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.
Google apologizes for ‘missing the mark’ after Gemini generated racially diverse Nazis
Screenshot: Adi Robertson / The Verge
Google has apologized for what it describes as “inaccuracies in some historical image generation depictions” with its Gemini AI tool, saying its attempts at creating a “wide range” of results missed the mark. The statement follows criticism that it depicted specific white figures (like the US Founding Fathers) or groups like Nazi-era German soldiers as people of color, possibly as an overcorrection to long-standing racial bias problems in AI.
“We’re aware that Gemini is offering inaccuracies in some historical image generation depictions,” says the Google statement, posted this afternoon on X. “We’re working to improve these kinds of depictions immediately. Gemini’s AI image generation does generate a wide range of people. And that’s generally a good thing because people around the world use it. But it’s missing the mark here.”
Read Article>
One month with Microsoft’s AI vision of the future: Copilot Pro
The Verge
Microsoft’s Copilot Pro launched last month as a $20 monthly subscription that provides access to AI-powered features inside some Office apps, alongside priority access to the latest OpenAI models and improved image generation.
I’ve been testing Copilot Pro over the past month to see if it’s worth the $20 subscription for my daily needs and just how good or bad the AI image and text generation is across Office apps like Word, Excel, and PowerPoint. Some of the Copilot Pro features are a little disappointing right now, whereas others are truly useful improvements that I’m not sure I want to live without.
Read Article>
Sora can create video collages, too.
One of OpenAI’s employees showed off another of the company’s new text-to-video generator’s abilities.
This is some impressive AI creation of course, but what in blue blazes is happening in the upper right frame here?
OpenAI can’t register ‘GPT’ as a trademark — yet
Image: OpenAI
The US Patent and Trademark Office (PTO) has denied OpenAI’s application to register the word GPT, which means generative pre-trained transformer, saying GPT is too general a term to register and can prevent competitors from correctly describing their products as a GPT.
OpenAI argued in its application that GPT is not a descriptive word — that GPT isn’t such a general term that consumers would “immediately understand” what it means.
Read Article>
At least in Canada, companies are responsible when their customer service chatbots lie to their customer.
A man was booking an Air Canada flight and asked for a reduced rate because of bereavement. The chatbot assured him this was possible — the reduced fare would be a rebate. When he went to submit the rebate, the airline refused to refund him.
In February of 2023, Moffatt sent the airline a screenshot of his conversation with the chatbot and received a response in which Air Canada “admitted the chatbot had provided ‘misleading words.’”
He took the airline to court and won.
Scientists are extremely concerned about this rat’s “dck.”
And for good reason — this, and several other nonsensical AI-generated images were openly credited to Midjourney in a peer-reviewed science paper published by the Frontiers Journal this week. The gibberish annotations and grotesquely inaccurate images it included are one example of the risks that generative AI poses to the accuracy of academic research.
Frontiers has responded and removed the offending paper:
Our investigation revealed that one of the reviewers raised valid concerns about the figures and requested author revisions. The authors failed to respond to these requests. We are investigating how our processes failed to act on the lack of author compliance with the reviewers’ requirements.
Sora’s AI-generated video looks cool, but it’s still bad with hands.
OpenAI’s still-in-limited-testing new text-to-video generation model, Sora, is very impressive, especially compared to widely available AI video generators like Runway Gen-2 and Google’s Imagen.
As you can see in the clips, though, there are issues — basketballs go through the sides of metal hoops, dogs pass through each other while walking, and hands are…. not always hands.
You sound like a bot
Illustration by Erik Carter
In 2018, a viral joke started going around the internet: scripts based on “making a bot watch 1,000 hours” of just about anything. The premise (concocted by comedian Keaton Patti) was that you could train an artificial intelligence model on vast quantities of Saw films, Hallmark specials, or Olive Garden commercials and get back a bizarre funhouse-mirror version with lines like “lasagna wings with extra Italy” or “her mouth is full of secret soup.” The scripts almost certainly weren’t actually written by a bot, but the joke conveyed a common cultural understanding: AI was weird.
Strange AI was everywhere a few years ago. AI Dungeon, a text adventure game genuinely powered by OpenAI’s GPT-2 and GPT-3, touted its ability to produce deeply imagined stories about the inner life of a chair. The first well-known AI art tools, like Google’s computer vision program Deep Dream, produced unabashedly bizarre Giger-esque nightmares. Perhaps the archetypal example was Janelle Shane’s blog AI Weirdness, where Shane trained models to create physically impossible nuclear waste warnings or sublimely inedible recipes. “Made by a bot” was shorthand for a kind of free-associative, nonsensical surrealism — both because of the models’ technical limitations and because they were more curiosities than commercial products. Lots of people had seen what “a bot” (actually or supposedly) produced. Fewer had used one. Even fewer had to worry about them in day-to-day life.
Read Article>
How much electricity does AI consume?
Illustration by Erik Carter
It’s common knowledge that machine learning consumes a lot of energy. All those AI models powering email summaries, regicidal chatbots, and videos of Homer Simpson singing nu-metal are racking up a hefty server bill measured in megawatts per hour. But no one, it seems — not even the companies behind the tech — can say exactly what the cost is.
Estimates do exist, but experts say those figures are partial and contingent, offering only a glimpse of AI’s total energy usage. This is because machine learning models are incredibly variable, able to be configured in ways that dramatically alter their power consumption. Moreover, the organizations best placed to produce a bill — companies like Meta, Microsoft, and OpenAI — simply aren’t sharing the relevant information. (Judy Priest, CTO for cloud operations and innovations at Microsoft said in an e-mail that the company is currently “investing in developing methodologies to quantify the energy use and carbon impact of AI while working on ways to make large systems more efficient, in both training and application.” OpenAI and Meta did not respond to requests for comment.)
Read Article>
In defense of busywork
Illustration by Erik Carter
In the show Severance’s dystopian workplace — is there any other kind? — employees spend their days studying arrays of numbers bobbing on their screens. Whenever a cluster of numbers makes an employee feel unsettled, the employee clicks on it to discard it. The work’s value is not apparent to the workers, who are told only that they are “refining macro-data files,” but the job is nevertheless satisfying to complete. When one protagonist, Helly, tosses enough bad numbers, she is greeted with a Game Boy-esque animation of the company’s founder and CEO, who tells her, “I love you.”
The task is a parody of corporate busywork, the time-consuming, mind-numbing, manager-placating chores that fill our days. Most jobs involve some degree of busywork, and it is generally maligned. A Microsoft WorkLab survey published last January reported that 85 percent of respondents said they hoped artificial intelligence tools would automate all busywork, freeing up their time for more fulfilling activities such as “engaging with others.” These respondents have clearly never sat through a five-hour conversation about a three-word headline, but I digress: busywork has been cast as the enemy of innovation, and AI has been cast as the solution. “Eliminating busywork” has become AI proponents’ “making the world a better place.” But would it?
Read Article>
How AI can make history
Illustration by Erik Carter
Like millions of other people, the first thing Mark Humphries did with ChatGPT when it was released in late 2022 was ask it to perform parlor tricks, like writing poetry in the style of Bob Dylan — which, while very impressive, did not seem particularly useful to him, a historian studying the 18th-century fur trade. But Humphries, a 43-year-old professor at Wilfrid Laurier University in Waterloo, Canada, had long been interested in applying artificial intelligence to his work. He was already using a specialized text recognition tool designed to transcribe antiquated scripts and typefaces, though it made frequent errors that took time to correct. Curious, he pasted the tool’s garbled interpretation of a handwritten French letter into ChatGPT. AI corrected the text, fixing all the Fs that had been misread as an S and even adding missing accents. Then Humphries asked ChatGPT to translate it to English. It did that, too. Maybe, he thought, this thing would be useful after all.
For Humphries, AI tools held a tantalizing promise. Over the last decade, millions of documents in archives and libraries have been scanned and digitized — Humphries was involved in one such effort himself — but because their wide variety of formats, fonts, and vocabulary rendered them impenetrable to automated search, working with them required stupendous amounts of manual research. For a previous project, Humphries pieced together biographies for several hundred shellshocked World War I soldiers from assorted medical records, war diaries, newspapers, personnel files, and other ephemera. It had taken years and a team of research assistants to read, tag, and cross-reference the material for each individual. If new language models were as powerful as they seemed, he thought, it might be possible to simply upload all this material and ask the model to extract all the documents related to every soldier diagnosed with shell shock.
Read Article>
Stability AI tries to stay ahead of the pack with a new image-generating AI model
Stability AI
Stability AI’s newest model for image generation is Stable Cascade promises to be faster and more powerful than its industry-leading predecessor, Stable Diffusion, which is the basis of many other text-to-image generation AI tools.
Stable Cascade can generate photos and give variations of the exact image it created, or try to increase an existing picture’s resolution. Other text-to-image editing features include inpainting and outpainting, where the model will fill edit only a specific part of the image, as well as canny edge, where users can make a new photo just by using the edges of an existing picture.
Read Article>
Is OpenAI the next challenger trying to take on Google Search?
A source for The Information says OpenAI is working on web search (partially powered by Bing) that would more directly compete with Google. It’s unclear if it would be standalone, or a part of ChatGPT.
This comes one year after Microsoft CEO (and OpenAI backer) Satya Nadella targeted Google by adding Copilot AI tools to Bing, saying on Decoder, “I want people to know that we made them dance.”
Between Google’s Bard / Gemini, Copilot, and newcomers like Perplexity, the dance floor is filling up quickly.
Gemini Advanced is most impressive when it’s working with Google
The Verge
Chatbots occupy a tricky space for users — they have to be a search engine, a creation tool, and an assistant all at once. That’s especially true for a chatbot coming from Google, which is increasingly counting on AI to supplement its search engine, its voice assistant, and just about every productivity tool in its arsenal.
Right now, the ultimate version of Google’s AI is Gemini Advanced, which launched last week for users willing to pay $20 per month for the privilege — the same price OpenAI charges for its upgraded ChatGPT Plus. So I plunked down $20 and decided to see how Gemini Advanced stood up to the rival service.
Read Article>
Google Gemini can hang onto your chats for up to three years.
Google recently updated its privacy policy for Gemini — the chatbot it previously called Bard. It details just how long the company will keep conversations that are “reviewed or annotated” by human reviewers, despite whether you’ve deleted your app activity or not:
Conversations that have been reviewed or annotated by human reviewers (and related data like your language, device type, location info, or feedback) are not deleted when you delete your Gemini Apps activity because they are kept separately and are not connected to your Google Account. Instead, they are retained for up to three years.
To compare, ChatGPT lets you permanently delete conversations every 30 days.
OpenAI’s Dall-E sent a “shock wave of panic” through Adobe.
That’s according to a new Bloomberg report, detailing how Adobe concentrated its efforts to build Firefly, the company’s own “commercially safe” generative AI model used in tools like Photoshop, following the success of rival tools like Midjourney.
Analysts now anticipate that Adobe may be one of the first big tech companies to actually profit from AI. Meanwhile, Adobe Stock contributors who helped train Firefly, potentially unknowingly, receive annual payouts that are as low as $70.
ChatGPT is getting ‘memory’ to remember who you are and what you like
Illustration: The Verge
Talking to an AI chatbot can feel a bit like Groundhog Day after a while, as you tell it for the umpteenth time how you like your emails formatted and which of those “fun things to do this weekend” you’ve already done six times. OpenAI is trying to fix that and personalize its own bot in a big way. It’s rolling out “memory” for ChatGPT, which will allow the bot to remember information about you and your conversations over time.
Memory works in one of two ways. You can tell ChatGPT to remember something specific about you: you always write code in Javascript, your boss’s name is Anna, your kid is allergic to sweet potatoes. Or ChatGPT can simply try to pick up those details over time, storing information about you as you ask questions and get answers. In either case, the goal is for ChatGPT to feel a little more personal and a little smarter, without needing to be reminded every time.
Read Article>
The return of the (robot) travel agent
Illustration by Erik Carter
I was recently sitting in a hot tub with a friend — a glaciologist who studies how quickly the southern ice is sliding into the sea — when she mentioned that she had recently planned her honeymoon using ChatGPT. Our fellow bathers burst into laughter. “You’d done it, too?” This, apparently, is the present state of get-togethers among friends in their early 30s: six people and three AI-assisted honeymoons between them.
My friend is a pro at arranging helicopters and snow cat brigades to remote wedges of ice. But she was overloaded with decisions about charger plates and floral arrangements, and had put the task of arranging a 12-day trip to Tasmania on her husband-to-be. A statistician, he was using ChatGPT to answer questions at work, following the advice of a mentor who told him he’d better make a habit of it. So he asked it for an itinerary that would emphasize the couple’s love of nature, adventure, and (it being a honeymoon) luxury. A specific request: time for at least one lengthy trail run.
Read Article>
When a death is clickbait
Illustration by Erik Carter
In late December 2023, several of Brian Vastag and Beth Mazur’s friends were devastated to learn that the couple had suddenly died. Vastag and Mazur had dedicated their lives to advocating for disabled people and writing about chronic illness. As the obituaries surfaced on Google, members of their community began to dial each other up to share the terrible news, even reaching people on vacations halfway around the world.
Except Brian Vastag was very much alive, unaware of the fake obituaries that had leapt to the top of Google Search results. Beth Mazur had in fact passed away on December 21st, 2023. But the spammy articles that now filled the web claimed that Vastag himself had died that day, too.
Read Article>
Using generative AI to declare political victory.
The AI-generated stand-in voice for imprisoned Former Pakistan Prime Minister Imran Khan claimed victory on behalf of his party in Pakistan’s parliamentary elections on Saturday, according to The New York Times.
The party has used an AI version of his voice this way for months. As the Times writes, the use highlights both the usefulness and the danger of generative AI in elections.
Google’s Gemini for Android doesn’t require “Hey Google” to auto-send queries anymore.
Now, the chatbot formerly known as Bard will respond to your queries when you stop talking, regardless of how you summoned it. Before, that only worked when you invoked Google’s chatbot with the phrase “Hey Google.”
Microsoft’s Copilot AI can explain stuff to you in Notepad.
The rumors are true, even Notepad is getting a generative AI boost. A new update called “Explain with Copilot” will help users decipher any text, code segments, or log files they select within the text editor as Microsoft’s AI add-on enters its second year.
Microsoft announced the feature is in beta testing, available to Windows Insiders in the Canary and Dev Channels.
A screenshot of Notepad’s new “Explain with Copilot” feature.
Image: Microsoft
>>> Read full article>>>
Copyright for syndicated content belongs to the linked Source : The Verge – https://www.theverge.com/23610427/chatbots-chatgpt-new-bing-google-bard-conversational-ai