Group items tagged

"In fact, generating an image using a powerful AI model takes as much energy as fully charging your smartphone, according to a new study by researchers at the AI startup Hugging Face and Carnegie Mellon University. However, they found that using an AI model to generate text is significantly less energy-intensive. Creating text 1,000 times only uses as much energy as 16% of a full smartphone charge. "

"ow, legal redress is all very well, but it's usually beyond the resources of working artists. And lawsuits are almost always retrospective, after the damage has been done. It's sometimes better, as in rugby, to "get your retaliation in first". Which is why the most interesting news of the week was that a team of researchers at the University of Chicago have developed a tool to enable artists to fight back against permissionless appropriation of their work by corporations. Appropriately, it's called Nightshade and it "lets artists add invisible changes to the pixels in their art before they upload it online so that if it's scraped into an AI training set, it can cause the resulting model to break in chaotic and unpredictable ways" - dogs become cats, cars become cows, and who knows what else? (Boris Johnson becoming piglet, with added grease perhaps?) It's a new kind of magic. And the good news is that corporations might find it black. Or even deadly."

"NBC has come under fire over the weekend for allegedly altering a photo taken of controversial University of Pennsylvania swimmer Lia Thomas, a biological male who identifies as a woman, at the conclusion of a race. The photo appeared on a TODAY Show segment last week talking about Thomas ahead of the NCAA championships, which took place over the weekend."

"But ChatGPT and similar programs are, by design, unlimited in what they can "learn" (which is to say, memorize); they are incapable of distinguishing the possible from the impossible. Unlike humans, for example, who are endowed with a universal grammar that limits the languages we can learn to those with a certain kind of almost mathematical elegance, these programs learn humanly possible and humanly impossible languages with equal facility. Whereas humans are limited in the kinds of explanations we can rationally conjecture, machine learning systems can learn both that the earth is flat and that the earth is round. They trade merely in probabilities that change over time."

"A problem arises, though, when that content misleads us. When a purported "symptom" of anxiety is, actually, just a universal, everyday experience. When the information is flawed, or the people providing it are ill-informed. When viewers, many of whom are children and teens, don't realize that a TikTok diagnosis cannot replace treatment by a professional."

"The combination of ChatGPT with its Wolfram plug-in just scored 96% in a UK Maths A-level paper, the exam taken at the end of school, as a crucial metric for university entrance. (That compares to 43% for ChatGPT alone). If this doesn't shock you, it should. Maths A-level (like its equivalent in many other countries) is held up as the required and essential qualification for much of our populations-the way to be prepared for our upcoming AI age. And yet, here it is, done by those very AIs, better than most of our students."

"The question of why AI generates fake academic papers relates to how large language models work: they are probabilistic, in that they map the probability over sequences of words. As Dr David Smerdon of the University of Queensland puts it: "Given the start of a sentence, it will try to guess the most likely words to come next.""

"But for many scientists, Twitter has become an essential tool for collaboration and discovery - a source of real-time conversations around research papers, conference talks and wider topics in academia. Papers now zip around scientific communities faster thanks to Twitter, says Johann Unger, a linguist at Lancaster University, UK, who notes that extra information is also shared in direct private messages through the site. And its limit on tweet length - currently 280 characters - has pushed academics into keeping their commentary pithy, he adds."

"The music industry is urging streaming platforms not to let artificial intelligence use copyrighted songs for training, in the latest of a run of arguments over intellectual property that threaten to derail the generative AI sector's explosive growth. In a letter to streamers including Spotify and Apple Music, the record label Universal Music Group expressed fears that AI labs would scrape millions of tracks to use as training data for their models and copycat versions of pop stars."

"Researchers at the Citizen Lab at the University of Toronto's Munk School said the spyware, which is made by an Israeli company called QuaDream, infected some victims' phones by sending an iCloud calendar invitation to mobile users from operators of the spyware, who are likely to be government clients. Victims were not notified of the calendar invitations because they were sent for events logged in the past, making them invisible to the targets of the hacking. Such attacks are known as "zero-click" because users of the mobile phone do not have to click on any malicious link or take any action in order to be infected."

""Bots view, 'like,' subscribe and repost content and manipulate view counts to move content up in search results and recommendation lists," the analysis said. In some cases, Fabrika targets users with disinformation directly after gleaning their emails and phone numbers from databases. The campaign's goals include demoralising Ukrainians and exploiting divisions among western states, the document added. Experts have downplayed the 1% claim. Alan Woodward, a professor of cybersecurity at Surrey University, said the figure sounded implausible and that sock puppet accounts - a term for accounts with fake identities - need their content to be reposted by plausible accounts such as those operated by influencers."

"If you were doubting how important recommender systems are to social media companies, a lawsuit filed last week against Meta makes it crystal clear. At the heart of this legal battle is a fundamental question: Shouldn't users have the power to decide what they do and don't see online? The lawsuit filed by Knight First Amendment Institute at Columbia University on behalf of Professor Ethan Zuckerman directly challenges how social media feeds are curated. Professor Zuckerman's proposed browser extension, 'Unfollow Everything 2.0,' would enable Facebook users to disengage from the algorithmically driven content that dominates their feeds, by allowing them to unfollow friends, pages and groups en masse, thus resetting their digital interactions on their terms."

"A pilot study by academics at the University of Oxford found some care providers had been using generative AI chatbots such as ChatGPT and Bard to create care plans for people receiving care. That presents a potential risk to patient confidentiality, according to Dr Caroline Green, an early career research fellow at the Institute for Ethics in AI at Oxford, who surveyed care organisations for the study. "If you put any type of personal data into [a generative AI chatbot], that data is used to train the language model," Green said. "That personal data could be generated and revealed to somebody else." She said carers might act on faulty or biased information and inadvertently cause harm, and an AI-generated care plan might be substandard."

"Other far more sinister real-world effects of algorithms are well documented. In the US, pedestrians have been mowed down by robotaxis; prisoners denied bail on the advice, in part, of software; in Australia, welfare recipients incorrectly and illegally hounded by an algorithmic debt collector that came to be known as robodebt. In the UK, students took to the streets in 2020 after being denied places at universities by the calculations of digital minions - their chants of "fuck the algorithm" proving a "defining moment" for Kowalkiewicz and an inspiration for his book."

"The research has found many children do not even recognise these promotions, known as content marketing, as advertising. It warns that this may lead to children following betting companies on social media, making it more likely that they sign up with them when they turn 18 and can legally gamble. Dr Raffaello Rossi, lecturer in marketing at Bristol University, one of the report's authors, said content marketing was particularly popular with young people."

"To find more glyphs, researchers led by archaeologist Masato Sakai of Yamagata University trained an AI program to identify relief-type glyphs in high-resolution drone images taken of the entire region. The program identified 1309 possible geoglyphs, and the team confirmed 303 of them with on-the-ground surveys, almost doubling the number of known geoglyphs of this type, the researchers report today in the Proceedings of the National Academy of Sciences."

"Researchers at the University of Pennsylvania found that Turkish high school students who had access to ChatGPT while doing practice math problems did worse on a math test compared with students who didn't have access to ChatGPT. Those with ChatGPT solved 48 percent more of the practice problems correctly, but they ultimately scored 17 percent worse on a test of the topic that the students were learning. "

"But what is pitched as a more convenient way of looking up information online has prompted scrutiny over how and where these chatbots select the information they provide. Looking into the sort of evidence that large language models (LLMs, the engines on which chatbots are built) find most convincing, three computer science researchers from the University of California, Berkeley, found current chatbots overrely on the superficial relevance of information. They tend to prioritise text that includes pertinent technical language or is stuffed with related keywords, while ignoring other features we would usually use to assess trustworthiness, such as the inclusion of scientific references or objective language free of personal bias."

"When "I can has cheezburger?" became one of the first internet memes to blow our minds, it's unlikely that anyone worried about how much energy it would use up. But research has now found that the vast majority of data stored in the cloud is "dark data", meaning it is used once then never visited again. That means that all the memes and jokes and films that we love to share with friends and family - from "All your base are belong to us", through Ryan Gosling saying "Hey Girl", to Tim Walz with a piglet - are out there somewhere, sitting in a datacentre, using up energy. By 2030, the National Grid anticipates that datacentres will account for just under 6% of the UK's total electricity consumption, so tackling junk data is an important part of tackling the climate crisis. Ian Hodgkinson, a professor of strategy at Loughborough University has been studying the climate impact of dark data and how it can be reduced."

""An emotionally intelligent human does not usually claim they can accurately put a label on everything everyone says and tell you this person is currently feeling 80% angry, 18% fearful, and 2% sad," says Edward B Kang, an assistant professor at New York University writing about the intersection of AI and sound. "In fact, that sounds to me like the opposite of what an emotionally intelligent person would say." Adding to this is the notorious problem of AI bias. "Your algorithms are only as good as the training material," Barrett says. "And if your training material is biased in some way, then you are enshrining that bias in code.""