1/8 You know that frustrating feeling when you teach your smart assistant something, and the next day it's completely forgotten? It's because most AI models are like printed books: their knowledge is vast but frozen in time. And that has been AI's biggest weakness. Until now.

Is your LLM-based multi-agent system actually coordinating? That’s the question behind this paper. They use information theory to tell the difference between a pile of chatbots and a true collective intelligence. They introduce a clean measurement loop. First, test if the

AI companies are racing to build superintelligence...without knowing how to control it. Let's take our future back from Big Tech. Join the call for a ban on superintelligence, until it can be done safely with public buy-in. #KeepTheFutureHuman share.superintelligence-statement.org/s/giYSlN5s

In Science, researchers report the observation of dendritic nanotubes in brain tissue from mice and humans. The discovery suggests that the current understanding of the brain’s organization may be incomplete, overlooking a hidden layer of connectivity. Learn more in a new

In principle, a comparison of the genAI and (e.g.) AlphaZero approaches: there are tasks for which the "bigger brain" of genAI with stochastic pattern-based activity is suitable, and others for which a much smaller AI system working with problem logic is more appropriate. The

Just a month later and... 🇪🇺 ChatControl is back! Now they're trying to pass an even more far reaching ChatControl law through the back door, in a form even more intrusive than the originally rejected plan, without needing any of the EU countries votes The new proposal: -

Dnes som v Nitre prednášal na konferencii venujúcej sa legalizovaniu a deeskalácii vojen a ozbrojených konfliktov: frcth.uniba.sk/detail-aktuali… Kompletná prezentácia mojej prednášky na tému "Etika a dynamika využívania systémov AI v súčasných ozbrojených konfliktoch" je k dispozícii

🚨 THE CHINA MODEL IS HERE — AND THIS IS WHAT IT LOOKS LIKE A chilling look inside Shanghai’s AI-powered surveillance “Urban Brain” system — the blueprint for the social credit grid spreading across China… and soon, the West. In Pudong, a single control room can monitor every

Ako priamy a aktívny účastník udalostí z novembra 1989 by som odkazované vyhlásenie naplno podpísal: marker.sk/clanky/697/spo…

I just read a paper that completely broke my brain. It describes a system that solved an AI task with over 1,000,000 sequential steps... with ZERO errors. Using AI models that are known to be flaky and make mistakes. How is that even possible? 🤯 We all know LLMs have an

My most amusing interaction was where the model (I think I was given some earlier version with a stale system prompt) refused to believe me that it is 2025 and kept inventing reasons why I must be trying to trick it or playing some elaborate joke on it. I kept giving it images

A unified theoretical framework showing the fundamental limits of LLMs. Discusses hallucination, context compression, reasoning degradation- rooted in computability, information theory, and learning constraints. Nice read to understand limits in LLMs even under scaling. Very

Včera som prednášal v rámci 6. ročníka konferencie 3xKaM, ktorá bola venovaná téme: „Duševné zdravie mladých v digitálnej ére: prevencia a etické výzvy AI“ vudpap.sk/podporne-cinno… Kompletná prezentácia mojej prednášky na tému "Čo sme si objednali a čo v skutočnosti dostávame..."

Rada EÚ schválila pozmenený návrh #ChatControl, ktorý teraz mieri do EP. Rada EÚ schválila návrh nariadenia, ktoré umožňuje plošne skenovať a analyzovať súkromnú komunikáciu (ak je E2E šifrovaná, ešte pred jej zašifrovaním). Toto nariadenie - ak nadobudne účinnosť - má potenciál

Metodický deň pre učiteľov náboženstva a náboženskej výchovy na tému: AI (umelá inteligencia) - etické aspekty a potenciálne riziká pre mladých frcth.uniba.sk/detail-aktuali…

S adventom prichádzajú aj rorátne sv. omše - väčšinou slávim večerné, no raz do týždňa i takto ráno...

Výborný článok od Matúša Jablonského - hodný zamyslenia... standard.sk/1087975/preco-…