Skip to Content
Artificial intelligence

This Fun AI Tutorial Highlights the Limits of Deep Learning

October 9, 2017

Sure, neural networks can easily classify images—but they still don’t really understand what they see without human intervention. That much is made plain in Google’s new AI tutorial, called Teachable Machine, which was brought to our attention by the Verge. You can watch it in action in the video above, or try it out for yourself. It’s fun to play with: you train an AI to classify images by showing it objects via your webcam, which it then associates with a GIF or sound that it plays on demand when shown those objects again.

But that sentence pretty much says it all: the kind of deep neural network that’s being used to power this simple example—and, arguably, most machine learning that's currently in use around the world—is, essentially, performing advanced pattern recognition, but little else. Beyond identifying, say, your waving hand, and differentiating it from a view of your wall, the AI has no clue about what it’s seeing, unless it’s provided with huge quantities of labeled data. And it certainly can't abstract further, to work out the kinds of deeper meaning you might associate with a seemingly inanimate object being held in front of a web cam.

For a deeper dive into the limitations of deep learning, read our recent article, “Is AI Riding a One-Trick Pony?

Deep Dive

Artificial intelligence

This new data poisoning tool lets artists fight back against generative AI

The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models. 

Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist

An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.

Driving companywide efficiencies with AI

Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.

Unpacking the hype around OpenAI’s rumored new Q* model

If OpenAI's new model can solve grade-school math, it could pave the way for more powerful systems.

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 with a list of newsletters you’d like to receive.