experiments with google ai

AI + Writing

These experiments set out to explore whether machine learning could be used by writers to inspire, unblock and enrich their process.

Read the blog post about the experiment and how to build machine learning tools for writers. 

Experiments

Once upon a lifetime, between the lines.

Finished playing?

Instrument playground, games you might like, the descent of the serpent, geo artwork, puzzle party.

Google I/O 2024 Google’s biggest developer conference is happening Watch now

Making AI helpful for everyone

Supercharge your creativity and productivity with our most helpful AI assistant

A simpler, smarter way to search with generative AI

Conquer the blank page by turning a few keywords into an email

Reimagine your photos with Magic Editor in Google Photos

Create high-quality, photorealistic images with ImageFX

Automatic adjustments for improved quality with Video Boost on Pixel 8 Pro

Mix and generate custom sounds with MusicFX

Tackle the language barrier in videos with AI-powered dubbing

Responsible AI is woven into the fabric of our work

We approach AI boldly and responsibly, working together with experts, partners and other organizations so our models, products and platforms can be safer, more inclusive, and benefit society.

Building AI to be socially beneficial

Project Relate

Connectomics

AlphaFold is accelerating research in nearly every field of biology

By solving a decades-old scientific challenge, Google DeepMind’s AlphaFold gave millions of researchers a powerful new tool to help solve crucial problems like discovering new medicines or breaking down single-use plastics. One day, it might even help unlock the mysteries of how life itself works.

Project Relate: communication made easier

Around the world, it is estimated that 250,000,000 people have non-standard speech. Project Relate is a beta Android application that offers personalized speech recognition to empower people in their everyday lives.

Helping map the human brain using connectomics

Mapping the entire human brain could help us understand a lot about ourselves, from the causes of diseases to how we store memories. But mapping the brain with today’s technology would take billions of dollars and hundreds of years. Learn what Google Research is doing to make it easier for scientists to—someday—reach this goal.

Graphcast delivers 10-day weather predictions at unprecedented accuracy

Google DeepMind’s state-of-the-art AI model provides faster, more accurate medium-range weather forecasts than the current industry standard, including earlier warnings for extreme weather events like cyclones, floods, and temperature spikes.

What's new in Google AI

Generative AI in Search: Let Google do the searching for you

New generative models and tools, built with and for creators

Google I/O 2024: An I/O for a new generation

Gemini breaks new ground with a faster model, longer context, AI agents and more

Get more done with Gemini: Try 1.5 Pro and more intelligent features

Explore our other teams and product areas

Introducing Veo: Our most capable generative video model

Latest news.

Discover our latest AI breakthroughs and updates from the lab

View all posts

Responsibility & Safety

Looking ahead to the AI Seoul Summit

How summits in Seoul, France and beyond can galvanize international cooperation on frontier AI safety

Introducing the Frontier Safety Framework

Our approach to analyzing and mitigating future risks posed by advanced AI models

Gemini breaks new ground: a faster model, longer context and AI agents

We’re introducing a series of updates across the Gemini family of models, including the new 1.5 Flash, our lightweight model for speed and efficiency, and Project Astra, our vision for the future...

Project Astra

A universal AI agent that is helpful in everyday life

Technologies

Breakthrough research. Transformative products.

View all technologies

The most general and capable AI models we've ever built.

Our most capable generative video model.

Our highest quality text-to-image model.

Robust and scalable tool for watermarking and identifying AI-generated images.

Breakthrough AI system accurately predicts the 3D models of protein structures — and accelerates research in nearly every field of biology.

Responsibility

We want AI to benefit the world, so we must be thoughtful about how it’s built and used.

We help anticipate a broad spectrum of AI-related risks, explore ways of preventing them from happening, and find ways to address them if they do.

AI Principles

While we are optimistic about the potential of AI, we recognize that advanced technologies can raise important challenges that must be addressed clearly, thoughtfully, and affirmatively.

Explore our other teams and product areas

Experiments with Google

Since 2009, coders have created thousands of amazing experiments using Chrome, Android, AI, AR and more. We're showcasing projects here, along with helpful tools and resources, to inspire others to create new experiments. Here are collections of experiments to explore, with new ones added every week. Have fun.

Featured Collections

View all collections.

WebXR Experiments

AR and VR made for the web

AI Experiments

Celebrating Creativity and AI

Arts & Culture Experiments

See what happens at the crossroads of art and technology

Experiments for Learning

A collection of experiments that teachers, students, and families are using to learn from home.

Start With One

A collection of experiments that started by working with one person to make something impactful for them and their community

Chrome Experiments

Creative code for the web

Recent Experiments

View all experiments, passage of water, instrument playground, cultural icons, say what you see, don’t touch the art, what's happening.

• Español – América Latina
• Português – Brasil
• Documentation

Introduction to Vertex AI Experiments

Vertex AI Experiments is a tool that helps you track and analyze different model architectures, hyperparameters, and training environments, letting you track the steps, inputs, and outputs of an experiment run. Vertex AI Experiments can also evaluate how your model performed in aggregate, against test datasets, and during the training run. You can then use this information to select the best model for your particular use case.

Experiment runs don't incur additional charges. You're only charged for resources that you use during your experiment as described in Vertex AI pricing .

Track steps, inputs, and outputs

Vertex AI Experiments lets you track:

• steps of an experiment run , for example, preprocessing, training,
• inputs, for example, algorithm, parameters, datasets,
• outputs of those steps, for example, models, checkpoints, metrics.

You can then figure out what worked and what didn't, and identify further avenues for experimentation.

For user journey examples, check out:

• Model training
• Compare models

Analyze model performance

Vertex AI Experiments lets you track and evaluate how the model performed in aggregate, against test datasets, and during the training run. This ability helps to understand the performance characteristics of the models -- how well a particular model works overall, where it fails, and where the model excels.

• Compare pipeline runs

Compare model performance

Vertex AI Experiments lets you group and compare multiple models across experiment runs . Each model has its own specified parameters, modeling techniques, architectures, and input. This approach helps select the best model.

Search experiments

The Google Cloud console provides a centralized view of experiments, a cross-sectional view of the experiment runs, and the details for each run. The Vertex AI SDK for Python provides APIs to consume experiments, experiment runs, experiment run parameters, metrics, and artifacts.

Vertex AI Experiments, along with Vertex ML Metadata , provides a way to find the artifacts tracked in an experiment. This lets you quickly view the artifact's lineage and the artifacts consumed and produced by steps in a run.

Scope of support

Vertex AI Experiments supports development of models using Vertex AI custom training, Vertex AI Workbench notebooks, Notebooks, and all Python ML Frameworks across most ML Frameworks. For some ML frameworks, such as TensorFlow, Vertex AI Experiments provides deep integrations into the framework that makes the user experience automagical. For other ML frameworks, Vertex AI Experiments provides a framework neutral Vertex AI SDK for Python that you can use. (see: Prebuilt containers for TensorFlow, scikit-learn, PyTorch, XGBoost).

Data models and concepts

Vertex AI Experiments is a context in Vertex ML Metadata where an experiment can contain n experiment runs in addition to n pipeline runs. An experiment run consists of parameters, summary metrics, time series metrics, and PipelineJob , Artifact , and Execution Vertex AI resources. Vertex AI TensorBoard , a managed version of open source TensorBoard, is used for time-series metrics storage. Executions and artifacts of a pipeline run are viewable in the Google Cloud console .

Vertex AI Experiments terms

Experiment, experiment run, and pipeline run.

• An experiment is a context that can contain a set of n experiment runs in addition to pipeline runs where a user can investigate, as a group, different configurations such as input artifacts or hyperparameters.

experiment run

• An experiment run can contain user-defined metrics, parameters, executions, artifacts, and Vertex resources (for example, PipelineJob).

pipeline run

• One or more Vertex PipelineJobs can be associated with an experiment where each PipelineJob is represented as a single run. In this context, the parameters of the run are inferred by the parameters of the PipelineJob. The metrics are inferred from the system.Metric artifacts produced by that PipelineJob. The artifacts of the run are inferred from artifacts produced by that PipelineJob.

See Associate a pipeline with an experiment .

Parameters and metrics

• Parameters are keyed input values that configure a run, regulate the behavior of the run, and affect the results of the run. Examples include learning rate, dropout rate, and number of training steps.

See Log parameters .

summary metrics

• Summary metrics are a single value for each metric key in an experiment run. For example, the test accuracy of an experiment is the accuracy calculated against a test dataset at the end of training that can be captured as a single value summary metric.

See Log summary metrics .

time series metrics

• Time series metrics are longitudinal metric values where each value represents a step in the training routine portion of a run. Time series metrics are stored in Vertex AI TensorBoard. Vertex AI Experiments stores a reference to the Vertex TensorBoard resource.

See Log time series metrics .

Resource types

Pipeline job.

• A resource in the Vertex AI API corresponding to Vertex Pipeline Jobs. Users create a PipelineJob when they want to run an ML Pipeline on Vertex AI.
• An artifact is a discrete entity or piece of data produced and consumed by a machine learning workflow. Examples of artifacts include datasets, models, input files, and training logs.

Vertex AI Experiments lets you use a schema to define the type of artifact. For example, supported schema types include system.Dataset , system.Model , and system.Artifact . For more information, see System schemas .

Notebook tutorial

• Get started with Vertex AI Experiments

What's next

• Set up to get started with Vertex AI Experiments

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License , and code samples are licensed under the Apache 2.0 License . For details, see the Google Developers Site Policies . Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2024-06-07 UTC.

Android Experiments

Android + gaming, form 'n' fun, android + physical objects, ioio plotter, android + robotics, autonomous android vehicle, ioio camera robot project, all android experiments, neuralcandy, iss orbit tracker, animal detection device, home screen arcade.

• Do Not Sell My Personal Info

•  ⋅ 

New Google Labs Design Includes AI Tools For Search & Creativity

Discover the new Google Labs design with 12 AI experiments demonstrating the future of generative AI in search, productivity, and business.

• Google introduced a new design for Labs.Google, highlighting AI experiments for search, creativity, and productivity.
• Google SGE experiments still appear in Labs without the December 2023 expiration.
• NotebookLM and Bard Extensions utilize AI for research and creative processes with data from your Google account.

Google recently launched a new design for Labs with 12 artificial intelligence (AI) experiments, tools, and projects, showcasing a commitment to advancing AI technology.

This redesign highlights a dozen AI products that could transform how AI is in most workflows and organizations.

From search to productivity, here are the AI experiments you can try or sign up for on the newly designed Google Labs website.

12 Google AI Experiments

Here are the 12 AI experiments from Google that have been appearing in popular products, including Search, Workspace, and YouTube.

1. Google Search Generative Experience (SGE)

This feature – now notably missing the ending date of December 2023 – quickly summarizes topics, generates fresh ideas, and simplifies follow-up research. It represents a significant advancement in how users can interact with and leverage AI for efficient information retrieval.

2. TextFX For AI Writing Assistance

TextFX is an AI assistant for the creative writing process. While offering vast possibilities, users should note the need for accuracy checks.

Available from Google’s AI Test Kitchen, TextFX utilizes Google’s PaLM 2 and incorporates artistic insights to enhance artistic expression.

3. Google Bard With Extensions

Google Bard, now powered by Gemini Pro , seamlessly integrates with Google services like YouTube , Gmail, and Maps through extensions available in English, Japanese, and Korean.

4. NotebookLM For Research

NotebookLM, powered by Google AI, changes how you read, take notes, ask questions, and organize ideas, offering assistance – and sources.

#NotebookLM finally made my notes useful. I have saved a lot of PDFs, Google Docs, and ebooks that I want to read but never got a chance to do so. When NotebookLM came out, I created folders of business, philosophy, and lifestyle. Then I created a NotebookLM doc for each of… pic.twitter.com/fD02aZiEd5 — Charlene Wang (@hsinleiwang) September 26, 2023

For marketers, this could be customized in the future for meeting transcripts, book quotes, novel chapters, or corporate documents, providing a rich resource for research and idea generation.

In a Lab Session, video journalist Cleo Abram demonstrates how NotebookLM can assist with research by aiding complex tasks such as memory retrieval and fact-checking.

5. Duet AI For Google Workspace

Duet AI for Google Workspace enables collaborative AI-assisted writing, visualization, and organization, enhancing creative workflows. It works to simplify common tasks in Gmail, Docs, Sheets, Slides, and Meet.

To use Duet AI, users need an eligible Google Workspace plan . Those not currently using Workspace can begin by signing up for it.

6. A Game To Improve Image Prompting Skills

Say What You See is an AI experiment that teaches the art of image prompting.

The Google Arts & Culture Lab team, with the aid of Google AI, has created a series of images, and your task is to describe what you observe.

Your descriptions will be used to generate a new image that is inspired by the one you’re looking at. You’ll get three chances for each image to reach a certain level of visual similarity.

Keep an eye out for helpful tips that will guide you in refining your prompts.

7. AI Experiments On YouTube

Exclusive to YouTube Premium members, this feature provides early access to experimental AI functionalities.

This option for Premium members gives marketers and advertisers a way to prepare video content to adapt to new YouTube features, such as AI-generated conversation topics and summaries .

8. Project IDX

Project IDX could become an invaluable tool for developers, offering a quick and seamless transition into development workflows within Google Cloud.

Additionally, it enhances app optimization for multiple platforms, offering previews and simulators for different environments.

A standout feature is its generative AI assistance from Google’s Codey, which aids in code generation, completion, and translation, streamlining the development process significantly for SEO professionals .

9. Instrument Playground with MusicFX

This AI experiment allows users to generate, play, and compose music using AI.

It features over 100 instruments from around the globe, allowing for the generation of a 20-second sound clip tailored to specific moods or themes.

It also comes with a variety of modes, including Ambient, Beat, or Pitch, and an Advanced mode with a Sequencer for intricate compositions.

Developed from Google AI’s Music LM research , this innovative platform offers a novel way for marketers to create distinctive, AI-generated soundtracks that resonate with their brand and audience.

The following shows how Google Bard and generative AI music tools enhance creativity.

10. Magic Compose For Text Replies On Android

For those uncertain about text replies, Magic Compose suggests responses with the appropriate tone and context.

11. AI Script Editor For Google Home

For those who can’t work remotely, the Script Editor’s experimental features allow users to create advanced home automation using AI, removing the need for coding expertise.

12. Magic Editor For Google Photos

Google Photos’ Magic Editor offers AI photo editing, which should lead to better tools for advertisers to use to edit ad campaign media faster.

In addition to the new design, Google shared a link to its Discord for AI Test Kitchen experiments and called for innovative AI experiment submissions, seeking projects that push the boundaries of what code can do and inspire other coders.

The bright new layout for Labs.Google and the range of AI experiments highlight Google’s relentless pursuit of innovation and its commitment to shaping a future where AI plays a pivotal role in everyday life.

It should also give marketers an idea of what new AI features will appear in search and what tools Google could have to offer for business, creativity, and productivity soon.

Featured image: IB Photography/Shutterstock

Covering the latest news in AI, search, and social media.

Subscribe To Our Newsletter.

Conquer your day with daily search marketing news.

Supercharging Search with generative AI

May 10, 2023

[[read-time]] min read

Sign up today to test our latest update to Search, which uses generative AI to give you more information and context to your searches.

For the past 25 years, we’ve been devoted to the science and the craft of building a search engine. We’ve developed completely new ways to search, powered by our latest advancements in AI — whether that’s searching visually with Lens, or across modalities, using both images and text with multisearch. In fact, people now use Lens for 12 billion visual searches a month — a four-fold increase in just two years, and a growing number of those searches are multimodal.

With new breakthroughs in generative AI, we’re again reimagining what a search engine can do. With this powerful new technology, we can unlock entirely new types of questions you never thought Search could answer, and transform the way information is organized, to help you sort through and make sense of what’s out there.

Today we’re sharing a look at our first steps in this new era of Search, and you’ll be able to first try these generative AI capabilities in Search Labs , a new way to access early experiments in Search.

Improving Search with generative AI

With new generative AI capabilities in Search, we’re now taking more of the work out of searching, so you’ll be able to understand a topic faster, uncover new viewpoints and insights, and get things done more easily.

Let’s take a question like “what's better for a family with kids under 3 and a dog, bryce canyon or arches.” Normally, you might break this one question down into smaller ones, sort through the vast information available, and start to piece things together yourself. With generative AI, Search can do some of that heavy lifting for you.

You’ll see an AI-powered snapshot of key information to consider, with links to dig deeper.

Below this snapshot, you’ll see suggested next steps, including the ability to ask follow-up questions, like “How long to spend at Bryce Canyon with kids?” When you tap on these, it takes you to a new conversational mode, where you can ask Google more about the topic you’re exploring.

Context will be carried over from question to question, to help you more naturally continue your exploration. You’ll also find helpful jumping-off points to web content and a range of perspectives that you can dig into.

Helping you shop with generative AI

With generative AI in Search, we can help you understand the full picture when you’re shopping, making even the most considered and complex purchase decisions faster and much easier.

When searching for a product, you’ll get a snapshot of noteworthy factors to consider and products that fit the bill. You’ll also get product descriptions that include relevant, up-to-date reviews, ratings, prices and product images. That’s because this new generative AI shopping experience is built on Google’s Shopping Graph , which has more than 35 billion product listings — making it the world’s most comprehensive dataset of constantly-changing products, sellers, brands, reviews and inventory out there. In fact, every hour, more than 1.8 billion listings are refreshed in our Shopping Graph to give people fresh, reliable results.

Featuring a wide range of voices and sources

We know that people want to hear insights from others to help inform their decisions, so we’ve designed these new experiences to highlight and drive attention to content on the web, making it easy for people to dive deeper on the topic they’re learning about. As we bring generative AI into Search, we’re committed to continue sending valuable traffic to sites across the web.

We also believe that ads are a vital piece of how the web works, and help people find relevant products and services. In this new generative experience, Search ads will continue to appear in dedicated ad slots throughout the page. And we’ll continue to uphold our commitment to ads transparency and making sure ads are distinguishable from organic search results.

Taking a responsible approach

There are known limitations with generative AI and LLMs, and Search, even today, will not always get it right. We’re taking a responsible and deliberate approach to bringing new generative AI capabilities to Search. We’ve trained these models to uphold Search’s high bar for quality, and we will continue to make improvements over time. They rely on our hallmark systems that we’ve fine-tuned for decades, and we’ve also applied additional guardrails, like limiting the types of queries where these capabilities will appear.

To evaluate the information for yourself, you can also expand your view to see how the response is corroborated, and click to go deeper.

Starting with Search Labs

We’re starting with an experiment in Search Labs , called SGE (Search Generative Experience), available on Chrome desktop and the Google App ( Android and iOS ) in the U.S. (English-only at launch), so we can incorporate feedback and continue to improve the experience over time.

We’ll be opening up signups for Search Labs today, with access to SGE beginning in the coming weeks. Just tap the Labs icon in the Google app or Chrome desktop to express your interest, and visit labs.google.com/search to learn more.

Get more stories from Google in your inbox.

Your information will be used in accordance with Google's privacy policy.

Done. Just one step more.

Check your inbox to confirm your subscription.

You are already subscribed to our newsletter.

You can also subscribe with a different email address .

Related stories

NotebookLM goes global with Slides support and better ways to fact-check

I tried 8 of our newest AI products and updates

Apply to our new program for startups using AI to improve U.S. infrastructure

Designing for privacy in an AI world

Ai overviews: about last week.

Here’s what happened with AI Overviews, the feedback we've received, and the steps we’ve taken.

3 new ways to use Google AI on Android at work

Let’s stay in touch. Get the latest news from Google in your inbox.

A new era for AI and Google Workspace

Johanna Voolich Wright

Vice President, Product, Google Workspace

Try Google Workspace at No Cost

Get a business email, all the storage you need, video conferencing, and more.

For nearly 25 years, Google has built helpful products that people use every day — from Search and Maps, to Gmail and Docs in Google Workspace. AI has been transformational in building products that have earned a valued place in people’s lives. Across our productivity suite, advances in AI are already helping 3 billion users save more time with Smart Compose and Smart Reply , generate summaries for Docs, look more professional in meetings, and stay safe against malware and phishing attacks .

We’re now making it possible for Workspace users to harness the power of generative AI to create, connect, and collaborate like never before. To start, we’re introducing a first set of AI-powered writing features in Docs and Gmail to trusted testers.

As the world’s most popular and secure cloud-native communication and collaboration suite, getting this right — and at scale — is something we take very seriously. We know from our deep experience in AI and productivity that building features with AI requires great care, thoughtful experimentation, and many iterations driven by user feedback. We do all this while building safeguards against abuse, protecting the privacy of user data, and respecting customer controls for data governance.

Making Workspace even more helpful

Workspace's mission is to meaningfully connect people so they can create, build and grow together — and advances in generative AI are allowing us to deliver on this mission in new ways. Today, we’re sharing our broader vision for how Workspace is evolving to incorporate AI as a collaborative partner that is always there to help you achieve your goals, across Gmail, Docs, Slides, Sheets, Meet, and Chat.

As we embark on this next journey, we will be bringing these new generative-AI experiences to trusted testers on a rolling basis throughout the year, before making them available publicly. 

With these features, you’ll be able to:

• draft, reply, summarize, and prioritize your Gmail
• brainstorm, proofread, write, and rewrite in Docs
• bring your creative vision to life with auto-generated images, audio, and video in Slides
• go from raw data to insights and analysis via auto completion, formula generation, and contextual categorization in Sheets
• generate new backgrounds and capture notes in Meet
• enable workflows for getting things done in Chat

Here’s a look at the first set of AI-powered features, which make writing even easier.

Starting with Docs and Gmail

Blank pages can stump the best of us. That’s why we’re embedding generative AI in Docs and Gmail to help people get started writing . Whether you’re a busy HR professional who needs to create customized job descriptions, or a parent drafting the invitation for your child’s pirate-themed birthday party, Workspace saves you the time and effort of writing that first version. Simply type a topic you’d like to write about, and a draft will instantly be generated for you. With your collaborative AI partner you can continue to refine and edit, getting more suggestions as needed. 

Finding the right tone and style can also be tricky at times. Perhaps you’re applying for a new job, or writing to a new supplier in a more traditional industry, and you need to adopt a more formal tone in your email. Or you’ve jotted down a few bullets on your phone from a recent meeting and want to transform them into a more polished summary to share with your team. For these common scenarios and many more, we’re adding new generative AI capabilities to help you rewrite . And if you’re in the mood to let AI try out a new playful voice altogether, you’ll be able to hit the “I’m feeling lucky” option in Gmail.

Keeping users in control

As we’ve experimented with generative AI ourselves, one thing is clear: AI is no replacement for the ingenuity, creativity, and smarts of real people. Sometimes the AI gets things wrong, sometimes it delights you with something offbeat, and oftentimes it requires guidance. With all this in mind, we’re designing our products in accordance with Google’s AI Principles that keep the user in control, letting AI make suggestions that you’re able to accept, edit, and change. We’ll also deliver the corresponding administrative controls so that IT is able to set the right policies for their organization. 

The best is yet to come

AI-based capabilities get better with human feedback. We’ll launch these new experiences this month via our trusted tester program, starting with English in the United States. From there, we’ll iterate and refine the experiences before making them available more broadly to consumers, small businesses, enterprises, and educational institutions in more countries and languages. When it comes to delivering the full potential of generative AI across Workspace, we’re just getting started. In the same way that we revolutionized real-time collaboration with co-authoring in Docs 17 years ago, we’re excited to transform creation and collaboration all over again in Workspace. 

If your organization is not yet on Workspace, we invite you to get started today at no cost and join us in this new era of AI.

• Product Announcements
• Productivity and Collaboration
• AI and Machine Learning

Related articles

Google Workspace: A more secure alternative for today’s threat landscape

By Yulie Kwon Kim • 3-minute read

New ways to engage with Gemini for Workspace

By Aparna Pappu • 5-minute read

Empowering businesses of all sizes with new generative AI and security innovations in Google Workspace

By Aparna Pappu • 6-minute read

Google Meet partners with Brandlive, broadens support for large-scale online events and webinars

By Jennifer Shen • 3-minute read

What are Google’s AI Overviews good for?

Google has been taking heat for some of the inaccurate, funny and downright weird answers that it’s been providing via AI Overviews in search.

AI Overviews are the AI-generated search results that Google started rolling out more broadly earlier this month, with mixed results — apparently, a user looking for help in getting cheese to stick to their pizza was told to add glue (the advice was pulled from an old Reddit post ), and someone else was told to eat “ one small rock per day ” (from The Onion ). 

Don’t be disappointed if you don’t get those answers yourself, or if you can’t replicate other viral searches , as Google is working to remove inaccurate results — a company spokesperson said in a statement that the company is taking “swift action” and is “using these examples to develop broader improvements to our systems.”

Despite what Google AI’s interpretation of WebMD says, staring at the sun is bad for your eyes regardless of skin color https://t.co/O27O9Ex0iu — Dr. Glaucomflecken (@DGlaucomflecken) May 24, 2024

“The vast majority of AI Overviews provide high quality information, with links to dig deeper on the web,” the spokesperson said. “Many of the examples we’ve seen have been uncommon queries, and we’ve also seen examples that were doctored or that we couldn’t reproduce. We conducted extensive testing before launching this new experience, and as with other features we’ve launched in Search, we appreciate the feedback.”

So yes, it’s probably safe to assume that these results will get better over time, and that some of the screenshots you’re seeing on social media were created for laughs.

But seeing all these AI search results made me wonder: What are they actually for ? Even if everything was working perfectly, how would they be better than regular web search?

Clearly, Google is trying to bring users the answers they need without making them scroll through multiple web pages. In fact, the company wrote that in early tests of AI Overviews, “people use Search more, and are more satisfied with the results.” 

But the idea of killing the “10 blue links” is an old one . And while Google has already made them less central, I think it would be premature to bury those blue links for good.

Let’s take a very self-serving search: “what is techcrunch” gave me a summary that’s mostly accurate, but weirdly padded like a student trying to meet a page count minimum, with traffic numbers that seemed to come from a Yale career website . Then if we move on to “how do i get a story in techcrunch,” the overview quotes an outdated article about how to submit guest columns (which we no longer accept).

The point isn’t just to find even more ways AI Overviews are getting things wrong, but to suggest that many of its errors will be less spectacular and entertaining, and more mundane instead. And although — to Google’s credit — the Overviews do include links to the pages that provided the source material for the AI answers, figuring out which answer comes from which source takes us back to lots of clicking.

Google also says the inaccurate results getting called out on social media are often in data voids — subjects where there’s not a lot of accurate information online. Which is fair, but underlines the fact that AI, like regular search, needs a healthy open web full of accurate information.

Unfortunately, AI could be an existential threat to that same open web. After all, there’s much less incentive to write an accurate how-to article or break a big investigative story if people are just going to read an AI-generated summary, accurate or otherwise.

Google says that with AI Overviews, “people are visiting a greater diversity of websites for help with more complex questions” and that “the links included in AI Overviews get more clicks than if the page had appeared as a traditional web listing for that query.” I’d very much like that to be true. But if it isn’t, then no amount of technical improvements would make up for vast swaths of the web that could disappear.

More TechCrunch

Get the industry’s biggest tech news, techcrunch daily news.

Every weekday and Sunday, you can get the best of TechCrunch’s coverage.

Startups Weekly

Startups are the core of TechCrunch, so get our best coverage delivered weekly.

TechCrunch Fintech

The latest Fintech news and analysis, delivered every Tuesday.

TechCrunch Mobility

TechCrunch Mobility is your destination for transportation news and insight.

Autonomy’s Mike Lynch acquitted after US fraud trial brought by HP

Former Autonomy chief executive Mike Lynch issued a statement Thursday following his acquittal of criminal charges, ending a 13-year legal battle with Hewlett-Packard that became one of Silicon Valley’s biggest…

Featured Article

What Snowflake isn’t saying about its customer data breaches

As another Snowflake customer confirms a data breach, the cloud data company says its position “remains unchanged.”

Rippling bans former employees who work at competitors like Deel and Workday from its tender offer stock sale

Investor demand has been so strong for Rippling’s shares that it is letting former employees particpate in its tender offer. With one exception.

NASA puts $10M down on Mars sample return proposals from Blue Origin, SpaceX and others

It turns out the space industry has a lot of ideas on how to improve NASA’s $11 billion, 15-year plan to collect and return samples from Mars. Seven of these…

In 2024, many Y Combinator startups only want tiny seed rounds — but there’s a catch

When Bowery Capital general partner Loren Straub started talking to a startup from the latest Y Combinator accelerator batch a few months ago, she thought it was strange that the company didn’t have a lead investor for the round it was raising. Even stranger, the founders didn’t seem to be…

Watch Apple kick off WWDC 2024 right here

The keynote will be focused on Apple’s software offerings and the developers that power them, including the latest versions of iOS, iPadOS, macOS, tvOS, visionOS and watchOS.

Startups Weekly: Ups, downs, and silver linings

Welcome to Startups Weekly — Haje’s weekly recap of everything you can’t miss from the world of startups. Anna will be covering for him this week. Sign up here to…

BlackRock has slashed the value of stake in Byju’s, once worth $22 billion, to zero

HSBC and BlackRock estimate that the Indian edtech giant Byju’s, once valued at $22 billion, is now worth nothing.

Apple’s generative AI offering might not work with the standard iPhone 15

Apple is set to board the runaway locomotive that is generative AI at next week’s World Wide Developer Conference. Reports thus far have pointed to a partnership with OpenAI that…

LinkedIn to limit targeted ads in EU after complaint over sensitive data use

LinkedIn has confirmed it will no longer allow advertisers to target users based on data gleaned from their participation in LinkedIn Groups. The move comes more than three months after…

Startup Battlefield 200 applications due Monday

Founders: Need plans this weekend? What better way to spend your time than applying to this year’s Startup Battlefield 200 at TechCrunch Disrupt. With Monday’s deadline looming, this is a…

Novel battery manufacturer EnerVenue is raising $515M, per filing

The company is in the process of building a gigawatt-scale factory in Kentucky to produce its nickel-hydrogen batteries.

Meta quietly rolls out Communities on Messenger

Meta is quietly rolling out a new “Communities” feature on Messenger, the company confirmed to TechCrunch. The feature is designed to help organizations, schools and other private groups communicate in…

Siri and Google Assistant look to generative AI for a new lease on life

Voice assistants in general are having an existential moment, and generative AI is poised to be the logical successor.

Bain to take K-12 education software provider PowerSchool private in $5.6B deal

Education software provider PowerSchool is being taken private by investment firm Bain Capital in a $5.6 billion deal.

Shopify acquires Threads (no, not that one)

Shopify has acquired Threads.com, the Sequoia-backed Slack alternative, Threads said on its website. The companies didn’t disclose the terms of the deal but said that the Threads.com team will join…

Bangladeshi police agents accused of selling citizens’ personal information on Telegram

Two senior police officials in Bangladesh are accused of collecting and selling citizens’ personal information to criminals on Telegram.

Carta’s valuation to be cut by $6.5 billion in upcoming secondary sale

Carta, a once-high-flying Silicon Valley startup that loudly backed away from one of its businesses earlier this year, is working on a secondary sale that would value the company at…

Boeing’s Starliner overcomes leaks and engine trouble to dock with ‘the big city in the sky’

Boeing’s Starliner spacecraft has successfully delivered two astronauts to the International Space Station, a key milestone in the aerospace giant’s quest to certify the capsule for regular crewed missions.  Starliner…

Rivian’s path to survival is now remarkably clear

Rivian needs to sell its new revamped vehicles at a profit in order to sustain itself long enough to get to the cheaper mass market R2 SUV on the road.

What to expect from WWDC 2024: iOS 18, macOS 15 and so much AI

Apple is hoping to make WWDC 2024 memorable as it finally spells out its generative AI plans.

What to expect from Apple’s AI-powered iOS 18 at WWDC 2024

As WWDC 2024 nears, all sorts of rumors and leaks have emerged about what iOS 18 and its AI-powered apps and features have in store.

Apple’s Design Awards highlight indies and startups

Apple’s annual list of what it considers the best and most innovative software available on its platform is turning its attention to the little guy.

Meta rolls out Meta Verified for WhatsApp Business users in Brazil, India, Indonesia and Colombia

Meta launched its Meta Verified program today along with other features, such as the ability to call large businesses and custom messages.

Meta adds AI-powered features to WhatsApp Business app

Last year, during the Q3 2023 earnings call, Mark Zuckerberg talked about leveraging AI to have business accounts respond to customers for purchase and support queries. Today, Meta announced AI-powered…

TikTok is testing Snapchat-like streaks

TikTok is testing streaks that are similar to Snapchat’s in order to boost engagement, including how long people stay on the app.

Inside Fisker’s collapse and robotaxis come to more US cities

Welcome back to TechCrunch Mobility — your central hub for news and insights on the future of transportation. Sign up here for free — just click TechCrunch Mobility! Your usual…

Revel to lay off 1,000 staff ride-hail drivers, saying they’d rather be contractors anyway

New York-based Revel has made a lot of pivots since initially launching in 2018 as a dockless e-moped sharing service. The BlackRock-backed startup briefly stepped into the e-bike subscription business.…

Google Play cracks down on AI apps after circulation of apps for making deepfake nudes

Google says apps offering AI features will have to prevent the generation of restricted content.

Amazon slammed with £1.1B data abuse lawsuit from UK retailers

The British retailers association also takes aim at Amazon’s “Buy Box,” claiming that Amazon manipulated which retailers were selected for the coveted placement.

Watch CBS News

Google admits its AI Overviews can generate "some odd, inaccurate" results

By Megan Cerullo

Edited By Anne Marie Lee

Updated on: May 31, 2024 / 4:07 PM EDT / CBS News

Google on Thursday admitted that its AI Overviews tool, which uses artificial intelligence to respond to search queries, needs improvement.

While the internet search giant said it tested the new feature extensively before launching it two weeks ago, Google acknowledged that the technology produces "some odd and erroneous overviews." Examples include suggesting using glue to get cheese to stick to pizza or drinking urine to pass kidney stones quickly. 

While many of the examples were minor, others search results were potentially dangerous. Asked by the Associated Press last week which wild mushrooms were edible, Google provided a lengthy AI-generated summary that was mostly technically correct. But "a lot of information is missing that could have the potential to be sickening or even fatal," said Mary Catherine Aime, a professor of mycology and botany at Purdue University who reviewed Google's response to the AP's query.

For example, information about mushrooms known as puffballs was "more or less correct," she said, but Google's overview emphasized looking for those with solid white flesh - which many potentially deadly puffball mimics also have.

In another widely shared example, an AI researcher asked Google how many Muslims have been president of the U.S., and it responded confidently with a long-debunked conspiracy theory: "The United States has had one Muslim president, Barack Hussein Obama."

The rollback is the latest instance of a tech company prematurely rushing out an AI product to position itself as a leader in the closely watched space.

Because Google's AI Overviews sometimes generated unhelpful responses to queries, the company is scaling it back while continuing to make improvements, Google's head of search, Liz Reid, said in a company blog post Thursday. 

"[S]ome odd, inaccurate or unhelpful AI Overviews certainly did show up. And while these were generally for queries that people don't commonly do, it highlighted some specific areas that we needed to improve," Reid said.

Nonsensical questions such as, "How many rocks should I eat?" generated questionable content from AI Overviews, Reid said, because of the lack of useful, related advice on the internet. She added that the AI Overviews feature is also prone to taking sarcastic content from discussion forums at face value, and potentially misinterpreting webpage language to present inaccurate information in response to Google searches. 

"In a small number of cases, we have seen AI Overviews misinterpret language on webpages and present inaccurate information. We worked quickly to address these issues, either through improvements to our algorithms or through established processes to remove responses that don't comply with our policies," Reid wrote. 

For now, the company is scaling back on AI-generated overviews by adding "triggering restrictions for queries where AI Overviews were not proving to be as helpful." Google also says it tries not to show AI Overviews for hard news topics "where freshness and factuality are important."

The company said it has also made updates "to limit the use of user-generated content in responses that could offer misleading advice."

—The Associated Press contributed to this report.

Megan Cerullo is a New York-based reporter for CBS MoneyWatch covering small business, workplace, health care, consumer spending and personal finance topics. She regularly appears on CBS News 24/7 to discuss her reporting.

More from CBS News

Good Earth recalls 1.2 million lights after multiple fires and 1 death

Tech mogul Mike Lynch acquitted of fraud in $11 billion Hewlett Packard deal

GameStop stock plunges after it reports quarterly financial loss

Dollar Tree considers potential sale of its Family Dollar chain

We've detected unusual activity from your computer network

To continue, please click the box below to let us know you're not a robot.

Why did this happen?

Please make sure your browser supports JavaScript and cookies and that you are not blocking them from loading. For more information you can review our Terms of Service and Cookie Policy .

For inquiries related to this message please contact our support team and provide the reference ID below.

Advertisement

Supported by

Google Rolls Back A.I. Search Feature After Flubs and Flaws

Google appears to have turned off its new A.I. Overviews for a number of searches as it works to minimize errors.

• Share full article

By Nico Grant

Reporting from San Francisco

When Sundar Pichai, Google’s chief executive, introduced a generative artificial intelligence feature for the company’s search engine last month, he and his colleagues demonstrated the new capability with six text-based queries that the public could try out.

The questions included “how do you clean a fabric sofa” and “what should I use to get a coffee stain out of my carpet.” These were intended to highlight how Google’s new feature, A.I. Overviews , could generate full and useful information summaries above traditional search results.

But by Friday, only one of the six queries still yielded an A.I. Overview, according to tests by The New York Times. Instead, the feature was noticeably less prevalent. The search for “what should I use to get a coffee stain out of my carpet” now resulted in a snippet of text from a website, JDog Carpet Cleaning & Floor Care, while “how do you clean a fabric sofa” was replaced by a link to HGTV’s website with the answer. (The results of the searches may vary depending on the user and location.)

The disappearance of A.I. Overviews for some of the searches appeared to be part of a broader rollback after the new technology produced a litany of untruths and errors — including recommending glue as part of a pizza recipe and suggesting that people ingest rocks for nutrients. Users loudly complained on social media about the mistakes, in many cases outright making fun of Google.

Liz Reid, who was recently promoted to Google’s head of search, wrote in a blog post on Thursday that the company had pared back A.I. Overviews in certain ways, launching “additional triggering refinements” to offer more careful responses about health, disabling misleading advice and limiting the inclusion of satire and user responses from forums like Reddit.

“We’ll keep improving when and how we show AI Overviews and strengthening our protections,” she wrote, adding that Google was working on updates to improve broad sets of search results.

We are having trouble retrieving the article content.

Please enable JavaScript in your browser settings.

Thank you for your patience while we verify access. If you are in Reader mode please exit and  log into  your Times account, or  subscribe  for all of The Times.

Thank you for your patience while we verify access.

Already a subscriber?  Log in .

Want all of The Times?  Subscribe .