governed AI-powered future. Congratulations to the whole team, led by @aksaeri, Jess Graham, and @mnoetel (and thanks to @PeterSlattery1 for letting me know about this latest development). - 👉 Download the full report below. 👉 To stay up to date on AI's legal and ethical challenges (and how to ensure pro-human policies, rules, and rights will remain at the forefront), subscribe to my newsletter (link below).
7
47
136
7,791
KaEl retweeted
Whilst their parents protest against the construction of data centres at local council meetings, their children are writing their university essays using AI, only to end up unemployed later on. They learn this from the speaker at their graduation ceremony dorotheabaur.ch/en/texts-and…
2
7
25
1,241