Google’s new AI tools hint at where search is headed

It’s an interesting time in the world of online search: Microsoft is betting its LLM-powered, intriguing-yet-sometimes-disturbing Bing chatbot will help it level the playing field on which Google currently enjoys around 85% of market share, according to Statista. Meanwhile, Google maintains that its own chatbot, Bard, isn’t a replacement for a  search engine.

Against that backdrop, the eyes of the tech world turned to Mountain View on Wednesday for Google I/O, Google’s annual developer conference, hoping to glean intel on its plans for keeping up with the rapid pace of the AI search wars.

The pitch for yet another new era in search, in the form of an AI-powered tool called Search Generative Experience (SGE), followed a somewhat different strategy from the one reportedly used to launch Bard: Move fast but not too fast, temper consumer expectations, and emphasize a “responsible” approach.

SGE, a search powered by large language models including Google’s new general-use model, PaLM2, is an experimental tool that will be available to users who sign up for Search Labs, a new program for consumers who want to try out Google’s latest experiments.

Search users who opt in to SGE can use “conversational mode” and see “AI-powered snapshots” that summarize search results.

Shoptalk

One of the key use cases Google is testing for SGE is shopping, which Senior Director of Product for Consumer Shopping Lilian Rincon described as one of the company’s “hero verticals for search,” with over a billion sessions daily.

“Our longer-term goal is to provide more natural and intuitive ways of getting product recommendations, and making these purchase decisions just like a personal shopper, or a knowledgeable store associate might help you when you go to the mall,” Rincon explained.

This experimental “shopping companion” is built on Google’s Shopping Graph, a dataset that includes billions of product listings and powers the Google shopping experience that exists today.

In its current iteration, SGE is most useful for what Rincon called “complex purchases”—those that may require more consideration, like a bike, for example. In the future, a “more advanced iteration” could help shoppers in apparel, which is one of Google’s most-searched shopping categories, she added.

Drip feed

With Microsoft sprinting back into search-engine relevance via Bing and OpenAI posting record-breaking user growth, Google has struggled to balance the Silicon Valley adage of “move fast and break things” with CEO Sundar Pichai’s emphasis on responsible AI development (evidence of this tension can be found in the Bard rollout, which has been criticized as “rushed”).

With SGE, Google appears to have leaned toward “slow and steady.” According to Rincon, the tool has been in the works “for a few years,” but the restrained rollout seems like an acknowledgement that AI search is a completely new experience both for consumers and brands.

The question of when SGE will graduate out of Search Labs and into Google’s core search capabilities is one Rincon said the company will be “working through” in coming months, adding that they’re waiting on user feedback in addition to internal user experience research.

“We’ve taken a lot of care and tried to be really responsible about the things we’re launching,” she said, pointing to the constraints that differentiate SGE from other search chatbots. “We really aim for the responses to be objective, neutral, and not provide personal opinions.”

“[Search Labs] is a new tool in Google’s belt,” Rincon added, one that aligns with its “responsible development” approach to innovation, letting the company “get feedback and iterate quickly before just putting this on the 1 billion users that we get every day,” she said.