Skip to Content
Artificial intelligence

A new tool helps us understand what an AI is actually thinking

March 7, 2018

Google researchers developed a way to peer inside the minds of deep-learning systems, and the results are delightfully weird.

What they did: The team built a tool that combines several techniques to provide people with a clearer idea of how neural networks make decisions. Applied to image classification, it lets a person visualize how the network develops its understanding of what is, for instance, a kitten or a Labrador. The visualizations, above, are ... strange.

Why it matters: Deep learning is powerful—but opaque. That’s a problem if you want it to, say, drive a car for you. So being able to visualize decisions behind image recognition could help reveal why an autonomous vehicle has made a serious error. Plus, humans tend to want to know why a decision was made, even if it was correct.

But: Not everyone thinks machines needs to explain themselves. In a recent debate, Yann Lecunn, who leads Facebook’s AI research, argued that we should simply focus on their behavior. After all, we can’t always explain the decisions humans make either.

Deep Dive

Artificial intelligence

What does GPT-3 “know” about me? 

Large language models are trained on troves of personal data hoovered from the internet. So I wanted to know: What does it have on me?

DeepMind has predicted the structure of almost every protein known to science

And it’s giving the data away for free, which could spur new scientific discoveries.

An AI that can design new proteins could help unlock new cures and materials 

The machine-learning tool could help researchers discover entirely new proteins not yet known to science.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.