Filed under:By Umar Shakir, a news writer fond of the electric vehicle lifestyle and things that plug in via USB-C. He spent over 15 years in IT support before joining The Verge. Big players, including Microsoft, with its Bing AI (and Copilot), Google, with Bard, 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 there are going to be problems — but you can be sure to see it all unfold here on The Verge.TODAY, 22 seconds agoRichard LawlerMicrosoft has announced that Bing Chat now supports Visual Search (still based on OpenAI’s GPT-4 model that was announced in March with image inputs, while Google Bard added image search in May) to “understand the context of an image, interpret it, and answer questions about it.”The new multimodal search capabilities are rolling out now on desktop and in the Bing mobile app, and Microsoft says it’s working to add it to Bing Chat Enterprise “over time.”TODAY, Two hours agoRichard LawlerThe Vox podcast Unexplainable has started a two-part series on AI by interviewing Sam Bowman, a researcher for Anthropic and professor at NYU, about how AI tools work and how you get from “a really fancy autocomplete tool” to something that seems more like a virtual assistant.With these neural networks [e.g., the type of AI ChatGPT uses], there’s no concise explanation. There’s no explanation in terms of things like checkers moves or strategy or what we think the other player is going to do. All we can really say is just there are a bunch of little numbers and sometimes they go up and sometimes they go down. And all of them together seem to do something involving language. We don’t have the concepts that map onto these neurons to really be able to say anything interesting about how they behave.TODAY, Two hours agoEmilia DavidMeta announced it’s open-sourcing its large language model LLaMA 2, making it free for commercial and research use and going head-to-head with OpenAI’s free-to-use GPT-4, which powers tools like ChatGPT and Microsoft Bing. Meta announced the move as part of Microsoft’s Inspire event, noting its support for Azure and Windows and a “growing” partnership between the two companies. At the same time, Microsoft revealed more details about the AI tools built into its 360 platform and how much those will cost. Qualcomm also announced it is working with Meta to bring LLaMa to laptops, phones, and headsets starting from 2024 onward for AI-powered apps that work without relying on cloud services. Jul 17Wes DavisTemplate-based website builder Wix has announced that, soon, it’ll let you create entire websites by typing a description into a box and answering a few follow-up questions. Everything, from the design to text and images, will then be automatically generated for you, and from the looks of things, it’ll be pretty fast.In the past, companies like Wix and WordPress.com have let you create websites using templates you can tweak to your liking. But Wix says its new AI Site Generator feature goes beyond templates, using AI and algorithms to create a “unique” website. Jul 16Wes DavisCode in Samsung Internet Browser version 22.0.0.54 has several references to ChatGPT, as spotted by Android Authority.Android Authority speculates that the integration as is would only be used to generate page summaries. Samsung is reportedly open to replacing Google as its default search engine, possibly due to Microsoft’s quick AI moves, so the idea of the company exploring its options isn’t far-fetched.[Android Authority]Jul 15Wes DavisNobody wants AI companies to scrape their data. Not artists, not writers, not social media companies, not news outlets. There are many ways parts of the internet are pushing back against AI data scraping, says The New York Times:Their protests have taken different forms. Writers and artists are locking their files to protect their work or are boycotting certain websites that publish A.I.-generated content, while companies like Reddit want to charge for access to their data. At least 10 lawsuits have been filed this year against A.I. companies, accusing them of training their systems on artists’ creative work without consent.That’s to say nothing of the numerous strikes going on in Hollywood over, in large part, the growing use of AI in entertainment.[The New York Times]Jul 14Umar ShakirAs rumors mount for a “Windows 12” release for next year, Microsoft confirmed and named a new Windows 11 23H2 update planned for the fourth quarter of 2023. Microsoft hasn’t yet revealed what’s in it, but the company does have many unreleased features in the beta testing pipeline that could be included. One of the big new items is Windows Copilot, which takes the GPT generative AI tech being used in Bing and Microsoft (Office) 365 and gives users an OS-wide personal assistant — and is available to test now.Jul 13Emilia DavidThe Federal Trade Commission (FTC) is investigating ChatGPT creator OpenAI over possible consumer harm through its data collection and the publication of false information. First reported by The Washington Post, the FTC sent a 20-page letter to the company this week. The letter requests documents related to developing and training its large language models, as well as data security. Jul 13Emma RothOpenAI will train its AI models on The Associated Press’ news stories for the next two years, thanks to an agreement first reported by Axios. The deal between the two companies will give OpenAI access to some of the content in AP’s archive as far back as 1985.As part of the agreement, AP will gain access to OpenAI’s “technology and product expertise,” although it’s not clear exactly what that entails. AP has long been exploring AI features and began generating reports about company earnings in 2014. It later leveraged the technology to automate stories about Minor League Baseball and college sports. Jul 13Jay PetersGoogle is adding some new features to its Bard AI chatbot, including the ability for Bard to speak its answers to you and for it to respond to prompts that also include images. The chatbot is also now available in much of the world, including the EU.In a blog post, Google is positioning Bard’s spoken responses as a helpful way to “correct pronunciation of a word or listen to a poem or script.” You’ll be able to hear spoken responses by entering a prompt and selecting the sound icon. Spoken responses will be available in more than 40 languages and are live now, according to Google.Jul 12David PierceProject Tailwind, the AI-backed note-taking tool that Google launched at this year’s I/O developer conference, is rebranding. It’s now known as NotebookLM, and it’s launching today to “a small group of users in the US,” according to a Google blog post. (The LM stands for Language Model because Google really wants to make sure you don’t forget about all the AI in here.) The product hasn’t changed, though: Google’s still trying to give users their own personal AI, trained on their data and notes and able to help them make sense of it all.The core of NotebookLM seems to actually start in Google Docs. (“We’ll be adding additional formats soon,” the blog post says.) Once you get access to the app, you’ll be able to select a bunch of docs and then use NotebookLM to ask questions about them and even create new stuff with them. Jul 12Jay PetersSeriously! Sadly, the company has apparently deprioritized the work on the project, according to CNBC. I guess I’ll just have to ask Bard to respond to me in a Gen Z way instead.[CNBC]Jul 11Emma RothClaude, the AI chatbot that Anthropic bills as easier to talk to, is finally available for more people to try. The company has announced that everyone in the US and UK can test out the new version of its conversational bot, Claude 2, from its website.Its public availability allows Claude to join the ranks of ChatGPT, Bing, and Bard, all of which are available to users across numerous countries. That means we all have one more AI chatbot to play around with, but Anthropic says to “think of Claude as a friendly, enthusiastic colleague or personal assistant who can be instructed in natural language to help you with many tasks.”Jul 10Emma RothJust a couple of months ago, a leaked memo said to be from a Google researcher cast doubt on the company’s future in AI, stating that it has “no moat” in the industry — and now, we seemingly have confirmation that it was real. In an interview with Decoder, Demis Hassabis, the CEO of Google’s DeepMind, told The Verge that although he believes the memo was legitimate, he disagrees with its conclusions.“I think that memo was real. I think engineers at Google often write various documents, and sometimes they get leaked and go viral,” Hassabis said. “I think it’s interesting to listen to them, and then you’ve got to chart your own course. And I haven’t read that specific memo in detail, but I disagree with the conclusions from that.”Jul 10Nilay PatelToday, I’m talking to Demis Hassabis, the CEO of Google DeepMind, the newly created division of Google responsible for AI efforts across the company. Google DeepMind is the result of an internal merger: Google acquired Demis’ DeepMind startup in 2014 and ran it as a separate company inside its parent company, Alphabet, while Google itself had an AI team called Google Brain. Google has been showing off AI demos for years now, but with the explosion of ChatGPT and a renewed threat from Microsoft in search, Google and Alphabet CEO Sundar Pichai made the decision to bring DeepMind into Google itself earlier this year to create… Google DeepMind.Jul 9Wes DavisComedian and author Sarah Silverman, as well as authors Christopher Golden and Richard Kadrey — are suing OpenAI and Meta each in a US District Court over dual claims of copyright infringement. The suits alleges, among other things, that OpenAI’s ChatGPT and Meta’s LLaMA were trained on illegally-acquired datasets containing their works, which they say were acquired from “shadow library” websites like Bibliotik, Library Genesis, Z-Library, and others, noting the books are “available in bulk via torrent systems.”Jul 8Wes DavisGoogle’s Med-PaLM 2, an AI tool designed to answer questions about medical information, has been in testing at the Mayo Clinic research hospital, among others, since April, The Wall Street Journal reported this morning. Med-PaLM 2 is a variant of PaLM 2, which was announced at Google I/O in May this year. PaLM 2 is the language model underpinning Google’s Bard.WSJ reports that an internal email it saw said Google believes its updated model can be particularly helpful in countries with “more limited access to doctors.” Med-PaLM 2 was trained on a curated set of medical expert demonstrations, which Google believes will make it better at healthcare conversations than generalized chatbots like Bard, Bing, and ChatGPT. Jul 5Richard LawlerHowever, as James Vincent noted previously:According to OpenAI’s critics, this talk of regulating superintelligence, otherwise known as artificial general intelligence, or AGI, is a rhetorical feint — a way for Altman to pull attention away from the current harms of AI systems and keep lawmakers and the public distracted with sci-fi scenarios.Also today, CNBC reported SensorTower data shows ChatGPT and Bing app installs dropped 38 percent in June, while SimilarWeb data shows worldwide traffic to its website dropped 9.7 percent in June, along with a similar decline in minutes spent on the site. Is it because kids are out of school and don’t need the bot to do their homework, or has the AI trend already found a peak for now?Jul 5Elizabeth LopattoPersonally, I think it would be more effective for the government to deal with the proliferation of spam calls that have made most of us stop picking up our phones, but this is fun too:Whitebeard stalls for time at the start of phone calls, using chatbot inanities about TV remotes and the like to give a couple of minutes for GPT-4, the OpenAI software, to process the telemarketer’s spiel and generate responses. Once ready, the AI text is fed into a voice cloner, which carries on the conversation. [WSJ]Jul 5Jess WeatherbedOn Monday, Gizmodo spotted that Google updated its privacy policy to disclose that its various AI services, such as Bard and Cloud AI, may be trained on public data that the company has scraped from the web.“Our privacy policy has long been transparent that Google uses publicly available information from the open web to train language models for services like Google Translate,” said Google spokesperson Christa Muldoon to The Verge. “This latest update simply clarifies that newer services like Bard are also included. We incorporate privacy principles and safeguards into the development of our AI technologies, in line with our AI Principles.”Jul 4Wes DavisApparently users could get around paywalls by using ChatGPT’s Bing integration, which the company just introduced in May (via Windows Central). Today, OpenAI announced it was disabling the beta integration so it could fix the issue:We have learned that the ChatGPT Browse beta can occasionally display content in ways we don’t want. For example, if a user specifically asks for a URL’s full text, it might inadvertently fulfill this request.As of July 3, 2023, we’ve disabled the Browse with Bing beta feature... while we fix this in order to do right by content owners. We are working to bring the beta back as quickly as possible...Oh, Bing. You’re always a good time.[help.openai.com]Jul 1Wes Davis and Richard LawlerElon Musk continues to blame Twitter’s new limitations on AI companies scraping “vast amounts of data” as he announced new “temporary” limits on how many posts people can read. Now unverified accounts will only be able to see 600 posts per day, and for “new” unverified accounts, just 300 in a day. The limits for verified accounts (presumably whether they’re bought as a part of the Twitter Blue subscription, granted through an organization, or verification Elon forced on people like Stephen King, LeBron James, and anyone else with more than a million followers) still allow reading only a maximum of 6,000 posts per day.Jun 29Emma Rothwhat 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.”[ai.googleblog.com]Jun 29Emma RothMicrosoft 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.Jun 28Justine CalmaPeople 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. / Sign up for Verge Deals to get deals on products we've tested sent to your inbox daily.The Verge is a vox media network© 2023 Vox Media, LLC. All Rights Reserved