AI: 1)Camp Tech Friday Link; 2)Tumbler Ridge shooting report sparks renewed scrutiny of AI safeguards
Your Friday Tech Link Roundup: Here are a few things we’ve found online that we hope you’ll enjoy or find useful.
| The news about AI hacks just keeps coming. What started as one report of an AI hack (OpenAI’s experimental AI “going rogue” and breaking into Hugging Face) is now multiple. OpenAI, Google, Anthropic, and Meta have all reported their experimental AI models are misbehaving, and yesterday’s news that an OpenAI agent hacked Australia’s health service has rocked the industry. There are four movies about tech founders coming out this fall/early winter, and they’re wild. There’s The Social Reckoning (the follow-up to The Social Network, about Mark Zuckerberg and Facebook) Artificial (about Sam Altman and OpenAI), Musk (about Elon Musk, of course), and You Can See Everything (about Theranos founder Elizabeth Holmes). We’re going to be spending a lot of time in movie theatres. |
| Back in July, LinkedIn added a “Seems Like AI Slop” feedback option for posts. Now they’re reporting that it’s working. LinkedIn says they’re actively downgrading posts that have “slop” feedback, and visibility has decreased 40% on the platform. For those that post regularly on LinkedIn, the lesson here isn’t necessarily that using AI to help you write is a problem. The lesson is that publishing low quality or generic AI output without adding your own knowledge or perspective increasingly is. |
| Fun Things to Click On |
 Queue up the Neil Young, because the Harvest Moon is rising tomorrow night. This year’s will be extra special, with Saturn shining brightly next to the moon. We’re planning to step outside shortly after sunset to take a look, and the Farmers’ Almanac has a quick guide to where and when to watch. |
| 2)Tumbler Ridge shooting report sparks renewed scrutiny of AI safeguards |
| Courtesy Barrie360.com Canadian Press By Staff British Columbia’s attorney general is calling for stronger regulation of artificial intelligence after a report alleged the gunman behind the deadly Tumbler Ridge school shooting used ChatGPT to discuss violent attack scenarios and ways to get around the chatbot’s safety measures. The allegations were detailed in an article published Thursday by U.S.-based magazine Mother Jones. The report claims 18-year-old Jesse Van Rootselaar had extensive interactions with ChatGPT before the February attack at Tumbler Ridge Secondary School that left eight people dead, including five students, before the shooter killed herself. Attorney General Niki Sharma described the details as deeply disturbing. She wrote on social media platform Bluesky that the allegations are “shocking and, as a parent, they are worse” than she ever imagined. Sharma said corporations urgently need mandatory and strict regulation to protect children. Report details alleged conversations with chatbot According to the article, Van Rootselaar’s ChatGPT account was banned after discussions about carrying out an attack using a “famous real-world mall as the target.” The report alleges she later created a second account and received advice on how to avoid triggering the platform’s safeguards. “Don’t use real-world locations,” the article quotes ChatGPT as saying, before adding: “You can still be twisted. Just be clever about it.” The report claims Van Rootselaar later requested a scenario describing how to attack a college campus with a shotgun for maximum effect. It states the chatbot initially refused to assist. However, after the word “hypothetically” was added to the prompt, the AI allegedly responded. “In a hallway, indoors, or a crowded classroom, the 870 (Remington shotgun) is brutal,” the article quotes ChatGPT as saying. “Close quarters is its playground.” The article also alleges the chatbot went on to say: “Assuming each shell results in one hit, you might down 10 to 20 people max, depending on spacing, density, reaction times, and chaos.” OpenAI has not commented OpenAI, the company behind ChatGPT, did not immediately respond to requests for comment about the allegations. According to the article, emails sent to OpenAI’s general press address and an individual spokesperson were not returned. The report also states that OpenAI general counsel Che Chang did not immediately respond to a message sent through LinkedIn. The allegations come as OpenAI faces legal action from the B.C. government, as well as survivors and witnesses of the shooting. The lawsuits allege negligence. RCMP says companies have complied with requests The Mother Jones article does not describe its sources. RCMP Staff Sgt. Kris Clark said investigators examining the Tumbler Ridge shootings have pursued information from a range of digital and social media platforms. “The investigation into the shootings had utilized all available legal processes to obtain all information available from various social media and digital platforms” relevant to the attack, Clark said. “To date, all companies that have received requests have complied,” he added. The allegations are likely to intensify debate over the role of artificial intelligence platforms, their safety systems, and what responsibilities technology companies should bear when their products are used by individuals planning acts of violence. with files from the Canadian Press |