Skip to Content
Uncategorized

How Do You Turn a Dog into a Car? Change a Single Pixel.

October 30, 2017

Thank a new approach to spoofing image recognition AIs, developed by a team from Kyushu University in Japan, for that joke.

Trying to catch out AIs is a popular pastime for many researchers, and we've reported  machine-learning spoofs in the past. The general approach is to add features to images that will incorrectly trigger a neural network and have it identify what it sees as something else entirely.

The new research, published on the arXiv, describes an algorithm that can efficiently identify the best pixels to alter in order to confuse an AI into mislabeling a picture. By changing just one pixel in a 1,024-pixel image, the software can trick an AI about 74 percent of the time. That figure rises to around 87 percent if five pixels are tweaked.

You can see some examples above, which spoofed an AI into calling an airplane a dog, a frog a truck, and a horse a car, among other things.

As the Register points out, 1,024-pixel images are pretty damn small, which means that larger images would need hundreds of pixels tweaked. But the push to make AIs fall over using the smallest number of possible changes is interesting and worrying, in equal measure. As we've explained in the past, finding a way to protect AIs from these kinds of tricks is quite difficult, because we still we don’t truly understand the inner workings of deep neural networks.

Until we do, such hacks will be hard to avoid.

Deep Dive

Uncategorized

stock art of market data
stock art of market data

Maximize business value with data-driven strategies

Every organization is now collecting data, but few are truly data driven. Here are five ways data can transform your business.

image of library due date card on fire over black background
image of library due date card on fire over black background

The book ban movement has a chilling new tactic: harassing teachers on social media

Educators who stand up to conservative activists are being harassed and called “groomers” online, turning them into potential targets for real-world violence.

two images made by DALL-E 2
two images made by DALL-E 2

OpenAI is ready to sell DALL-E to its first million customers

But the company has had to rush out fixes to the image-making model’s worst flaws to do so.

Breaking new ground: Sustainability in Malaysia

How global companies in Greater Kuala Lumpur are testing new ideas to advance their sustainability goals.

Stay connected

Illustration by Rose WongIllustration 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.