📰 What Happened
A company in Palo Alto called Elorian has released its first AI model, named Glimpse. The startup raised $55 million before it even had a product. Glimpse is built to look at an image again and again while it thinks. It can zoom into part of a photo, pull frames from a video, or run code to measure what it sees. The chat version is waitlist-only for now.
🔍 The Backstory
Most AI models look at an image once, turn it into text, then reason about it, so they often miss details. Elorian's chief executive, Andrew Dai, once showed a top model a finished board game and asked who was winning. It could not say. Zooming during reasoning is not brand new, but Elorian built its whole model around it. The company says a small team trained Glimpse on a few hundred GPUs, and it beats much bigger models at reading charts.
🎯 Why It Matters
Models that truly understand images could help doctors study scans and robots see the world. Making AI cheaper to run would help small firms too.

A company in Palo Alto called Elorian has released its first AI model, named Glimpse. The startup raised $55 million before it even had a product. Glimpse is built to look at an image again and again while it thinks. It can zoom into part of a photo, pull frames from a video, or run code to measure what it sees. The chat version is waitlist-only for now.

Most AI models look at an image once, turn it into text, then reason about it, so they often miss details. Elorian's chief executive, Andrew Dai, once showed a top model a finished board game and asked who was winning. It could not say. Zooming during reasoning is not brand new, but Elorian built its whole model around it. The company says a small team trained Glimpse on a few hundred GPUs, and it beats much bigger models at reading charts.

Models that truly understand images could help doctors study scans and robots see the world. Making AI cheaper to run would help small firms too.

📰 Source: News Source
techcrunch.com ↗
Was this article useful?