Skip to Content
MIT Technology Review

Here are the secrets to useful AI, according to designers

Thinking about how us meatbots actually work could deliver AI that’s more useful and fair.

Secrets to good AI: Designers at Google spent three years helping build AI for the firm’s Clips wearable camera, and they published a blog post about what they learned. They say the best AIs solve a real human problem, do things in a way that makes sense to humans, and build trust with users. Easy, right?

And how to build ’em: A new tool called Cortex could get designers more involved in building AI. Its graphical user interface makes building complex algorithms simpler and should help non-nerds create AI that’s helpful for humans.

Why it matters: Many AI problems, like bias, stem from the fact that such a narrow group of people are building the technology. Opening up design and focusing on end users could make AI both better and fairer.