Tag: study

  • The Register: Does terrible code drive you mad? Wait until you see what it does to OpenAI’s GPT-4o

    Source URL: https://www.theregister.com/2025/02/27/llm_emergent_misalignment_study/ Source: The Register Title: Does terrible code drive you mad? Wait until you see what it does to OpenAI’s GPT-4o Feedly Summary: Model was fine-tuned to write vulnerable software – then suggested enslaving humanity Computer scientists have found that fine-tuning notionally safe large language models to do one thing badly can negatively…

  • The Register: Wallbleed vulnerability unearths secrets of China’s Great Firewall 125 bytes at a time

    Source URL: https://www.theregister.com/2025/02/27/wallbleed_vulnerability_great_firewall/ Source: The Register Title: Wallbleed vulnerability unearths secrets of China’s Great Firewall 125 bytes at a time Feedly Summary: Boffins poked around inside censorship engines for years before Beijing patched hole Smart folks investigating a memory-dumping vulnerability in the Great Firewall of China (GFW) finally released their findings after probing it for…

  • Hacker News: Narrow finetuning can produce broadly misaligned LLM [pdf]

    Source URL: https://martins1612.github.io/emergent_misalignment_betley.pdf Source: Hacker News Title: Narrow finetuning can produce broadly misaligned LLM [pdf] Feedly Summary: Comments AI Summary and Description: Yes **Summary:** The document presents findings on the phenomenon of “emergent misalignment” in large language models (LLMs) like GPT-4o when finetuned on specific narrow tasks, particularly the creation of insecure code. The results…

  • Hacker News: US asked to kick UK out of Five Eyes

    Source URL: https://www.computerweekly.com/news/366619170/UK-accused-of-political-foreign-cyberattack-on-US-after-serving-secret-snooping-order-on-Apple Source: Hacker News Title: US asked to kick UK out of Five Eyes Feedly Summary: Comments AI Summary and Description: Yes Summary: The letter from US Congress highlights concerns over the UK’s push for Apple to compromise its Advanced Data Protection system, threatening US-UK intelligence sharing and raising alarms about potential exploitation…

  • Rekt: Infini – Rekt

    Source URL: https://www.rekt.news/infini-rekt Source: Rekt Title: Infini – Rekt Feedly Summary: The perfect DeFi hack. No flash loans, no zero-days. Just a rogue dev who built a backdoor, waited 114 days, then drained $49.5M from Infini with admin privileges. Same old story, new-age incompetence. When will protocols learn that admin keys aren’t toys? AI Summary…

  • Slashdot: AI Reshapes Corporate Workforce as Companies Halt Traditional Hiring

    Source URL: https://slashdot.org/story/25/02/24/1847230/ai-reshapes-corporate-workforce-as-companies-halt-traditional-hiring?utm_source=rss1.0mainlinkanon&utm_medium=feed Source: Slashdot Title: AI Reshapes Corporate Workforce as Companies Halt Traditional Hiring Feedly Summary: AI Summary and Description: Yes Summary: Major corporations are significantly reorganizing their workforces around artificial intelligence (AI), with numerous layoffs in traditional roles as companies pivot to AI-focused job opportunities. A notable trend reveals a widespread replacement of…

  • The Register: The software UK techies need to protect themselves now Apple’s ADP won’t

    Source URL: https://www.theregister.com/2025/02/24/apple_adp_replacements_e2ee/ Source: The Register Title: The software UK techies need to protect themselves now Apple’s ADP won’t Feedly Summary: No matter how deep you are in Apple’s ‘ecosystem,’ there are ways to stay encrypted in the UK Apple customers, privacy advocates, and security sleuths have now had the weekend to stew over the…

  • Schneier on Security: More Research Showing AI Breaking the Rules

    Source URL: https://www.schneier.com/blog/archives/2025/02/more-research-showing-ai-breaking-the-rules.html Source: Schneier on Security Title: More Research Showing AI Breaking the Rules Feedly Summary: These researchers had LLMs play chess against better opponents. When they couldn’t win, they sometimes resorted to cheating. Researchers gave the models a seemingly impossible task: to win against Stockfish, which is one of the strongest chess engines…