When OpenAI debuted its ChatGPT bot earlier this year, the startup attracted enormous attention and launched a wave of interest in artificial intelligence.
But Google, which has long led the way in rolling out new technologies to support its search business, was caught flat-footed. Now, as the tech giant celebrates its 25th anniversary amid a race to catch up, the engineers at Google’s facility in Kendall Square are playing a major role in how to best incorporate AI into search.
Their efforts may very well determine whether Google maintains its grip on the $280 billion global search advertising market.
Specifically, the Cambridge team is working on ways for Google’s search engine to produce much more detailed answers to the user’s initial query expressed in text, photos, and videos.
“You get a much more context-rich experience,” said Tali Sason, director of search engineering and Cambridge site leader. “I think that’s really exciting. And Cambridge is really at the forefront of what that will look like.”
At issue is the emergence of generative AI, which powers ChatGPT.
Generative AI can create text, images, and other media by learning the patterns and structures of data fed into the system and then produce something original that resembles that data. The program also allows people to conduct much more interactive, complex conversations with the bot.
Microsoft, a major investor in OpenAI, integrated the startup’s technology into its Bing search engine and released it in February. Google debuted its Bard chatbot in late March and then rolled out its generative AI-powered search engine in late May.
“Over the last five years, we’ve essentially experienced decades of advancement in this field,” said Matthew Gray, director of engineering for search and site leader. “We have this sort of radical new set of capabilities. There’s a lot of cool tricks and we need to figure out how we turn that into real user value, real value for the world, real ability to sort of understand and do new things.”
With the new AI advances, users can query the search engine with more complicated questions and then possibly see results that are not only more accurate but also contain greater depth and sweep.
For example, a person wants to plan a trip to New York. Currently, the user would have to type individual queries in the search engine and then click on related individual links. With new AI advancements, one query can produce a personalized vacation plan, including best times to travel by train, suggestions for hotels depending on different budgets and dates of travel, and activities related to weather and time of year.
The Kendall Square campus, which opened in 2007, has long led Google’s efforts to advance search. Gray was part of the office’s first group of engineers, who worked to improve the quality of searches, especially how the company’s algorithms ranked websites.
Today, more than 2,100 people work in Cambridge.
One of the office’s most visible achievements was the creation of the popular People Also Ask feature for Google search, which provides users with additional questions related to their original search query and offers quick answers. In many ways, People Also Ask was the precursor to what Google is trying to do now with AI: give users information that they might find useful beyond their original search.
But despite Google’s historical dominance of search, Microsoft beat the company to the AI punch.
Google had previously invested in AI, but “they would have never done more with it had it not been for ChatGPT,” said Michael Jacobides, a professor of strategy and innovation at London Business School. “They didn’t know what they could do with it.”
Google doesn’t have a lot of room for error. Despite years of experimentation and investments into new technologies and markets, the company is still primarily a search firm, at least from a financial perspective.
Last year, Google poured nearly $40 billion into research and development, about 14 percent of the company’s total revenue of $162.45 billion, according to company filings. Yet search ads accounted for 58.1 percent of total revenue. In other words, not much has changed from the time Google went public in 2007: Search advertising still dominates the company’s balance sheet.
“There is a lot of exploration that goes on at Google,” said Michael Cusumano, a professor of management and deputy dean at MIT Sloan School of Management. “It just did not have a large impact on its business model. Google is kind of a one trick pony.”
That’s why Microsoft’s early lead on integrating more advanced AI into search is so concerning, at least to Google’s shareholders. The company can’t afford to lose its edge in search.
“AI could represent an existential threat,” Jacobides said.
Gray acknowledges that AI can fundamentally alter search. But he says the Kendall Square team is working hard to preserve Google’s leadership with this new technology.
“These new AI experiences open up a ton of opportunity,” Gray said. “We get to be on that cutting edge.”
For example, the team is trying to figure out whether users want to use a chatbot for searches or stick with typing queries into a search engine. And given intense regulatory scrutiny of AI, Google wants to make sure it can roll out the technology in a safe and effective way, Sason says.
“This idea of move fast and break things, that is a big part of that we have to be a lot more careful about,” Sason said. “It’s really important to us that we don’t break the user experience and that we can at all times make sure that user’s information is being protected, that users are having a trustworthy and safe experience.”
Thomas Lee can be reached at firstname.lastname@example.org.