Yhn first seen 7 h ago, last 5 min ago, peak #20

AI models fall short of human algorithmic innovation

Original: Recent AI models struggled to match a human algorithmic innovation

A new study by Epoch AI finds that recent AI models struggled to match a human algorithmic innovation when tested on reproducing novel research ideas. The findings suggest frontier models still lag behind humans on tasks requiring genuine creative problem-solving, even as they excel at routine coding and reasoning benchmarks.

Why now: The study touches the live debate over how close AI really is to matching human research ability.

Epoch AIOpenAIAnthropicGoogle DeepMind

Open on hn →

Rank over time, top of the chart is #1. 30 snapshots from 7 h ago to 5 min ago.

Evidence

• Recent AI models struggled to match a human algorithmic innovation · merksittich · 82

API: https://socialmediatrends-api.osmike.com/v1/trends/1772371