Step 1: Google rolls out new AI-powered products.
Step 2: Users quickly spot a flaw in the product and point it out in a social media post, which then makes the news.
Step 3: Google admits that its new AI-powered product is fundamentally flawed and stops developing it.
Now we're in the second phase of the same process: Google is broadly rolling out its AI Summary feature, which replaces regular answers to search queries (lists of links to sites where you might actually find the answer you're looking for) with AI-generated answers that try to summarize the content of those sites, and people are finding examples of Google generating incorrect, sometimes comically bad, answers.
That's why my coworker Katie Notopoulos made a pizza out of glue and ate it. (Congratulations Katie! That's a truly heroic act. Use your hazard pay wisely (we all get hazard pay, right?).)
So the $2 trillion question: Will Google have to back down on this one too?
Google says that's not the case, claiming that the stupid answers it generates are rare, and that most people don't know or care about a search answer that tells them how many rocks they should eat, or that they should stare at the sun for five to 15 minutes, although darker-skinned people can stare for twice that amount of time.
Google AI gave an interesting answer to a question about sun gazing. Overview of Google AI
Google also says it will be cleaning up bad answers as they appear, especially ones where someone smart enough to use a phone but stupid enough to follow the answer could hurt themselves.
Below is the official response from Google spokesperson Lara Levin:
“The majority of our AI summaries provide high-quality information with links to explore further on the web. Many of the examples we saw were unusual queries, and we also saw examples that were doctored or couldn't be reproduced. We conducted extensive testing before releasing this new experience, and as with other features we've released in Search, we appreciate your feedback. We respond promptly where appropriate, in accordance with our content policies, and we're using these examples to develop broader improvements to our system, some of which we've already started rolling out.”
got it.
But like I said, we've seen versions of this story before: What happens if people keep finding bad answers on Google and Google can't whack them off quickly enough? And, importantly, what happens if regular people who don't spend time reading or talking about tech news start hearing about Google's bad and potentially dangerous answers?
That's a really big problem. Google does a lot of different things, but the reason it's worth over $2 trillion remains its two core products: search, and the ads that are generated alongside those search results. And if the public loses trust in Google as a search/answer machine…
Well, that would really be a problem.
Privately, Googlers are increasingly convinced that these “bad answers” are actually a peripheral issue. And unlike the “woke Google” issue a few months back (where there was a real problem with the model Google was using to create its images), this isn't the case here. Google never gets things 100% right (they say even more quietly), because at the end of the day, it's only dependent on what people put out there on the internet. It's just that some people are now paying a lot more attention because there's something new to look at.
I believe the answer. I've had Google's AI answer appear in my search results for about a month now, and it's generally fine.
However, this isn't always the case.
and, very The difference between the old Google search results and the new is the responsibility and authority that Google has taken. Before, Google was saying that someone else could answer your question. Now, Google is answering your question.
That's the difference between me giving you a map and giving you directions that will drive your car off a cliff.
Like my 15-year-old son (we talk about this at home because we're weirdos), you could argue that Google shouldn't replace its old-fashioned perfect search results with AI-generated answers: If you wanted AI-generated answers, you'd go to ChatGPT, right?
But of course, what Google is worried about is people going to ChatGPT, which is why they're making this big pivot to disrupt themselves before ChatGPT or some other AI engine can disrupt them.
You can argue that they moved too fast, or were too sloppy, or whatever, but it's hard to imagine Google backing away from this now.
