The Death of Search – The Atlantic

For nearly two years, the world's largest technology companies have been saying that AI will change the internet, your lives and the world. But first, they're redesigning the humble search engine.

In theory, chatbots and search go together perfectly. A standard Google search interprets a search query and returns relevant results; Technology companies have invested tens or hundreds of millions of dollars into developing chatbots that interpret human input, synthesize information and provide fluid, useful responses. No more refining keywords or searching Wikipedia – ChatGPT does it all. Search is also an attractive target: shaping the way people navigate the Internet is tantamount to designing the Internet itself.

Months of prophecies about generative AI have almost suddenly led to perhaps the clearest view of the future of the Internet. After a series of limited releases and product demos that were riddled with various setbacks and embarrassing mistakes, tech companies are introducing AI-powered search engines as fully realized, all-inclusive products. Last Monday, Google announced that it would be rolling out its AI Digests in more than 100 new countries; This feature will now reach more than 1 billion users monthly. Days later, OpenAI announced a new search feature in ChatGPT, initially available to paying users and soon rolling out to the public. That same afternoon, AI search startup Perplexity released instructions on how to make its “answer engine” the default search tool in your web browser.

Read: The AI ​​search war has begun

Over the past week, I've been using these products in a variety of ways: researching articles, tracking the election, and conducting everyday searches. In turn, I looked, as best I could, into the future of how billions of people will access, interact with, and synthesize information. I've learned that these products are unexpectedly convenient, frustrating, and strange all at the same time. The current versions of these tools have surprised and sometimes impressed me, but even if they work perfectly, I'm not convinced that AI search is a wise endeavor.

The search bar has been a familiar entity for decades. People all over the world are used to it; Multiple generations implicitly view Google as the first and best way to learn about virtually anything. Enter a request, browse a list of links, enter a follow-up request, receive more links, etc. until your question is answered or your request is fulfilled. This indirectness and wide reach – all that clicking and scrolling – are in some ways the defining characteristics of a traditional Google search, allowing (or even forcing) you to traverse the depth and breadth of connections that justify the term worldwide Web. In this sense, the hyperlink is the building block of the modern Internet.

This vastness is wonderful when delving into Lucrezia de Medici, as I did on a trip to Florence last year, or when diving deep into a scholarly dilemma. It's perfect for stumbling across adorable video clips, magazine posts, and social media posts. And it's annoying when all you need is a simple biographical answer, or a brunch recommendation without the backstory of three different chefs, or a brief overview of a complex field of research without having to wade through obscure documents.

In recent years, more and more Google Search users have found that the frustration outweighs the joy – describing a growing number of paid ads, seemingly relevant links designed to outperform the search algorithm, and erroneous results. Generative AI promises to address these moments of frustration by providing a completely different experience. When I asked ChatGPT to search the Internet for reasons why Kamala Harris lost the presidential election, I received a short list of four factors: “economic concerns,” “demographic changes,” “swing state dynamics,” and “campaign strategies.” “. It was an easy-to-understand answer, but not a particularly insightful one; In response to a follow-up question about voter demographics, ChatGPT provided a stream of statistics without context or analysis. A similar Google search now brings up a variety of news analyzes that you have to read through. If you follow Google's links, you will develop a much deeper understanding of American economics and politics.

Another example: I recently read about a controversial proposed infrastructure project in Maryland. Google searches led me through a maze of public documents, corporate presentations, and hours of City Council meeting recordings that took ages to review but piqued my curiosity and kept me well informed. When asked, ChatGPT provided an accurate summary and timeline of events and cited the sources – which was an extremely useful way of organizing the reading I had already done, but in itself could have spelled the end of my explorations.

I have long been a critic of AI-powered search. Technology has repeatedly fabricated information and struggled to accurately attribute its sources. The authors are accused of plagiarism and violating the intellectual property rights of major news organizations. None of these concerns have been fully addressed: the new ChatGPT search function has made some mistakes in my own use and in other reports, mixing up data, misreporting sports scores, and telling me that Brooklyn's Prospect Park is bigger than Manhattan's (much bigger). ) Central Park. The links offered by traditional search engines are also full of errors – but search bots implicitly ask for your trust without verification. The quotes are not particularly inviting to click on. And while OpenAI and Perplexity have partnered with a number of media organizations including The Atlantic— perhaps competing for the high-quality, human-generated content their search bots rely on — it's unclear to me exactly how sites that once relied on advertising revenue and subscriptions will fare in an AI-filtered web. (The editorial team of The Atlantic operates independently of the division, which announced its corporate partnership with OpenAI in May.)

Read: AI search is becoming the problem everyone is worried about

Although ChatGPT and Perplexity as well as Google AI Overviews cite their sources with (small) footnotes or clickable bars, not Clicking on these links is the whole point. OpenAI wrote in the announcement of its new search feature that “getting useful answers on the web can be a lot of work.” Finding quality sources and the right information for you often requires multiple searches and browsing through links. Now chat can help you get a better answer.” Google claims its AI “does the Googling for you.” Perplexity's chief business officer told me this summer that “people don't come to Perplexity to consume journalism” and that the AI ​​tool will generate less traffic than traditional search. For curious users, Perplexity suggests follow-up questions so that instead of opening a footnote, you can continue reading in Perplexity.

The change represents the transition from navigating a library using the Dewey Decimal System and finding related books on adjacent shelves to requesting books for pickup through a digital catalog. It could completely reorient our relationship to knowledge, prioritizing quick, detailed and abridged answers over deep understanding and consideration of different sources and viewpoints. The great thing about searching the internet is that you get caught up in ridiculous Reddit debates and develop unforeseen obsessions over a topic you first heard about six hours ago through another search; constantly falling into disarray and treasure without ever meaning to. AI search could block these pathways not only to discovery but also to its drive, curiosity.

The factuality and attribution issues may well be resolved – but even if they aren't, tech companies show no signs of backing down. Controlling search means controlling how most people access all other digital things – it's an incredible platform for gaining trust and visibility, advertising or influencing public opinion.

The internet is changing and no one outside of these companies has any say in it. And the biggest, most useful, and scariest change could come from making AI search engines work properly. With AI, the goal is to keep you in a tech company's ecosystem – so that you continue to use the AI ​​interface and receive the information that the AI ​​finds relevant and necessary. The best searches are goal-oriented; The best answers are short. Which perhaps shouldn't be surprising given that Silicon Valley giants are all about optimizing their companies, products and user lives.

A little or even a lot of inefficiency in search has long been the norm; The AI ​​will wipe it out. Our lives will be more comfortable and simpler, but perhaps a little less wonderful and full of wonder, a little less enlightened. A process that was once focused on exploration will shift to extraction. Meander less, chase more. No more unknown unknowns. If these companies have their way, there will be no more hyperlinks – and therefore no real web.