AI arms race kicks off as Google, Microsoft make dueling announcements

Google offered a glimpse of its new artificial intelligence chatbot search tool on Wednesday at a presentation that sought to underscore its prowess in both search engine and AI tech, a day after archrival Microsoft unveiled its own search chatbot aimed at eroding Google’s dominance.

But a highly visible mistake Google’s bot, named “Bard,” made in a Monday blog post the company put out announcing the product, and the fact that Google hasn’t let outsiders test its tool yet, potentially contributed to concerns from the company’s investors, who sold off shares, leading to an 8% drop in the company’s value Wednesday morning.

The dueling announcements, hyperfocus on whether Google is dipping behind Microsoft and twitchy Wall Street reaction show how the battle over AI has become the central obsession of the tech industry.

“The AI arms race between Microsoft and Google (and the rest of Big Tech) has begun,” said Dan Ives, an analyst with Wedbush Securities, in a Wednesday note.

The chatbot search tools announced by the companies this week are different from regular search engines in that they create longer, contextualized answers in response to queries, cutting out the need for people to click through to a publisher’s website. On Tuesday, Microsoft held a major event at its Redmond, Washington, headquarters, where Chief Executive Officer Satya Nadella touted its chatbot search feature, based on tech developed by smaller AI company OpenAI. Google, meanwhile, appeared to attempt to front-run the announcement with the unveiling of its own chatbot on Monday – something it then demoed at an event in France on Wednesday.

The competition between the two tech giants reflects the excitement and hype around technology called generative AI, which uses massive computer programs trained on reams of text and images to build bots that conjure content of their own based on relatively complex questions. Google first unveiled its chatbot LaMDA in 2021, but didn’t make it available to the public. Last year, smaller AI company OpenAI made its chatbot ChatGPT and image generator DALL-E available to the public, spurring a burst of interest in the technology, which in turn pushed Microsoft and Google to rush out their products.

Now, the tech giants are trying to bring that same type of experience to search, which has slowly transformed over the past decade to offer more and more sophisticated answers to users. But experts caution that widespread public availability of these types of AI – which harbor biases formed from the information they’re trained on and have been shown to consistently make factual errors and invent information – opens a potential Pandora’s box.

Microsoft is already rolling out its tool – available only on its Edge browser – to select users. The tech is based on smaller AI company OpenAI’s ChatGPT, which Microsoft signed a multibillion-dollar deal with recently. Google said its Bard search tool powered by LaMDA would be available in the “coming weeks” and that it is still testing it with paid contractors.

Microsoft, which has a tiny sliver of the search market with its engine Bing and would require significant gains to catch up, still seemed to have impressed investors, leading to Google’s stock drop.

Google’s announcement did not include a hands-on demonstration like Microsoft’s did. Its chatbot also made a mistake in the example the company showed off in its initial blog post – incorrectly saying that the James Webb telescope was the first to take a picture of an exoplanet, when really it was a different telescope.

Chatbots routinely make factual mistakes or mix incorrect information into their answers, a problem skeptics of the technology say suggests it is clearly not ready to be incorporated into search engines.

At the Wednesday event, Prabhakar Raghavan, who is one of Google’s most powerful executives as the head of search and ads, said the bot would be used for questions where there is no single, direct answer. He showed a video of the bot answering questions about what type of car someone should buy, which stops to visit along the way during a road trip from San Francisco to Santa Cruz and which constellations to look up at in the night sky.

But each of those queries, typed into Google currently, provide links to numerous blog posts written by real-life content creators. The demo video showed Google’s AI-generated answers pushing down the links to other websites as it generated more answers on its own.

Google and Microsoft have both used AI in their search engines for years to help parse peoples’ queries, decide which content is best for which questions and offer other services like translation. But the chatbots are the first case of the companies using generative AI.

Both search engines moved away from the “ten blue links” model of simply providing links to other websites years ago, and now often provide direct answers for questions about the weather, sports scores and the ages of famous people. But some AI entrepreneurs believe generative AI will create a world where linking back to original source material becomes obsolete, with chabots like ChatGPT or Bard simply answering people’s questions directly based on the knowledge they’ve gained from hoovering up the internet’s collective knowledge.

That is prompting concerns from internet publishers that the new systems will simply steal their work and present it as their own, without sending any traffic back to the original content creators. Google’s example did not show it citing sources, while Microsoft’s did.

But the tech is still very early, and it’s likely Google will cite sources and link back to original authors when its bot does officially debut, said Ross Hudgens, a search engine optimization expert and chief executive of Siege Media, a content marketing company. The company’s business model of getting users to click on links to ads is too important to jeopardize, he said.

“Google needs to maintain the experience of sending people to outside websites,” he said.