Skip to main content

Filed under:

From ChatGPT to Gemini: how AI is rewriting the internet

Big players, including Microsoft, with Copilot, Google, with Gemini, and OpenAI, with ChatGPT-4, are making AI chatbot technology previously restricted to test labs more accessible to the general public.

How do these large language model (LLM) programs work? OpenAI’s GPT-3 told us that AI uses “a series of autocomplete-like programs to learn language” and that these programs analyze “the statistical properties of the language” to “make educated guesses based on the words you’ve typed previously.” 

Or, in the words of James Vincent, a human person: “These AI tools are vast autocomplete systems, trained to predict which word follows the next in any given sentence. As such, they have no hard-coded database of ‘facts’ to draw on — just the ability to write plausible-sounding statements. This means they have a tendency to present false information as truth since whether a given sentence sounds plausible does not guarantee its factuality.”

But there are so many more pieces to the AI landscape that are coming into play (and so many name changes — remember when we were talking about Bing and Bard last year?), but you can be sure to see it all unfold here on The Verge.

  • Emma Roth

    Jun 29, 2023

    Emma Roth

    Forget machine learning...

    what about machine unlearning?

    Since larger AI models “tend to memorize details of their training set,” including information that may have been deleted from the database they're trained on, Google wants to find ways for AI to “forget” certain things.

    That’s why the company is holding a machine unlearning challenge, in which researchers must make AI systems forget “a certain subset of the training images... to protect the privacy or rights of the individuals concerned.”


  • Emma Roth

    Jun 29, 2023

    Emma Roth

    Bing will now surface AI-generated buying guides

    Bing’s AI-generated buying guide for the Surface Headphones 2
    Image: Microsoft

    Microsoft is bringing AI-generated buying guides to Bing, the company announced on Thursday. Now when you search for things like “college supplies,” Bing will surface AI-made guides that offer comparisons between products in the same categories, such as headphones or laptops.

    The link to Bing’s buying guide will appear at the very top of the search engine’s results. Clicking into the guide reveals an AI-generated summary about whatever you’re looking for, along with a list of products relevant to your search. Hit the Compare button, and you’ll see a chart that pulls specs from the product manufacturers’ websites and compares them side by side. Microsoft notes that you can access buying guides both through Bing chat and in the Edge sidebar.

    Read Article >
  • Justine Calma

    Jun 28, 2023

    Justine Calma

    AI-generated tweets might be more convincing than real people, research finds

    An image of OpenAI’s logo, which looks like a stylized and symmetrical braid.
    Image: OpenAI

    People apparently find tweets more convincing when they’re written by AI language models. At least, that was the case in a new study comparing content created by humans to language generated by OpenAI’s model GPT-3.

    The authors of the new research surveyed people to see if they could discern whether a tweet was written by another person or by GPT-3. The result? People couldn’t really do it. The survey also asked them to decide whether the information in each tweet was true or not. This is where things get even dicier, especially since the content focused on science topics like vaccines and climate change that are subject to a lot of misinformation campaigns online.

    Read Article >
  • Emma Roth

    Jun 27, 2023

    Emma Roth

    ChatGPT on iOS now comes with Bing built-in

    An image of OpenAI’s logo, which looks like a stylized and symmetrical braid.
    Image: OpenAI

    The ChatGPT app on iOS now lets paid users access information surfaced by Microsoft’s Bing. In the app’s most recent update, OpenAI says subscribers who pay for its $20 per month ChatGPT Plus plan receive up-to-date information from the web.

    Microsoft, which made a multibillion-dollar investment into OpenAI earlier this year, announced Bing would become the default search engine for ChatGPT during its Build event in May.

    Read Article >
  • Wes Davis

    Jun 24, 2023

    Wes Davis

    Google AI raters that were laid off for complaining about working conditions have gotten their jobs back.

    On May 31st, a group of six AI chatbot trainers employed by Google subcontractor Appen were fired after they spoke out about how their poor working conditions could make Google’s Bard chatbot dangerous.

    The Alphabet Workers Union now says the workers’ jobs have been reinstated with backpay. AWU provided The Verge with a copy of the email from Appen’s RaterLabs explaining the decision:

    Based on our continuing consideration and review of our business needs, we have determined that some recent reductions of our workforce were not necessary and can be reversed. Consequently, we are pleased to offer you the option to return to work with RaterLabs.

    Internal complaints about Bard previously called it “a pathological liar” and begged the company not to launch it.


  • Umar Shakir

    Jun 22, 2023

    Umar Shakir

    Google’s spreadsheet-generating AI feature is rolling out now.

    Google is rolling out an AI feature for Sheets that auto-builds custom templates based on a text prompt, but it’s only coming to Workspace Labs users to start (via 9to5Google).

    Last month, the company showed off Duet AI, its answer to Microsoft’s Copilot (and perhaps Clippy), which contains a set of generative AI tools to help users write email responses in Gmail, create images from text in Google Slides, proofread documents in Google Docs, and more.


  • Adi Robertson

    Jun 22, 2023

    Adi Robertson

    Remember the lawyer who submitted fake case law from ChatGPT?

    The sanctions for both lawyers involved are in, and they’re pretty embarrassing: a $5,000 fine, plus they have to send a letter to every real judge named in the made-up cases. For what it’s worth, Steven Schwartz says he thought ChatGPT was just a “super search engine.”


  • Emma Roth

    Jun 21, 2023

    Emma Roth

    Opera launches revamped browser equipped with an AI sidekick

    An image showing the Aria AI tool in Opera One.
    Image: Opera

    Opera has launched Opera One — a new version of the browser that comes packaged with an AI-powered chatbot called Aria. Just like the Bing chatbot on Microsoft Edge, Opera’s AI assistant lives within the browser’s sidebar, where you can have it answer questions using real-time information, generate text or code, brainstorm ideas, and more.

    The built-in chatbot is powered by Opera’s Composer AI engine and connects to OpenAI’s GPT model. To use the tool, you need to sign up for an Opera account if you don’t have one already. Once that’s done, you can click the Aria icon on the left side of the screen to start chatting. While Opera first started testing the revamped version of the browser in May, now it’s available to everyone who downloads it.

    Read Article >
  • Emma Roth

    Jun 21, 2023

    Emma Roth

    You.com, the AI search engine, is launching a paid subscription.

    You.com was one of the first search engines to incorporate a ChatGPT-style chatbot, and now it's one of the first to put unlimited access to AI tools behind a paywall.

    The new YouPro subscription costs $14.99 / month (but currently costs $9.99 / month for a limited time), and gives you access to unlimited AI chat searches, unlimited AI image generation, unlimited text generation, and more. Will Microsoft and Google be next?


  • Elizabeth Lopatto

    Jun 20, 2023

    Elizabeth Lopatto

    The former president of FTX US is back with — you guessed it! — an AI company.

    Yep, it uses ChatGPT. It only raised $5 million of the $10 million it wanted earlier this year so it seems like I’m not the only who’s a little skeptical of this.


  • Umar Shakir

    Jun 16, 2023

    Umar Shakir

    The new Bing widget on iOS is a chatbot shortcut.

    Microsoft’s releasing a new Bing widget for iPhone users — it was already available on Android — designed to get you straight to engaging with its AI chatbot. There are two styles of the widget, one with wallpaper and a plain one.

    Additionally, Microsoft has improved text-to-speech support in 38 languages, including Arabic, Croatian, Hebrew, Hindi, Korean, Lithuanian, Polish, Tamil, and Urdu, and said it improved the responsiveness of the voice input button.


    animated GIF showing the iOS Bing widget, which launches the Bing app when touched
    The iOS Bing app auto-launches if you tap on the widget, and lets you initiate a chat quickly. The experience feels tighter assuming the Bing app is already running.
    GIF: Umar Shakir / The Verge
  • Richard Lawler

    Jun 15, 2023

    Richard Lawler

    Google recently warned employees about submitting sensitive data to its Bard chatbot.

    Elizabeth Lopatto wrote just last month that Big Tech is warning all of us about the privacy issues with AI chatbots, as companies like Samsung warn employees about giving them sensitive information that could expose trade secrets.

    Now Reuters reports:

    A Google privacy notice updated on June 1 also states: “Don’t include confidential or sensitive information in your Bard conversations.”


  • James Vincent

    Jun 15, 2023

    James Vincent

    Google’s EU launch of AI chatbot Bard delayed by privacy concerns

    A screenshot of Bard saying, “Bard is an experiment.”
    Image: Google

    Google recently expanded access to its AI chatbot Bard to 180 new countries and territories. But not featured on the list? Any European Union (EU) nations.

    This is due to Google failing to answer privacy concerns from the Irish Data Protection Commission or DPC — the regulator for Google’s Dublin-based EU operations.

    Read Article >
  • James Vincent

    Jun 14, 2023

    James Vincent

    OpenAI reportedly trained its AI models on YouTube.

    That’s according to a report from The Information on the value for Google of YouTube as an AI training dataset. The fact that OpenAI scraped YouTube isn’t surprising, but the company is famously secretive about its training data, partly for competition reasons, and partly, it’s thought, to stymie potential lawsuits.

    YouTube’s terms of service forbid using content for anything other than “personal, non-commercial use,” but it’s an open secret in the AI industry that everyone is scraping the web constantly. If Google protests too much, it would end up incriminating itself.


  • Alex Heath

    Jun 11, 2023

    Alex Heath

    Meta is giving its employees access to an AI chatbot that is trained on internal data.

    Another scoop from the last issue of Command Line:

    Meta has built an internal AI chatbot called Metamate that uses company data to help employees summarize meetings, write code, and debug features. Employees will be able to create their own prompts and share them with colleagues. 

    The company is starting to roll it out internally to a small group now. I wrote in a previous issue that Meta was talking to Microsoft and OpenAI about powering this tool, but now I’m told that it landed on using its own separate, in-house model.

    You can subscribe at the link below to get the full issue in your inbox.


  • Emma Roth

    Jun 9, 2023

    Emma Roth

    Bing’s chatbot now lets you ask questions with your voice on desktop

    An image showing the Edge logo
    Image: The Verge

    Microsoft is bringing Bing’s voice mode to Edge on desktop. With this latest update, you can now ask Bing’s chatbot questions using your voice — and it will respond in a voice of its own.

    Just like the feature on the mobile version of Edge, you can select the microphone button on Bing’s chat box and then ask your question aloud. Microsoft says its chatbot currently supports English, Japanese, French, German, and Mandarin but notes that additional languages are “on the way.”

    Read Article >
  • James Vincent

    Jun 9, 2023

    James Vincent

    OpenAI sued for defamation after ChatGPT fabricates legal accusations against radio host

    ChatGPT Official App Photo Illustrations
    Photo by Jakub Porzycki/NurPhoto via Getty Images

    OpenAI has been hit with what appears to be the first defamation lawsuit responding to false information generated by ChatGPT.

    A radio host in Georgia, Mark Walters, is suing the company after ChatGPT stated that Walters had been accused of defrauding and embezzling funds from a non-profit organization. The system generated the information in response to a request from a third party, a journalist named Fred Riehl. Walters’ case was filed June 5th in Georgia’s Superior Court of Gwinnett County and he is seeking unspecified monetary damages from OpenAI.

    Read Article >
  • Jun 8, 2023

    Emma Roth and Alex Heath

    Meta’s first generative AI feature will be AI stickers in Messenger

    An image showing the Messenger logo
    Illustration by Alex Castro / The Verge

    Meta is bringing AI-generated stickers to Messenger as part of a slew of related features it’s working on for its social apps.

    During a companywide meeting today that The Verge listened to, Ahmad Al-Dahle, Meta’s vice president of AI, told employees that the company will leverage its image generation model to let users create stickers based on text prompts. Employees will begin testing the feature internally before it’s made available to the public.

    Read Article >
  • Jess Weatherbed

    Jun 5, 2023

    Jess Weatherbed

    Bing AI’s training wheels are slowly being removed.

    Microsoft has once again softened the limitations it placed on the Bing AI Chatbot in February as part of its efforts to correct the search engine’s weird behavior.

    Chat’s with the bot can now run for 30 turns (up from the previous 20), and the daily turn limit has increased to 300. This doesn’t mean the previous issues have been resolved though, so proceed with caution.


  • Wes Davis

    Jun 4, 2023

    Wes Davis

    Engineered Arts plugged its lifelike robot into ChatGPT.

    When we wrote about Engineered Arts’ Ameca android last year, the company said it wants to integrate chatbot functionality, and since then it has done so, using one of the most prominent chatbots, ChatGPT-3 (GPT-4 was too slow). In his conversation with The Verge, Engineered Arts CEO Will Jackson said:

    It’s amazing the simple things you can do to make a machine look sentient.

    The most human-like trick from the video below was not so much its response, but its double-take after being told “you stink.”


  • JPMorgan, Deutsche Bank, and others are pushing into AI.

    There’s a bunch of examples here but frankly, the nut of the piece is pretty far down:

    Bankers have a fiduciary duty not to trade on unreliable information. That’s an issue as use of AI expands.

    If you can’t explain where your information came from, is it reliable? Oh well, the hedge funds are at it, too.

    “Isn’t this sort of exciting?” writes Matt Levine. “The widespread use of relatively early-stage AI will introduce new ways of making mistakes into finance.”


  • Google has already driven ad prices into the basement. Now, with AI, will it destroy online media completely?

    “This is a facet of the larger AI story — which is to say it’s about automation,” writes John Herrman “But it’s also a story of a large platform deciding to compete more aggressively in the marketplace it controls.”

    It’s unclear whether this is what users want. But there is a dark new future forming, one that might suggest fun new antitrust laws!


  • Elizabeth Lopatto

    May 31, 2023

    Elizabeth Lopatto

    Meet the kids following the Pied Piper!

    “One appeal of generative A.I. is that it offers something for every would-be entrepreneur.” Yes, it is nice to see the hype machine in full effect, isn’t it? Here’s a sentence we should revisit in a year: “And unlike crypto, especially now, A.I. is a more credible field to be in for mainstream techies.” Anyway, every single one of these children would have gone to work on Wall Street before the year of our lord 2008.


  • Elizabeth Lopatto

    May 31, 2023

    Elizabeth Lopatto

    Hey, remember that eating disorder helpline that fired its unionizing staff and replaced them with a chatbot?

    You are never gonna guess what happened.

    That’s right, the chatbot started giving people with eating disorders advice on losing weight! (See also: Vice, Daily Dot)


  • May 30, 2023

    Wes Davis and Richard Lawler

    Nvidia became a $1 trillion company thanks to the AI boom

    Illustration of the Nvidia wordmark on a green and black background.
    Illustration by Alex Castro / The Verge

    Nvidia temporarily became a $1 trillion company on Tuesday morning, with its rising valuation fueled by tech companies big and small racing to add generative artificial intelligence tools to their products. AI tools made up the vast bulk of recent Google I/O and Microsoft Build presentations, and Nvidia’s chips make it a key supplier for companies trying to build something with AI.

    Its valuation pushed past the trillion-dollar benchmark as trading opened today at just over $405 per share, putting it in the rarified air previously occupied by only a few large companies such as Apple and Microsoft after they surpassed the significant mark in August 2018 and August 2019, respectively. Amazon and Google are the other tech stocks in the club, and Meta is a former member.

    Read Article >