Google’s New AI is Already Outperforming Human Scientists

The system found 40 novel methods in genomics and beat the CDC’s best COVID forecasts. We’re entering a new era of research.

Photo by Karollyne Videira Hubert on Unsplash

I just found a research paper that really surprised me. Not in a “Hey, that’s a neat trick” kind of way, but in a “Wait, did that just happen?” way.

It’s a paper from Google titled “An AI system to help scientists write expert-level empirical software,” and honestly, the title is the most boring thing about it.

What it describes is an AI that is already stepping into the role of a brilliant, tireless, and creative research scientist. It’s not just automating tasks; it’s making genuine discoveries.

I’ve been thinking about this nonstop. Is this one of those “feel the AGI” moments Sam Altman talks about? I think for scientists, this might be the moment.

Let’s break down what this AI does, how it works, and why it feels like we’ve quietly crossed a major threshold without anyone noticing.

The Big Problem That No One Talks About

First, why does this AI even need to exist?