Google uses AI to answer some search queries

• Google is testing AI-generated answers in regular search results.
• The experiment, dubbed AI Overview, began about a month ago in the US and UK.
• Remember when Google’s purpose was to create a list of sites that provided the information you needed? That was a long time ago.

Sure, you’ve heard about AI and technologies like ChatGPT. More and more people are trying to use technology in their daily lives, whether it’s for entertainment, school, or work.

However, if you want to use AI in your daily life, you will most likely have to look for it. If you want AI-generated results, you’ll need to use the ChatGPT website or app or click on a specific tab, such as Microsoft’s Bing.

Now, that’s changing, at least for some Google search users like me. On Monday morning, using regular old Google, I typed in “Truman Doctrine Vietnam War” and it returned “AI Overview.” It wasn’t something Google found on another site, but something it wrote itself.

I haven’t spent much time reading about post-World War II US foreign policy lately. That’s why we asked Google. But actually, I think that was a pretty good answer.

But far more interesting than the answer was how Google came up with it.

We knew Google was scrambling to catch up in the AI ​​product wars, but in a world where there is no longer a need to generate links to websites containing information, the AI ​​engine just needs to generate the information. was worried about the fate of its core search product. Get the answers you need. And we also knew that Google was developing an AI-powered version of search. You can try this yourself.

But I didn’t know that Google started making this information available to the public, mixed in with all other search results.

After all, they lasted about a month. And the fact that Google hasn’t made any big announcements about this may or may not be telling. For now, the only place you can find out about this on the web is this update from trade magazine Search Engine Land.

Results: Google is currently testing self-generated “AI summaries” for some common search queries that are considered complex (but answerable). A Google representative said AI answers are being deployed to a “very limited percentage of search traffic” in the US and UK.

And I shouldn’t be that surprised, actually: ever since OpenAI started wowing people with ChatGPT in late fall 2022, it was clear that this technology was coming to search. That’s the whole point of Microsoft’s big partnership with OpenAI, bringing technology to Bing.

But knowing it is one thing and actually starting to see it is another. And you begin to see it as a normal outcome rather than something special.

Google seems to be doing a good job of addressing this issue, addressing many of the obvious problems that Google-created answers would cause in search.

For example, clearly label results as experimental and generated by AI. And if you click on the “Learn more” link, you’ll see a well-written explanation like “Generative AI is not human. It can’t think for itself or feel emotions. It’s just good at finding patterns.” The sentence will be displayed. (And below that it says, “Generative AI is experimental and a work in progress, so it can and will make mistakes.”)

And Google is showing results. Clicking on the buttons within the results will provide links to helpful related sites such as the National Archives.

Since I’m a responsible journalist, I also wonder how Google Magi, which drives these results, will work with, or won’t work with, the technology that powers Google Gemini, the much-maligned “woke AI” chatbot. asked. I didn’t get a very satisfying answer other than that Google wants you to consider these as separate products.

However, it’s nice to have other people worried about Google’s wakefulness (we hear a lot less shouting about this now, right?). What I’m interested in is how these kinds of answers accelerate existing ways of using Google. This means using Google as a one-stop answer shop instead of a place that helps you find other places to find answers.

This trend has been well-documented over the years, and many people prefer Google to hang out with them through “knowledge panels” or something similar, rather than go elsewhere to get answers. It revolves around what Google has determined. Even though Google’s business model revolves around selling links to other places.

Now Google is trying to keep you at Google By displaying text from a site that purports to answer your question. In theory, if you want to know more, you can click through. But Google snippets often don’t motivate people to click through. We have everything you need. That’s it!

(On the other hand, the information that Google cites in text snippets mistaken — at least in part because Google relies on highly ranked web pages, whose primary purpose is to be ranked highly by Google, not to be accurate. For example, if you ask Google “What is Jason Kelsey’s net worth,” it will highlight a (not very helpful) answer from Taylor’s Debit His Cards site, which also provides search fodder for his Swift cat. )

And things get really interesting when you start imagining Google constantly providing completely AI-generated answers like this to all kinds of queries.

On the other hand, Google may become even more valuable now that it no longer even pretends to be a “search engine.” Google is just an answering machine. And I go there because I’m used to it.

On the other hand: In that scenario, Google certainly isn’t. only Answer Machine. This means that the entire Empire and the many companies that depend on it (such as digital publishers?) are up for grabs.

You can see why Google is testing things like this carefully and quietly, and why they need to figure out answers quickly.