Group items tagged

"A source told Reuters that OpenAI has tested a model internally that achieved a 90 percent score on a challenging test of AI math skills, though it again couldn't confirm if this was related to project Strawberry. But another two sources reported seeing demos from the Q* project that involved models solving math and science questions that would be beyond today's leading commercial AIs."

"In one activity, my students drafted a paragraph in class, fed their work to ChatGPT with a revision prompt, and then compared the output with their original writing. However, these types of comparative analyses failed because most of my students were not developed enough as writers to analyze the subtleties of meaning or evaluate style. "It makes my writing look fancy," one PhD student protested when I pointed to weaknesses in AI-revised text. My students also relied heavily on AI-powered paraphrasing tools such as Quillbot. Paraphrasing well, like drafting original research, is a process of deepening understanding. Recent high-profile examples of "duplicative language" are a reminder that paraphrasing is hard work. It is not surprising, then, that many students are tempted by AI-powered paraphrasing tools. These technologies, however, often result in inconsistent writing style, do not always help students avoid plagiarism, and allow the writer to gloss over understanding. Online paraphrasing tools are useful only when students have already developed a deep knowledge of the craft of writing."

"Writing in the journal Science, Summerfield and colleagues from Google DeepMind report how they built the "Habermas Machine" - an AI system named after the German philosopher Jürgen Habermas. The system works by taking written views of individuals within a group and using them to generate a set of group statements designed to be acceptable to all. Group members can then rate these statements, a process that not only trains the system but allows the statement with the greatest endorsement to be selected."

"TikTok quantified the precise amount of viewing it takes for someone to form a habit: 260 videos. Kentucky authorities note that while it might seem a lot, TikTok videos can be just a few seconds long. "Thus, in under 35 minutes, an average user is likely to become addicted to the platform," the state investigators concluded."

"In addition to their usual job of grubbing out bad human edits, they're having to spend an increasing amount of their time trying to weed out AI filler. 404 Media has talked to Ilyas Lebleu, an editor at the crowdsourced encyclopedia who was involved in founding the "WikiProject AI Cleanup" project. The group is trying to come up with best practices to detect machine-generated contributions. (And no, before you ask, AI is useless for this.)"

"The climate crisis could prove AI's greatest challenge. While Google publicises AI-driven advances in flooding, wildfire and heatwave forecasts, like many big tech companies, it uses more energy than many countries. Today's large models are a major culprit. It can take 10 gigawatt-hours of power to train a single large language model like OpenAI's ChatGPT, enough to supply 1,000 US homes for a year."

"Sara needed some chocolate - she had had one of those days - so wandered into a Home Bargains store. "Within less than a minute, I'm approached by a store worker who comes up to me and says, 'You're a thief, you need to leave the store'." Sara - who wants to remain anonymous - was wrongly accused after being flagged by a facial-recognition system called Facewatch."

"Police officers are watching TikTok in an attempt to catch far-right demonstrators livestreaming self-incriminating footage of their illegal behaviour. TikTok's Live function has become one of the defining outlets for coverage of this summer's riots, with hundreds of thousands of viewers watching live streams of rioting over the last week in cities such as Stoke, Leeds, Hull and Nottingham."

"Pliny the Prompter is known for his ability to disrupt the world's most robust artificial intelligence models within approximately thirty minutes. This pseudonymous hacker has managed to manipulate Meta's Llama 3 into sharing instructions on creating napalm and even caused Elon Musk's Grok to praise Adolf Hitler. One of his own modified versions of OpenAI's latest GPT-4o model, named "Godmode GPT," was banned by the startup after it started providing advice on illegal activities."

"AI Outperforms Humans in Theory of Mind Tests Large language models convincingly mimic the understanding of mental states"

"ChatGPT can help with reframing thoughts and situations, similar to cognitive behavioural therapy - but "some clients can start to use it as a substitute for therapy", Masterson says. "I've had clients telling me they've already processed on their own, because of what they've read - it's incredibly dangerous." She has had to ask some clients to cease their self-experiments while in treatment with her. "It's about you and me in the room," she says. "You just cannot have that with text - let alone a conglomeration of lots of other people's texts." Self-directed chatbot therapy also risks being counterproductive, shrinking the area of inquiry. "It's quite affirmative; I challenge clients," says Masterson. ChatGPT could actually cement patterns as it draws, over and again, from the same database: "The more you try to refine it, the more refined the message becomes.""

"Her intervention emerged after Sir Paul McCartney became the latest star to back calls for laws to stop mass copyright theft by generative AI companies, warning the technology "could just take over". Bush, who shot to fame with Wuthering Heights in 1978 but whose last album was released in 2011, gave a rare interview this year in which she said she was "very keen" to make a new album, saying: "I've got lots of ideas … it's been a long time.""

"A recent incident at a company has led to the dismissal of half its HR team after a manager discovered a significant flaw in the applicant tracking system (ATS) used for hiring. This system, intended to improve the recruitment process, was automatically rejecting all job candidates, including the manager's own application."

"A complex problem that took microbiologists a decade to get to the bottom of has been solved in just two days by a new artificial intelligence (AI) tool. Professor José R Penadés and his team at Imperial College London had spent years working out and proving why some superbugs are immune to antibiotics. He gave "co-scientist" - a tool made by Google - a short prompt asking it about the core problem he had been investigating and it reached the same conclusion in 48 hours."

"Some might scoff at this, and point out that news organisations make their own mistakes all the time - more consequential than my physicist/physician howler, if less humiliating. But cases of bad journalism are almost always warped representations of the real world, rather than missives from an imaginary one. Crucially, if an outlet gets big things wrong a lot, its reputation will suffer, and its audience are likely to vote with their feet, or other people will publish stories that air the mistake. And all of it will be out in the open. You may also note that journalists are increasingly likely to use AI in the production of stories - and there is no doubt that it is a phenomenally powerful tool, allowing investigative reporters to find patterns in vast financial datasets that reveal corruption, or analyse satellite imagery for evidence of bombing attacks in areas designated safe for civilians. There is a legitimate debate over the extent of disclosure required in such cases: on the one hand, if the inputs and outputs are being properly vetted, it might be a bit like flagging the use of Excel; on the other, AI is still new enough that readers may expect you to err on the side of caution. Still, the fundamental difference is not in what you're telling your audience, but what degree of supervision you're exercising over the machine."

"There is also a lot of research that both third-party fact-checking and Community Notes can be really effective at reducing misperceptions. But - and this is a significant caveat - neither works well as a complete solution for lies on social media. When Twitter was working on Birdwatch, they claimed it would "not replace other labels and fact checks Twitter currently uses". But as I've written about before, Musk scaled back Twitter's Trust and Safety team significantly and positioned Community Notes as the replacement. As Yoel Roth, Twitter's former head of Trust and Safety, told WIRED, "The intention of Birdwatch was always to be a complement to, rather than a replacement for, Twitter's other misinformation methods." In fact, research on various attempts to mitigate COVID misinformation found that a layered, "Swiss cheese" approach might work best, where some efforts work well sometimes, but collectively the system catches most falsehoods."

""All I kept coming back to was, what would Chris say?" Wales said. As AI spreads across society and enters the courtroom, the US judicial conference advisory committee has announced that it will begin seeking public comment as part of determining how to regulate the use of AI-generated evidence at trial."

"Brands such as fintech company Ramp, jobs search site Indeed and Pernod Ricard-owned Scottish whisky maker Chivas Brothers have adopted the software. They are hoping to reach millions of users who regularly use generative AI products as a new method to search for information online - a shift that poses a long-term threat for Google's main business. "This is about much more than just getting your website indexed in their results. This is about recognising large language models as the ultimate influencer," said Jack Smyth, partner at marketing technology group Brandtech, which has created its own interface for brands. "

"I believe the answer has a lot to do with the fact that few things created by humans with artsy aspirations are worth anything in isolation. Their worth comes from what surrounds them, from the human context that birthed them and the human lives that were impacted by them. That's how masterpieces are judged as such."

"'Profiting from misery': how TikTok makes money from child begging livestreams Exploitation fears as people in extreme poverty perform stunts and beg for virtual gifts"