Meta can now secretly watermark deepfake audio Researchers have found a way to imperceptibly watermark fake audio
From chaos to clarity with AI-driven categorization I’ll show you how I used Replicate’s API to push my AI model categorizer from 22% to 78% accuracy.
Teaching AI to see websites like a human made it more capable Tencent's AI can now complete the majority of its tasks on Google, Amazon, and Wikipedia
"Flow engineering" doubles code generation accuracy (19% vs 44%) The authors of a new paper present an approach that "intensifies" code generation.
Google's new LLM doctor is right way more often than a real doctor The LLM's differential diagnosis list had the correct diagnosis 59% of the time, vs. 34% for human doctors.
All LLM improvements are just task contamination? Current benchmarks are probably overestimating the true capabilities of LLMs
Prompting with unified diffs makes GPT-4 write much better code A developer for an open-source paired-programming tool discovered the trick