Google is in too much of a hurry on AI search

Google is deep in the AI arms race now.

The search giant made a flurry of announcements this week about enhancing its main products with artificial intelligence. In line with Bloomberg reporting that the company was rushing to stuff generative AI into as many services as it could, the company revealed at I/O, its annual developers conference, that new technology for generating content was coming to Gmail, Google Docs, Google Maps, Google Photos and more.

It also unveiled the biggest change to Google Search in years, the awkwardly named “search generative experience,” essentially a single, AI-generated answer to queries that will sit above the usual search results of ads and links.

Google has been making one thing clear: As much as the company is under pressure to be careful about how it deploys this potent technology, it is forging ahead, and quickly. But it ought to be careful about moving so fast that it undermines its standing with both users and advertisers.

Executives haven’t been shy about telegraphing their priorities. Note how Google Senior Vice President Prabhakar Raghavan answered a question from Bloomberg News about allowing outside scrutiny of its AI. He said Google would contribute to the scientific community but “look even harder at moving things into product, rapidly.”

When OpenAI launched ChatGPT six months ago, Google looked like a laggard. Now Sundar Pichai, chief executive officer of Google parent Alphabet Inc., has pledged to “reimagine all our core products” with generative AI. Some observers are impressed. Citigroup analysts said in a report that they came away from the event “confident in Google’s generative AI strategy.” (Bloomberg has announced its own language model for finance, which will likely compete with OpenAI’s GPT-4.)

On the plus side, Google does appear stronger in the cloud businesses, where it trails Amazon Inc.’s AWS and Microsoft Corp.’s Azure. Companies can use Google’s Duet AI service to build their apps without any serious coding knowledge. Gmail users can get their emails drafted by AI. A text-to-speech tool from Google called Chirp is taking drive-thru orders for Wendy’s. These features could give Google a boost in an industry that’s notoriously entrenched thanks to legacy contracts with companies.

But the real question mark is over planned changes to search, Google’s biggest product.

When users type in a query, Google’s search engine will soon churn out AI answers synthesized from other text across the web. Here’s an example of a query about visiting a national park with kids and a dog:

The answer is displayed at the top, and on the left are links to sites from which it drew its answer. But this will look very different on the smaller screen of a mobile device. Users will need to scroll down to see those sources, never mind other sites that might be useful to their search.

That should worry both Google’s users and paying customers like advertisers and website publishers. More than 60% of Google searches in the US occur on mobile phones. That means for most people, Google’s AI answer will take up most of the phone screen. Will people keep scrolling around, looking for citations to tap? Probably not.

When asked about whether people will visit those sources, a Google executive told the Washington Post, “I really, genuinely think that users want to know where their information comes from.”

I don’t buy it. Search users glance at the links to double-check that they look legitimate, but most probably won’t click on them. That takes a few extra seconds of scrolling and loading, an eternity online. There is, after all, a reason for the joke that the best place to store a dead body is page two of Google’s search results. Most people can’t be bothered to keep searching beyond the first list of links.

“It’s safe to bet on laziness,” Silicon Valley luminary Paul Graham tweeted on Tuesday, the day before Google’s announcement. He said generative AI was “one of the biggest bets on laziness in history.”

Graham is right. And the price could be the ire of advertising customers if Google’s AI search becomes so good that users stop clicking around on links. It could also risk eroding the trust of consumers if the tool starts generating too many erroneous answers, a problem OpenAI’s ChatGPT, Microsoft’s Bing and Google’s own Bard have all encountered. In a sign Google has yet to even nail down the basics, a tech journalist who asked Google’s new AI search for a chocolate chip cookie recipe got one with no chocolate chips.

That illustrates what uncertain times these are for Google. Company executives have been saying over and over that AI search is experimental. But for Google and hundreds of millions of users, taking part in the “experiment” might feel like traveling in a car whose steering wheel is being designed on the fly. We are in for a wild ride.

More From Bloomberg Opinion:

• AI Music Brings the Sound of Scammers to Spotify: Lionel Laurent
• A Drone Superhighway Gets the UK Where It Needs to Go: Dave Lee
• AI’s ‘Godfather’ Should Have Spoken Up Sooner: Parmy Olson

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

Parmy Olson is a Bloomberg Opinion columnist covering technology. A former reporter for the Wall Street Journal and Forbes, she is author of “We Are Anonymous.” Bloomberg