By the time you read this, hundreds of millions of web surfers will have had their questions answered by robots — and it might not be a moment too soon. The classic search and click made ubiquitous by the Google behemoth is getting a profound reset thanks to artificial intelligence.
The question is which bot will prevail. For the first time, AI competitors like ChatGPT, Bard, or Bing are giving Google a run for its money. These new search-adjacent bots aren’t just looking up information but generating it for the user.
This opens up whole new search-adjacent markets. It could change how we do everything from seeking information (ChatGPT can write poetry and essays in the style of your favorite author) to productivity tools for business, such as writing presentations or even coding. The potential is so great that Washington is now talking about it. Senators, regulators, and White House officials have held numerous meetings on how to govern AI.
The first significant shift is in how people search. In a Pew Internet survey 2012, search was second only to email among the activities most often done on any given day by online adults. By 2016, that figure had grown to more than eight in ten.
Moreover, people are more confident today that they can find what they want compared to 2004. For example, a 2012 survey found that 29% of search engine users said they “always” find what they need, while in 2004, that number was just 17%.
Google, however, has taken steps to defend its search throne. The Mountain View company has incorporated generative AI into its bots, which generate text-based answers to queries. The results, dubbed Knowledge Graph Cards, are displayed as the most relevant search results in the sidebar. In addition, Microsoft has loaded generative AI into its Bing search engine and invested in an Anthropic startup to develop its own generative AI.
Nevertheless, the new AI search bots can still be unreliable. The bots can’t always distinguish between fact and fiction and sometimes develop deceptively effective responses. They can also give false answers if the question is open-ended or vague and fail to grasp the context of the query.
The good news is that these early bots are improving. ChatGPT’s underlying AI, GPT-3.5, is constantly being updated, while Bard uses Google’s LaMDA model and will eventually get a makeover to PaLM. These improvements are likely to improve the results.
And the competition is far from over. Microsoft is preparing to launch its generative AI search engine. At the same time, Alphabet’s stock price tumbled on February 8 after the company announced it would include generative AI in searches within weeks. Microsoft CEO Satya Nadella is betting that the shift will boost search traffic, which is already growing faster than before.