This category is about Explainable and Trustworthy AI
A Stanford team led by Fei-Fei Li reveals that frontier multimodal models—GPT-5, Gemini-3-Pro, Claude Opus 4.5—confidently describe images that were never provided, achieving top benchmark scores without visual input. The implications for medical AI are alarming.
The dream of AI has always been to match human efficiency—learning a new concept from a single glance. In my Istanbul lab, I recently tackled the reproduction of the paper “Learning Conditional Class Dependencies: A Breakthrough in Few-Shot Classification.” Standard models treat every class as an isolated island. If a model sees a “Scooter” for the … Read more
We’ve all seen the benchmarks. The new “Reasoning” models (like the o1 series or fine-tuned Llama-3 variants) claim to possess human-like logic. But after building my dual-RTX 4080 lab and running these models on bare-metal Ubuntu, I’ve started to see the cracks in the mirror. Is it true “System 2” thinking, or just an incredibly … Read more