AI

Google admits its AI Overviews need work, but we’re all helping it beta test

Comment

Sundar Pichai onstage at Google IO
Image Credits: Google

Google is embarrassed about its AI Overviews, too. After a deluge of dunks and memes over the past week, which cracked on the poor quality and outright misinformation that arose from the tech giant’s underbaked new AI-powered search feature, the company on Thursday issued a mea culpa of sorts. Google — a company whose name is synonymous with searching the web — whose brand focuses on “organizing the world’s information” and putting it at user’s fingertips — actually wrote in a blog post that “some odd, inaccurate or unhelpful AI Overviews certainly did show up.”

That’s putting it mildly.

The admission of failure, penned by Google VP and Head of Search Liz Reid, seems a testimony as to how the drive to mash AI technology into everything has now somehow made Google Search worse.

In the post titled “About last week,” (this got past PR?), Reid spells out the many ways its AI Overviews make mistakes. While they don’t “hallucinate” or make things up the way that other large language models (LLMs) may, she says, they can get things wrong for “other reasons,” like “misinterpreting queries, misinterpreting a nuance of language on the web, or not having a lot of great information available.”

Reid also noted that some of the screenshots shared on social media over the past week were faked, while others were for nonsensical queries, like “How many rocks should I eat?” — something no one ever really searched for before. Since there’s little factual information on this topic, Google’s AI guided a user to satirical content. (In the case of the rocks, the satirical content had been published on a geological software provider’s website.)

It’s worth pointing out that if you had Googled “How many rocks should I eat?” and were presented with a set of unhelpful links, or even a jokey article, you wouldn’t be surprised. What people are reacting to is the confidence with which the AI spouted back that “geologists recommend eating at least one small rock per day” as if it’s a factual answer. It may not be a “hallucination,” in technical terms, but the end user doesn’t care. It’s insane.

What’s unsettling, too, is that Reid claims Google “tested the feature extensively before launch,” including with “robust red-teaming efforts.”

Does no one at Google have a sense of humor then? No one thought of prompts that would generate poor results?

In addition, Google downplayed the AI feature’s reliance on Reddit user data as a source of knowledge and truth. Although people have regularly appended “Reddit” to their searches for so long that Google finally made it a built-in search filter, Reddit is not a body of factual knowledge. And yet the AI would point to Reddit forum posts to answer questions, without an understanding of when first-hand Reddit knowledge is helpful and when it is not — or worse, when it is a troll.

Reddit today is making bank by offering its data to companies like Google, OpenAI and others to train their models, but that doesn’t mean users want Google’s AI deciding when to search Reddit for an answer, or suggesting that someone’s opinion is a fact. There’s nuance to learning when to search Reddit and Google’s AI doesn’t understand that yet.

As Reid admits, “forums are often a great source of authentic, first-hand information, but in some cases can lead to less-than-helpful advice, like using glue to get cheese to stick to pizza,” she said, referencing one of the AI feature’s more spectacular failures over the past week.

Google AI overview suggests adding glue to get cheese to stick to pizza, and it turns out the source is an 11 year old Reddit comment from user F*cksmith 😂 pic.twitter.com/uDPAbsAKeO

— Peter Yang (@petergyang) May 23, 2024

If last week was a disaster, though, at least Google is iterating quickly as a result — or so it says.

The company says it’s looked at examples from AI Overviews and identified patterns where it could do better, including building better detection mechanisms for nonsensical queries, limiting the user of user-generated content for responses that could offer misleading advice, adding triggering restrictions for queries where AI Overviews were not helpful, not showing AI Overviews for hard news topics, “where freshness and factuality are important,” and adding additional triggering refinements to its protections for health searches.

With AI companies building ever-improving chatbots every day, the question is not on whether they will ever outperform Google Search for helping us understand the world’s information, but whether Google Search will ever be able to get up to speed on AI to challenge them in return.

As ridiculous as Google’s mistakes may be, it’s too soon to count it out of the race yet — especially given the massive scale of Google’s beta-testing crew, which is essentially anybody who uses search.

“There’s nothing quite like having millions of people using the feature with many novel searches,” says Reid.

More TechCrunch

Tags

If you spend time on X or Threads, where snarky memes rise and fall, you’ve probably seen posts referencing “founder mode” over the last few days, like this: https://www.threads.net/@carnage4life/post/C_eaQAxyIcV Or…

Those ‘Founder mode’ memes keep coming

Andreessen Horowitz (a16z) has closed a satellite office in Miami Beach just two years into a five-year lease it signed for an 8,300-square-foot space. The reason? Disuse, reports Bloomberg. Miami…

Andreessen Horowitz shutters its Miami office after two years

These final maneuvers will bring to a close a troubled first crewed mission for the Boeing-made Starliner.

Boeing and NASA prepare to bring Starliner home without its crew on Friday

As Meta tries to rekindle the flame between Facebook and socially anxious youths, the company released a blog post Wednesday titled, “Navigating your 20s with Facebook.”

Facebook says, ‘How do you do, fellow kids?’

Cowboy has closed funding of around $5.5 million. With this recent funding round, Cowboy is now valued at €40 million on a pre-money basis.

E-bike maker Cowboy raises a small funding round as it targets profitability next year

HR and payroll software company Paylocity has agreed to acquire corporate spend startup Airbase for $325 million, the companies announced Wednesday. The deal is subject to regulatory approval and is…

Paylocity is acquiring corporate spend startup Airbase for $325M

A long-running lawsuit over the Internet Archive’s “emergency” e-book lending practices during the COVID-19 pandemic has ended in a loss for the website and a victory for publishers. The lawsuit…

Publishers prevail in lawsuit over Internet Archive’s ’emergency’ e-book lending

Ryan Breslow’s plan to get himself reinstalled as CEO of fintech company Bolt — and push through a $450 million fundraising deal that would value the startup at a staggering $14 billion…

Ryan Breslow’s $450M Bolt deal said to involve a restraining order now

Maybe a lack of AI characters is what Quibi got wrong. At least, that’s what one startup appears to believe.  My Drama is a new short series app with more…

Short series app My Drama takes on Character.AI with its new AI companions

A 23-year-old woman who allegedly killed two men in March while using Ford’s hands-free system, BlueCruise, has been charged with DUI homicide by Pennsylvania State Police. The woman, Dimple Patel,…

Woman who allegedly killed two people using Ford BlueCruise charged with DUI homicide

The hiring effort comes after X, formerly known as Twitter, laid off 80% of its trust and safety staff since Musk’s takeover.

X is hiring staff for security and safety after two years of layoffs

Hiya, folks, welcome to TechCrunch’s regular AI newsletter. If you want this in your inbox every Wednesday, sign up here. This week in AI, two startups developing tools to generate and…

This Week in AI: VCs (and devs) are enthusiastic about AI coding tools

The Cosmos Institute, a nonprofit whose founding fellows include Anthropic co-founder Jack Clark and former Defense Department technologist Brendan McCord, has announced a venture program and research initiatives to —…

The Cosmos Institute, whose founding fellows include Anthropic co-founder Jack Clark, launches grant programs and an AI lab

Once linked, parents will be alerted to their teen’s channel activity, including the number of uploads, subscriptions and comments.

YouTube debuts new parental controls aimed at teens

No one is putting the remote working genie back in the bottle. Which is good news for Oyster, a payroll and HR platform that specializes in distributed workforces — or…

As remote working keeps rolling, Oyster raises $59M Series D at $1.2B valuation

For the college students who are satisfied with dating apps, which may not be many, Tinder announced Wednesday a series of updates to Tinder U, its in-app feature that caters…

Tinder update targets college students as dating apps struggle

The exact contents of X’s (now permanent) undertaking with the DPC have not been made public, but it’s assumed the agreement limits how it can use people’s data.

Ireland’s privacy watchdog ends legal fight with X over data use for AI after it agrees to permanent limits

Years ago, Twitter tried but eventually walked away from building TV apps after getting a lukewarm reception. Now, as it looks to revive its advertising business, its new incarnation X…

X doubles down on video with a new TV app

Apple is likely to unveil its iPhone 16 series of phones and maybe even some Apple Watches at its Glowtime event on September 9.

Apple event 2024: How to watch the iPhone 16 launch

Korea’s Institute of Machinery and Materials this week showcased a robotic wheelchair with large, deformable wheels that can manage rocks, stairs and other obstacles. During normal operation, the wheel maintains…

Watch this robotic wheelchair’s compliant wheels take on bumps, rocks and stairs

Mayfield is launching AI Garage, a $100 million initiative for ideation-stage founders interested in building “AI teammate” companies.

Mayfield allocates $100M to AI incubator modeled after its entrepreneur-in-residence program

Anthropic is launching a new subscription plan for its AI chatbot, Claude, catered toward enterprise customers that want more administrative controls and increased security. Claude Enterprise will compete with OpenAI’s…

Anthropic launches Claude Enterprise plan to compete with OpenAI

Time is running out to take advantage of our Student Pass discount for TechCrunch Disrupt 2024. Students and recent graduates can still save up to $200 until September 6 at…

Students and recent grads: Only 3 days left to save on TechCrunch Disrupt 2024 Student Passes

Fast-forward to today, Slauson & Co. remains even more committed to the mission of inclusivity in its funding, and it seems limited partners have its back. 

Slauson & Co. raises $100M Fund II proving appetite for inclusion persists

Safe Superintelligence (SSI), the AI startup co-founded by former OpenAI chief scientist Ilya Sutskever, has raised over $1 billion in capital from investors including NFDG (an investment partnership run by…

Ilya Sutskever’s startup, Safe Superintelligence, raises $1B

The American sports betting market produced $10.9 billion in revenue in 2023 for casinos, sportsbooks and iGaming, according to the American Gambling Association. One of the reasons this industry is…

DubClub wants amateur sports bettors to win more

New climate tech VC firms have emerged in recent years, but existing ones are also raising larger funds. Founded in 2007, Dutch firm SET Ventures is one of the latter.…

Dutch clean energy investor SET Ventures lands new €200 million fund, which will go toward digital tech

Revefi connects to a company’s data stores and databases (e.g. Snowflake, Databricks and so on) and attempts to automatically detect and troubleshoot data-related issues.

Revefi seeks to automate companies’ data operations

If you build an AI search product, you compete with Google. But Google has a lot easier time answering queries with a single, simple answer, such as “how many is…

With $50M in new funding, You.com thinks its AI can beat Google on hard questions

Featured Article

reMarkable’s Paper Pro adds color, light and more but keeps the focus on ‘focus’

The $499 Paper Pro — a new naming convention to indicate it is a higher-end alternative to the now-$379 reMarkable 2, not a direct successor — is momentous for its addition of both color and a “frontlight,” though both features are what you might call muted.

reMarkable’s Paper Pro adds color, light and more but keeps the focus on ‘focus’