When fake data is a good thing – how synthetic data trains AI to solve real problems

You have just completed a strenuous hike to the top of a mountain. You are exhausted but elated. The view of the city below is beautiful and you want to capture this moment on camera. But it’s already pretty dark and you’re not sure if you’ll get a good shot. Luckily, your phone has an AI-powered night mode that allows you to capture stunning photos even after the sun sets.

Here’s what you may not know: This night mode may have been trained on synthetic night images, computer-generated scenes that were never actually photographed.

As artificial intelligence researchers exhaust the supply of real data on the Internet and in digitized archives, they are increasingly turning to synthetic data, artificially generated examples that mimic real data. But this creates a paradox. In science, inventing data is a cardinal sin. Fake data and misinformation are already undermining trust in information online. So how can synthetic data be good? Is that just a polite euphemism for…