Chrome’s dev team had to learn AI from the ground up to stay on top of trends

Tech moves fast. Fewer than two years ago, few could have imagined the ways in which AI would impact our lives. In 20 years, perhaps we’ll think of the launch of ChatGPT as a fault line in the zeitgeist of humanity, that is, the world before advanced large language models, and the world after. Few big companies moved as fast as Google did to embrace the potential of LLMs andintegrate them into its products. Typically, we only get to see these changes after the fact, when they’re being pushed out to the public, but occasionally, Google gives us a bit more.

Chrome AI features illustration

Set your student up for success with these great Chromebooks

Best-Chromebooks.-for-Students

What’s behind the curtain?

In arecent post on its The Keyword blog, Google shares a glimpse behind the curtain to see some of what went into making its Chrome AI tools work. Specifically, we get to learn a bit more about how the AI-powered tab organizer, AI themes, and the “Help me write” functionalities came to be. Adriana Porter Felt, Chrome’s engineering director, says that ideas for AI features were solicited from the Chrome team before these three features were settled upon.

help-me-write-chrome-ai

In the post, Porter Felt notes that the internet of today is vastly different from when Chrome launched back in 2008. Back then, the internet was largely information based, and Chrome designers wanted to “get content in front of a user and get the browser out of the way as much as possible.” To some extent that’s still true, but the internet has become so much more than a source of information that Chrome’s designers want to make Chrome more of an unobtrusive browsing assistant.

customize-chrome-ai-anim

The Chrome team’s pursuit of AI sounds very much like it was a game of catch-up. Porter Felt said “We had to learn not only how this technology works but also how to turn it into a product people can use.” The testing of the new features in particular was tricky for the team. Testing usually consists of making sure the same input always produces the same output, but generative AI doesn’t work like that. With these new features, the Chrome team had to look at things like creating the proper tone or length for an AI-generated text response. Porter Felt even goes into how her team worked with Google’s emoji team (Google has an entire team dedicated to emoji?) to use AI to select an appropriate emoji for AI-sorted tab groups.

organize-similar-tabs-chrome-ai-anim

If you run into issues with websites loading or formatting correctly, try clearing your cache

Most technology companies, Google included, aren’t too keen to share the development process of how their products are made, so this rare glimpse behind the scenes is refreshing. Now if we can just convince Spotify to explain how it designed its randomization algorithm…