💡 University of Florida’s Light-Powered AI Chip
Researchers at UF just showed off a chip that ditches electricity for light when running convolution ops. Early trials say it’s close to 100× more energy efficient - yet still hitting around 98% accuracy on stuff like digit recognition. It feels oddly sci-fi, but very real: swapping electrons for photons is a pretty wild pivot.
🔗 Read more
🦠 NYU Flags “Autonomous Ransomware”
NYU’s team cooked up a proof-of-concept where AI basically runs a ransomware campaign solo - scanning, picking targets, generating payloads, even sending the shakedown messages. They’re calling it “Ransomware 3.0.” It’s still lab-level work, but the idea is unsettling. Security pros repeat the classics: patch, backup, don’t slack.
🔗 Read more
🛡️ Booking.com vs. Fraudsters
Booking.com has been quietly beefing up its anti-fraud systems with AI. The setup now screens millions of bookings in the background, sniffing out suspicious behavior before it reaches users. It’s not exactly headline-grabbing AI magic, but it’s the kind of boringly crucial defense you’re glad someone’s handling.
🔗 Read more
🎙️ Alibaba’s “Qwen” Model for Speech
Alibaba rolled out its newest large model, “Qwen,” with a heavy push into speech-to-text. They’re boasting major gains in accuracy across multiple languages. For anyone drowning in meeting notes and endless call transcripts, this isn’t a toy - it’s oxygen.
🔗 Read more
⚖️ Indian Lawyers Try AI on the Quiet
In India, lawyers - whether from heavyweight firms or solo practices - are sliding AI into their everyday workflows. It’s not about replacing them, more like supercharging research and drafting. One attorney joked it feels like hiring a junior clerk who never clocks out.
🔗 Read more