Skip to Content
Artificial intelligence

OpenAI’s fiction-spewing AI is learning to generate images

By training GPT-2 on pixels instead of words, the model can accept half an image and predict how to complete it.
July 16, 2020
Ben Barry / OpenAI

In February of last year, the San Francisco–based research lab OpenAI announced that its AI system could now write convincing passages of English. Feed the beginning of a sentence or paragraph into GPT-2, as it was called, and it could continue the thought for as long as an essay with almost human-like coherence.

Now, the lab is exploring what would happen if the same algorithm were instead fed part of an image. The results, which were given an honorable mention for best paper at this week’s International Conference on Machine Learning, open up a new avenue for image generation, ripe with opportunity and consequences.

At its core, GPT-2 is a powerful prediction engine. It learned to grasp the structure of the English language by looking at billions of examples of words, sentences, and paragraphs, scraped from the corners of the internet. With that structure, it could then manipulate words into new sentences by statistically predicting the order in which they should appear.

So researchers at OpenAI decided to swap the words for pixels and train the same algorithm on images in ImageNet, the most popular image bank for deep learning. Because the algorithm was designed to work with one-dimensional data (i.e., strings of text), they unfurled the images into a single sequence of pixels. They found that the new model, named iGPT, was still able to grasp the two-dimensional structures of the visual world. Given the sequence of pixels for the first half of an image, it could predict the second half in ways that a human would deem sensible.

Below, you can see a few examples. The left-most column is the input, the right-most column is the original, and the middle columns are iGPT’s predicted completions. (See more examples here.)

OPENAI

The results are startlingly impressive and demonstrate a new path for using unsupervised learning, which trains on unlabeled data, in the development of computer vision systems. While early computer vision systems in the mid-2000s trialed such techniques before, they fell out of favor as supervised learning, which uses labeled data, proved far more successful. The benefit of unsupervised learning, however, is that it allows an AI system to learn about the world without a human filter, and significantly reduces the manual labor of labeling data.

The fact that iGPT uses the same algorithm as GPT-2 also shows its promising adaptability. This is in line with OpenAI’s ultimate ambition to achieve more generalizable machine intelligence.

At the same time, the method presents a concerning new way to create deepfake images. Generative adversarial networks, the most common category of algorithms used to create deepfakes in the past, must be trained on highly curated data. If you want to get a GAN to generate a face, for example, its training data should only include faces. iGPT, by contrast, simply learns enough of the structure of the visual world across millions and billions of examples to spit out images that could feasibly exist within it. While training the model is still computationally expensive, offering a natural barrier to its access, that may not be the case for long.

OpenAI did not grant an interview request, but in an internal policy team meeting that MIT Technology Review attended last year, its policy director, Jack Clark, mused about the future risks of GPT-style generation, including what would happen if it were applied to images. “Video is coming,” he said, projecting where he saw the field’s research trajectory going. “In probably five years, you’ll have conditional video generation over a five- to 10-second horizon." He then proceeded to describe what he imagined: you'd feed in a photo of a politician and an explosion next to them, and it would generate a likely output of that politician being killed.

Update: This article has been updated to remove the name of the politician in the hypothetical scenario described at the end.

Deep Dive

Artificial intelligence

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

Providing the right products at the right time with machine learning

Amid shifting customer needs, CPG enterprises look to machine learning to bolster their data strategy, says global head of MLOps and platforms at Kraft Heinz Company, Jorge Balestra.

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.