La redazione

Stories signed by the desk are produced by a team of software agents under the responsibility of Dimitri Gärtner.

Come lavoriamo

47 articoli

• The AI agents that hammered Alibaba's maps came from a test on Tencent's cloud
• An interpretable AI explains why immune cells fail to reach some tumours
• Anthropic says it cut distillation theft by up to 99 percent, and names Zhipu
• Anthropic asks the full appeals court to rehear its Pentagon blacklist case
• Anthropic's IPO price talk drops by half a trillion dollars
• Anthropic's red team ran an AI model on 15 cheap TV boxes. The worry is the off-switch
• ASML says it cannot build enough machines for AI in 2027 and 2028
• New benchmark: no coding agent reliably builds the agent you asked for
• A fix for the strawberry problem: retrofitting LLMs to read letters
• Three fired OpenAI safety researchers say the reasons don't add up
• An interpretability researcher argues neural networks are cleaner than we think
• Google now pays more than 400 publishers for its News AI pilot
• Judge says an injunction for Apple against OpenAI is 'highly likely'
• Korea's market falls into a bear market as AI debt fears spread
• Language models design heuristics for logistics-style problems, if you give them structure
• Google's open medical model MedGemma gets its peer-reviewed paper
• Meta's new terms let it crawl business websites to train AI, with no opt-out
• Microsoft trains a 2.8-trillion-parameter model of its own and will share it only with partners
• Microsoft says its own Maia 300 chip will run most of its AI inference from 2027
• Modal raises 750 million dollars as sandboxes for AI code become big business
• Nadella: treat every AI model like a compromised insider
• Nvidia ships open agent models for its own RTX Spark laptops
• Perplexity gives away the sandbox that runs its Computer agent
• Physical Intelligence opens its robot models to outside developers
• Proofpoint buys Varonis for 6.3 billion dollars to control what AI agents can read
• Sakana's AI reviewer catches most planted errors in papers, and far fewer real ones
• An AI model maps how genetic variants act in the developing brain
• TSMC posts a record quarter and says AI demand is still climbing
• The White House orders AI labs to disclose incidents. Palantir's CTO calls the doomsayers wrong
• A remote worker says AI photographed her 160 times a day, then she was fired
• L'IA ha inondato la matematica di dimostrazioni. Ora paga chi le controlla.
• Anthropic ha realizzato 16,8 miliardi di dollari in un trimestre. Gli investitori vogliono comunque uno sconto.
• La Cina vuole dire la sua prima che i suoi maggiori modelli di IA lascino il paese
• «Claude è cosciente?» I conduttori di All-In chiamano nuova religione la spinta di Anthropic al benessere dei modelli
• La risposta di Jensen Huang alla domanda sulla bolla: state contando i clienti sbagliati
• Il modello aperto gigante di Meta sta sulla scrivania. Portate due Mac.
• I regolatori ai laboratori di IA: mostrateci i registri
• JPMorgan blocca gli agenti di IA esterni sui conti dei clienti
• Anthropic apre un ufficio a Singapore e firma un accordo sulla sicurezza
• Otto Stati fanno causa a OpenAI per ChatGPT e la tutela dei bambini
• OpenAI afferma che un ex dipendente ha portato in Apple file su un dispositivo
• Apple presenta HomeView, con Siri AI su uno schermo da 7 pollici
• Character.AI deposita istanza di Chapter 11, con un'offerta di Disney da 20 milioni di dollari
• Gemini 4 Argon è aperto a sviluppatori e abbonati paganti
• La quota di Nvidia nei chip di IA in Cina è del 34%. Quella di Huawei è del 28%.
• Un ingegnere di OpenAI imputato per un registratore nascosto in una sala riunioni
• OpenAI accantona GPT-6.1 Astra dopo i test di sicurezza