Google plans to begin including AI-generated text answers to search queries in the coming months. The announcement came at the company’s (kind of slapdash) live event in Paris on Thursday morning, along with numerous other updates to Search, Maps, Translate and more.
Earlier this week, Google introduced Bard, its AI chatbot response to OpenAI’s wildly popular ChatGPT. But unlike Bard, these generative AI responses won’t be contained in a separate widget, they’ll become a part of standard searches.
What will generative AI in Google search look like?
For questions with no single correct answer (which Google has termed “No One Right Answer,” or NORA, queries), the company’s search engine will soon provide multi-sentence responses, exploring and explaining some of the many possibilities. These longform AI blurbs will appear in a box at the top of search, similar to where definitions for words or maths solutions might appear now.
Using a pre-recorded demo, Google exec Prabhakar Raghavan shared an example about stargazing. Offered the query, “What are the best constellations to look for, for stargazing?” Google’s generative AI spit out a bullet-pointed list of nuanced responses. It mentioned Orion, Cassiopeia, and other star formations notable for different reasons — their brightness, their shape, and so on.
On top of providing the initial text response, the search results for the stargazing query included a few suggested follow-up questions; also presumably AI-generated. These included, “What is the best time of year to see these constellations.” And, in response to that follow-up, the search engine offered a whole new list of bullet-pointed information which appeared to draw on knowledge of the previous answer. To the second query, Google didn’t just give the best time of year to see any old constellation — it offered descriptions specific to Orion, Cassiopeia, and the others it had first listed.
What’s happening with the Bard chatbot though?
Google didn’t provide much of an update on its AI chatbot. We already knew it had entered testing with a small group of “trusted” demo users, using a light version of the company’s LaMDA language model. And for now, that’s still the case.
Next month, Google said it will begin onboarding developers to interact with its generative language API to build their own tools and features.
Google’s recent rush to forefront AI is an obvious response to the rise of ChatGPT, and Microsoft’s swift investment in a multi-billion dollar partnership with OpenAI. On Tuesday, Microsoft’s ChatGPT-powered Bing search engine went live for a preview. Similar to Google, Microsoft also seems to be both offering an AI-chatbot option, separate from standard search, as well as AI-responses integrated into search.
At the moment, Microsoft appears to be taking a more aggressive approach to its AI-search responses. In an example on the company’s blog — a search for fitness advice yields a step-by-step workout plan that’s AI-generated, alongside standard video and link results. Whereas Google seems to be taking a more cautious approach — not using generative-AI to present definitive answers or guidance.
“We were one of the first companies to embrace AI principles,” Google’s Raghavan said. But he emphasised that the Google’s philosophy to developing generative-AI tools for the public is to be “bold and responsible.” As the legacy search engine dominator, Google has a reputation to uphold. Whereas Bing can likely be a little more free-wheeling in Microsoft’s bid for a real stake in the long-monopolized search space.
What else did Google announce?
In addition to the updates on AI-powered text search, Google introduced a whole suite of other improvements and expansions to existing features in Search, Maps and Translation.
While many of these aren’t new, they’re finally making their way to Australia.
Search:
-
Multisearch: Google last year launched multisearch, a way to search using text and images at the same time. In the Google app, you can use Lens to take a picture or use a screenshot and add text to it (the idea is that it’ll be just like when you put at something to ask a question about it). As of today, multisearch is live globally on mobile, in all languages and countries where Lens is available.
-
Multisearch near me: With multisearch near me you can snap a picture or take a screenshot of a dish or item then find it nearby instantly. Over the next few months, this feature will be coming to all languages and countries where Lens is available.
-
Multisearch on the web: Multisearch will also be expanding to images on the web globally on mobile.
-
Search your screen on Android: The company has upgraded its Lens and visual search tools to allow users to one-click search items from images and videos on their phone screens. Basically, if an Android phone user pulls up a picture of a historical building, they can tap it to find out what that landmark is called. This feature will be available in all languages where Lens is available.
Maps:
-
Immersive view: Google introduced a new way to interact with Maps, called “immersive view.” The mode, enabled by AI-fusing of 2-D images into a 3-D rendering, can take you from a birds’ eye perspective to the inside of a business in just a few clicks. But for now, the feature is limited to just a handful of cities: London, Los Angeles, New York, San Francisco, and Tokyo, coming soon to Florence, Venice, Amsterdam, and Dublin.
-
Live View: For those who often find themselves disoriented navigating large indoor spaces like airports, Google has also expanded its “Live View,” directions. The tool offers real-time directional arrows overlaid on your phone’s camera view of the world. Instead of looking at a 2-D map to orient themselves, Google users in a select number of locations like certain transit hubs and shopping centres now have the option of holding up their phone camera for an augmented reality directional guide. Google is expanding Search with Live View to more places in Europe in the coming months, including Barcelona, Madrid and Dublin. Indoor Live view is expanding to over 1,000 new airports, train stations, and malls in cities including London, Paris, Berlin, Madrid, Barcelona, Prague, Frankfurt, Tokyo, Sydney, Melbourne, São Paulo, Singapore, and Taipei in the coming months.
-
Glanceable directions: Track your journey from your route overview or lock screen, with updated ETAs and where to make your next turn now available without unlocking your phone. This feature starts rolling out globally on Android and iOS in the coming months.
-
New Maps features for EVs with Google built-in: Now, cars with Android Automotive will receive info on very fast charging stations, will show charging stations in search results and will start adding charging stops to shorter trips.
Translation:
-
Contextual translation options: In a translation update, Google has added more context to responses, long missing from the company’s tools. Instead of just receiving a single option for the translation of a word like “novel,” Google will offer information on all the possible definitions (i.e. a book, new, original) and when you might use each translated phrase. For now, the context update is limited to German, French, English, Japanese, and Spanish.
-
New design and features: Google has also redesigned the Translate app on Android and soon, it’ll be bringing a new design to iOS users.
-
Translation features in Lens: Lens will now have Google Translate baked in.
This article has been updated since it was first published.