Ben Cowley

Researchers set out to shrink an AI model that acts like a part of the brain's visual system.

When they started, the model used 60 million variables.

But Ben Cowley of Cold Spring Harbor Laboratory says the team was able to make a streamlined version that had just 10,000 variables.

And that is incredibly small.

This is something we could send in a tweet or an email.

Cowley says the key was finding redundant or unnecessary elements and applying statistical techniques like those used to compress digital photos.

The results suggest that a type of AI system that now requires a

torrent of electricity could get by with just a trickle.

John Hamilton, NPR News.

Researchers set out to shrink an AI model that acts like a part of the brain's visual system.

When they started, the model used 60 million variables.

But Ben Cowley of Cold Spring Harbor Laboratory says the team was able to make a streamlined version that had just 10,000 variables.

And that is incredibly small.

This is something we could send in a tweet or an email.

Cowley says the key was finding redundant or unnecessary elements and applying statistical techniques like those used to compress digital photos.

The results suggest that a type of AI system that now requires a torrent of electricity could get by with just a trickle.

John Hamilton, NPR News.