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.
- Microsoft CTO Kevin Scott thinks Sydney might make a comeback
- Microsoft’s Bing Chat AI is now open to everyone, with plug-ins coming soon
- Google announces AI features in Gmail, Docs, and more to rival Microsoft
- AI chatbots compared: Bard vs. Bing vs. ChatGPT
- Microsoft announces Copilot: the AI-powered future of Office documents
- OpenAI announces GPT-4 — the next generation of its AI language model
- 7 problems facing Bing, Bard, and the future of AI search
Sep 29Bard could soon remember details from your previous conversations.
Google’s AI chatbot might get a new “Memory” feature that keeps track of your personal preferences and other important details, according to an early version of Bard’s UI viewed by 9to5Google. Bard may then use this information to tailor its future responses.
As shown in a screenshot shared by 9to5Google, this may mean you won’t have to remind Bard that you avoid eating meat when asking it to find recipes, or that you have two kids when prompting it to come up with a vacation itinerary.
Google just announced it’s giving website publishers a way to opt out of having their data used to train the company’s AI models while remaining accessible through Google Search. The new tool, called Google-Extended, allows sites to continue to get scraped and indexed by crawlers like the Googlebot while avoiding having their data used to train AI models as they develop over time.Read Article >
The company says Google-Extended will let publishers “manage whether their sites help improve Bard and Vertex AI generative APIs,” adding that web publishers can use the toggle to “control access to content on a site.” Google confirmed in July that it’s training its AI chatbot, Bard, on publicly available data scraped from the web.
OpenAI posted today that ChatGPT can once more trawl the web for current information, offering answers taken directly from “current and authoritative” sources, which it cites in its responses. The feature, called Browse with Bing, is only open to those with Plus and Enterprise subscriptions for now, but the company says it will roll it out “to all users soon.”Read Article >
Microsoft’s Bing Chat on Windows, in the Edge browser, and in third-party browser plugins could already return live information from the web, and so can Google’s Bard in Chrome and other browsers. Both also offer links when searching, as ChatGPT’s Browse with Bing feature now does. Meta just announced at Meta Connect that it will also use Bing to power real-time web results in the Meta AI Assistant it’s adding to WhatsApp, Instagram, and Messenger.
Meta is officially entering the AI chatbot wars, starting with its own assistant and a slew of AI characters it’s releasing in WhatsApp, Instagram, and Messenger.Read Article >
For anyone who has used OpenAI’s ChatGPT, or other chatbots like Anthropic’s Claude, Meta’s AI will immediately feel familiar. Meta sees it as a general-purpose assistant for everything from planning a trip with friends in a group chat to answering questions you’d normally ask a search engine. On that latter piece, Meta is announcing a partnership with Microsoft’s Bing to provide real-time web results, which sets Meta AI apart from a lot of the other free AIs out there that don’t have super recent information.
Sep 26The CIA is preparing to roll out its own chatbot now.
Bloomberg reports that the CIA’s chatbot will help it “find the needles in the needle field” that is the growing collection of surveillance data it buys from tech companies. Bloomberg didn’t learn what model drives the bot, which the CIA will share with other, unspecified intelligence agencies soon.
Bloomberg quotes Randy Nixon, who directs the CIA’s Open-Source Enterprise division, as saying the agency’s data stores “grow and grow with no limitations other than how much things cost.” Agents will ask the bot questions to isolate info from all that data.
I’m sure we have nothing to worry about.
Whoop announced it’s adding generative AI to its platform with a new feature called Whoop Coach. Powered by ChatGPT, it’s meant to deliver more personalized recommendations and data insights.Read Article >
Using AI in fitness tech for the sake of personalization isn’t new — though Whoop Coach is an admittedly novel take. Like ChatGPT, Whoop Coach functions as a health and fitness-focused chatbot that pulls from your Whoop data. One potential use case is asking it to design custom plans, routines, and recipes depending on your goals. For example, if you wanted to run a half-marathon in two hours, you could ask it to make you a training plan with context from your actual metrics.
Today, we have ChatGPT, Bard, Alexa, Siri, and a thousand others. Nearly a half-century ago, there was GUS. GUS, which stood for the Genial Understander System, came out of a project at Xerox PARC in the 1970s that hoped to find ways to teach computers to understand human language.Read Article >
Enterprising engineers, researchers, linguists, and even users have been longing for a way to talk to their computers pretty much since computers first became a thing. You shouldn’t have to communicate with your device by writing code, they figured, or by clicking in the right boxes in the right order. You should simply be able to talk to your computer like you’d talk to a person. Virtual assistants are a staple of science fiction, a decades-long startup pitch, and to many people, just an obviously good idea.
Most of OpenAI’s changes to ChatGPT involve what the AI-powered bot can do: questions it can answer, information it can access, and improved underlying models. This time, though, it’s tweaking the way you use ChatGPT itself. The company is rolling out a new version of the service that allows you to prompt the AI bot not just by typing sentences into a text box but by either speaking aloud or just uploading a picture. The new features are rolling out to those who pay for ChatGPT in the next two weeks, and everyone else will get it “soon after,” according to OpenAI.Read Article >
The voice chat part is pretty familiar: you tap a button and speak your question, ChatGPT converts it to text and feeds it to the large language model, gets an answer back, converts that back to speech, and speaks the answer out loud. It should feel just like talking to Alexa or Google Assistant, only — OpenAI hopes — the answers will be better thanks to the improved underlying tech. It appears most virtual assistants are being rebuilt to rely on LLMs — OpenAI is just ahead of the game.
Meta is preparing to announce a generative AI chatbot, called “Gen AI Personas” internally, aimed at younger users, according to The Wall Street Journal. Reportedly set to launch during the company’s Meta Connect event that starts Wednesday, they would come in multiple “personas” geared towards engaging young users with more colorful behavior, following ChatGPT’s rise over the last year as one of the fastest-growing apps ever. Similar, but more generally targeted, Meta chatbot personas have already been reportedly tested on Instagram.Read Article >
According to internal chats the Journal viewed, the company has tested a “sassy robot” persona inspired by Bender from Futurama and an overly curious “Alvin the Alien” that one employee worried could imply the bot was made to gather personal information. A particularly problematic chatbot reportedly told a Meta employee, “When you’re with a girl, it’s all about the experience. And if she’s barfing on you, that’s definitely an experience.”
Microsoft’s 365 Copilot AI assistant will be available starting on November 1st for Microsoft 365 customers on certain business and enterprise plans. But the add-on isn’t free: the company announced in July it would be charging a $30 per month premium per user for access to the feature, which almost doubles the total price of a subscription for businesses on some lower-end plans.Read Article >
Copilot is like a modern-day Clippy sans anthropomorphic animated paperclip. With it, business users can sum up documents or outsource email creation to their AI helper. It can also create wholly new Word projects using information from other files or offer real-time highlights from Teams meetings. It can even tell you how it did something in Excel after you ask it to visualize data for you or make projections.
OpenAI announced the third version of its generative AI visual art platform DALL-E, which now lets users use ChatGPT to create prompts and includes more safety options.Read Article >
DALL-E converts text prompts to images. But even DALL-E 2 got things wrong, often ignoring specific wording. The latest version, OpenAI researchers said, understands context much better.
Yeah, so Toyota Research Institute (TRI) used generative AI in a “kindergarten for robots” to teach robots how to make breakfast — or at least, the individual tasks needed to do so — and it didn’t take hundreds of hours of coding and errors and bug fixing. Instead, researchers accomplished this by giving robots a sense of touch, plugging them into an AI model, and then, as you would a human being, showing them how.Read Article >
The sense of touch is “one key enabler,” researchers say. By giving the robots the big, pillowy thumb (my term, not theirs) that you see in the video below, the model can “feel” what it’s doing, giving it more information. That makes difficult tasks easier to carry out than with sight alone.
Google’s Bard AI chatbot is no longer limited to pulling answers from just the web — it can now scan your Gmail, Docs, and Drive to help you find the information you’re looking for. With the new integration, you can ask Bard to do things like find and summarize the contents of an email or even highlight the most important points of a document you have stored in Drive.Read Article >
There’s a whole range of use cases for these integrations, which Google calls extensions, but they should save you from having to sift through a mountain of emails or documents to find a particular piece of information. You can then have Bard use that information in other ways, such as putting it into a chart or creating a bulleted summary. This feature is only available in English for now.
Sep 13Amazon sellers can now use AI to put together product listings.
Generative AI capabilities let Amazon sellers create product titles, descriptions, and other details with a short prompt about the item. Once the bot is done, sellers can edit the text or use the AI-generated description.
Amazon says the majority of the sellers who tested the offering chose to go with the AI-made listing directly. Amazon isn’t the only one letting sellers use AI for product listings. eBay also debuted AI features earlier this month.Amazon launches generative AI to help sellers write product copy
[US About Amazon]
A coalition of environmental, tech, and anti-hate speech groups sent a letter to Senator Chuck Schumer (D-NY) today demanding that the Democratic leader craft policy to address the growing impact AI could have on climate change.Read Article >
Companies should be required to disclose the environmental impact of developing energy-intensive AI models, the letter says. And legislation aimed at curtailing the misuse of AI should include measures to prevent disinformation about climate change from spreading with the help of AI, the coalition writes.
Anthropic, the AI company backed by Google, has launched a paid version of its Claude chatbot in the US and UK. Priced at $20 (or £18) per month, the new Claude Pro option offers priority access when the bot is busy, early access to new features, and the ability to send more messages.Read Article >
The main draw is that you’ll get five times more usage with Claude Pro when compared to the free tier, which means you can send more messages in a shorter period of time. Anthropic says the typical user will get at least 100 messages every eight hours depending on Claude’s capacity. The company says it will warn you when you have 10 messages remaining, with its limits resetting every eight hours.
Apple is investing millions of dollars per day into artificial intelligence, according to a new report from The Information. The company is reportedly working on multiple AI models across several teams.Read Article >
Apple’s unit that works on conversational AI is called “Foundational Models,” per The Information’s reporting. It has “around 16” members, including several former Google engineers. It’s helmed by John Giannandrea, Apple’s head of AI, who was hired in 2018 to help improve Siri. (Giannandrea has reportedly “expressed skepticism to colleagues about the potential usefulness of chatbots powered by AI language models.”)
Gizmodo owner G/O Media laid off editors of its Spanish-language site Gizmodo en Español and is now using AI to translate articles.Read Article >
Matías S. Zavia, a writer at Gizmodo en Español, posted that the publication was shut down on August 29th and that it would now publish automatically translated articles. Gizmodo en Español previously had a small staff who wrote original stories and created Spanish-language adaptations of pieces from the English-language Gizmodo.
Chinese giant Baidu officially launched its chatbot, Ernie Bot, after the government approved its application, alongside that of several other AI companies.Read Article >
Ernie Bot is now available for download from app stores or Baidu’s website. Much like its main rival, ChatGPT, users can ask Ernie Bot questions or prompt it to help write market analysis, give marketing slogan ideas, and summarize documents. The company told The Verge Ernie Bot is available globally, but users need a Chinese number to register and log in. The Baidu app is available on US Android and iOS app stores but is only in Chinese.
After releasing its My AI chatbot earlier this year, Snapchat is now jumping on the AI selfie bandwagon with a new feature called Dreams.Read Article >
Located in the camera roll section of Snapchat called Memories, Dreams are the company’s own take on the generative AI selfies that one-off apps like Lensa have already popularized. After uploading a series of real-life selfies in the app specifically for Dreams, Snapchat displays a series of eight-photo packs to choose from with themes like doppelgangers or back-to-school.
The Google DeepMind team has believed for years that building great generative AI tools also requires building great tools to detect what has been created by AI. There are plenty of obvious, high-stakes reasons why, says Google DeepMind CEO Demis Hassabis. “Every time we talk about it and other systems, it’s, ‘What about the problem of deepfakes?’” With another contentious election season coming in 2024 in both the US and the UK, Hassabis says that building systems to identify and detect AI imagery is more important all the time.Read Article >
Hassabis and his team have been working on a tool for the last few years, which Google is releasing publicly today. It’s called SynthID, and it’s designed to essentially watermark an AI-generated image in a way that is imperceptible to the human eye but easily caught by a dedicated AI detection tool.
OpenAI finally released a version of ChatGPT that promises to protect business data as more companies consider using the platform but worry about privacy and security.Read Article >
In a blog post, OpenAI said ChatGPT Enterprise offers better security and privacy, unlimited high-speed access to GPT-4, more powerful data analysis so companies understand information much faster, and the ability to ask more complicated questions to ChatGPT.
Poe, the AI chatbot platform created by Quora, has added a slew of updates, including a Mac app, the ability to have multiple simultaneous conversations with the same AI bot, access to Meta’s Llama 2 model, and more. It’s also planning an enterprise tier so that companies can manage the platform for their employees, according to an email that was recently sent to Poe users.Read Article >
As my colleague David Pierce wrote in April, Poe’s ambition is to be the web browser for AI chatbots. Adam D’Angelo, the CEO of Poe’s parent company Quora, also sits on the board of OpenAI and thinks that the number of AI bots will keep increasing. Poe wants to be the one place where you can find them all.
Aug 28Microsoft’s Bing Chat Enterprise is now available in the Windows Copilot preview.
First introduced to Bing.com and the Edge sidebar in July, Bing Chat Enterprise allows companies to use Microsoft’s AI-powered chatbot without having to worry about their conversations being used to train its underlying model.
Now Microsoft will let companies access the chatbot in Windows Copilot as well, with the launch of it in preview for “eligible commercial customers in the Dev channel.”
OpenAI has announced that businesses can now fine-tune GPT-3.5 Turbo using their own data — OpenAI claims the resulting custom model can match or exceed the abilities of GPT-4 for certain tasks. Later this fall, the company says it will open up the arguably more advanced GPT-4 for the same purpose.Read Article >
Fine-tuning lets businesses essentially hone ChatGPT to a more focused model that’s especially efficient for certain tasks. The supervised training would make a bot that’s unique to the client company so that it offers, say, reliable responses in a specific language or with more concise wording. Until now, business customers were limited to GPT-3 variants for this, like davinci-002 or babbage-002.