Why is Google’s AI-powered Search Generative Experience (SGE) still a work in progress?

Google’s experimental Search Generative Experience in Search Labs brings AI-powered summaries and answers drawn from a wide range of dubious sources, but is a definite timesaver

October 11, 2023 03:46 pm | Updated 03:49 pm IST

Much like Bard, Google’s Search Generative Experience in Search Labs is still highly experimental and prone to both errors and illogical results [File]

Much like Bard, Google’s Search Generative Experience in Search Labs is still highly experimental and prone to both errors and illogical results [File] | Photo Credit: REUTERS

Google is feverishly injecting generative AI into the very heart of its business: internet search.

Users got their first taste of this with the company’s AI-powered chatbot Bard, which could collate information in seconds to answer queries in natural human speech.

The search giant also launched its experimental Search Generative Experience (SGE) that provides an enhanced AI search experience to those who opt in through Google Chrome.

Users can try out Google SGE by visiting Search Labs and signing up via their own Google account. After that, continue making Google searches as you normally would. Keep in mind not to input sensitive data or financial information.

(For top technology news of the day, subscribe to our tech newsletter Today’s Cache)

The feature can be tried on the Google app on Android and iOS, as well as the desktop version of Chrome.

The idea behind Google SGE

At the most basic level, internet users who have opted for Google SGE will get AI-powered responses when searching up subjects or questions across a variety of fields such as history, current events, household management, business, movie and celebrity gossip, mental health, mathematics, and more. Users can ask follow-up questions, click on related questions that will generate new answers, and upvote/downvote the responses to share their feedback. They can also try out the search generative experience with voice typing, or listen to the response being read aloud by the system.

Google’s generative search experience helps users quickly make sense of data and dense text content but can cite strange sources [File]

Google’s generative search experience helps users quickly make sense of data and dense text content but can cite strange sources [File] | Photo Credit: Sahana Venugopal

Not all queries result in SGE-powered answers, and users looking up sensitive information such as health queries or ongoing conflicts may have to rely on the top results, the news tab, Bard, or their own research.

““SGE while browsing” was specifically designed to help people more deeply engage with long-form content from publishers and creators, and make it easier to find what you’re looking for while browsing the web,” said Rany Ng, VP of Product Management at Google Search, in a blog post about the feature on August 15.

User experience

Google SGE’s AI-powered response takes several seconds to be generated and is the top search result, encased in a colourful box. Users get a quick answer to their query and additional context, as well as any necessary disclaimers.

Instead of clicking on web pages to find the needed information, questions are directly answered with Google SGE [File]

Instead of clicking on web pages to find the needed information, questions are directly answered with Google SGE [File] | Photo Credit: Sahana Venugopal

The responses come with thumbnail links at the end, pointing to their original sources. Users can also opt for section-by-section citations for more precision.

However, one major problem became immediately visible. Google SGE tries to provide localised sources in response to search queries. While a good idea in theory, this did not pan out well in reality. For example, an open-ended question about Mughal emperors generated a response which cited a popular far-right Indian news website—with an obvious anti-Mughal bias—as one of its sources.

Coming to other generalised fact-based questions, Google SGE responses cited dubious sources such as advertising agencies, ed-tech blog pages, Wikipedia profiles, local news websites from other countries, and even question-and-answer websites such as Quora.

Out of all these sources, Quora proved to be especially problematic as this website often directed readers to yet another AI-generated answer produced by its own ChatGPT bot. Instead of helping users engage critically with long-form content from trusted (and human) publishers/creators as Google had intended, we repeatedly found ourselves directed to other machine-generated answers, making it more likely that we would encounter additional errors and even hallucination. This recycling of AI-generated answers greatly degraded our user experience.

Google SGE often led us back to other AI-generated responses, defeating its purpose [File]

Google SGE often led us back to other AI-generated responses, defeating its purpose [File] | Photo Credit: Sahana Venugopal

On other occasions, the SGE-enabled answer was just a simple text-based list with little to no context and came right on top of the more engaging and informative links provided as part of Google’s default search results. Google SGE also once answered an English-language query in Hindi without being prompted to do so - another example of its localisation attempt gone awry.

Google SGE responses sometimes abruptly switched languages [File]

Google SGE responses sometimes abruptly switched languages [File] | Photo Credit: Sahana Venugopal

However, when the AI-powered responses were drawn from responsible sources such as well-edited news platforms or official websites, the Google SGE function helped us make sense of complex issues in simple language. Collecting information from several top websites, it summarised and presented data in a far more digestible format that saved us the time we would have otherwise spent clicking on links and fighting through a barrage of advertisements and pop-ups and sidebars on each new web page we opened. While there were occasional errors in the AI-generated responses, these were not glaring or overwhelming.

In conclusion

Much like Bard, Google’s Search Generative Experience in Search Labs is still highly experimental and prone to both errors and illogical results. While Google SGE’s ultimate aim is to provide internet users with faster and more concise responses based on multiple trusted sources, users will encounter factually incorrect answers pulled from unreliable websites, or even other machine-generated answers. However, Google SGE also saves time and effort by neatly compiling many of the answers users needed, so that they are not forced to trawl through websites filled with intrusive elements.

While the generative AI search experience is sure to be upgraded and improved - as was the case for Bard - Google’s manual search experience is still the gold standard for internet users.

Top News Today

Sign in to unlock member-only benefits!
  • Access 10 free stories every month
  • Save stories to read later
  • Access to comment on every story
  • Sign-up/manage your newsletter subscriptions with a single click
  • Get notified by email for early access to discounts & offers on our products
Sign in

Comments

Comments have to be in English, and in full sentences. They cannot be abusive or personal. Please abide by our community guidelines for posting your comments.

We have migrated to a new commenting platform. If you are already a registered user of The Hindu and logged in, you may continue to engage with our articles. If you do not have an account please register and login to post comments. Users can access their older comments by logging into their accounts on Vuukle.