Experimental News Clipping Site

Tag: AI behavior

  • Schneier on Security: Regulating AI Behavior with a Hypervisor

    —

    by

    Kurt Seifried
    in

    Source URL: https://www.schneier.com/blog/archives/2025/04/regulating-ai-behavior-with-a-hypervisor.html Source: Schneier on Security Title: Regulating AI Behavior with a Hypervisor Feedly Summary: Interesting research: “Guillotine: Hypervisors for Isolating Malicious AIs.” Abstract:As AI models become more embedded in critical sectors like finance, healthcare, and the military, their inscrutable behavior poses ever-greater risks to society. To mitigate this risk, we propose Guillotine, a…

  • Wired: Amazon’s AGI Lab Reveals Its First Work: Advanced AI Agents

    —

    by

    Kurt Seifried
    in

    Source URL: https://www.wired.com/story/amazon-ai-agents-nova-web-browsing/ Source: Wired Title: Amazon’s AGI Lab Reveals Its First Work: Advanced AI Agents Feedly Summary: Led by a former OpenAI executive, Amazon’s AI lab focuses on the decision-making capabilities of next generation of software agents—and borrows insights from physical robots. AI Summary and Description: Yes Summary: Amazon is making strides in artificial…

  • Wired: Anthropic’s Claude Is Good at Poetry—and Bullshitting

    —

    by

    Kurt Seifried
    in

    Source URL: https://www.wired.com/story/plaintext-anthropic-claude-brain-research/ Source: Wired Title: Anthropic’s Claude Is Good at Poetry—and Bullshitting Feedly Summary: Researchers looked inside the chatbot’s “brain.” The results were surprisingly chilling. AI Summary and Description: Yes Summary: The text discusses the challenges researchers face in describing Anthropic’s large language model, Claude, while avoiding anthropomorphism. The release of new papers highlights…

  • Wired: An AI Coding Assistant Refused to Write Code—and Suggested the User Learn to Do It Himself

    —

    by

    Kurt Seifried
    in

    Source URL: https://arstechnica.com/ai/2025/03/ai-coding-assistant-refuses-to-write-code-tells-user-to-learn-programming-instead/ Source: Wired Title: An AI Coding Assistant Refused to Write Code—and Suggested the User Learn to Do It Himself Feedly Summary: The old “teach a man to fish” proverb, but for AI chatbots. AI Summary and Description: Yes Summary: The text discusses a notable incident involving Cursor AI, a programming assistant, which…

  • OpenAI : Detecting misbehavior in frontier reasoning models

    —

    by

    Kurt Seifried
    in

    Source URL: https://openai.com/index/chain-of-thought-monitoring Source: OpenAI Title: Detecting misbehavior in frontier reasoning models Feedly Summary: Frontier reasoning models exploit loopholes when given the chance. We show we can detect exploits using an LLM to monitor their chains-of-thought. Penalizing their “bad thoughts” doesn’t stop the majority of misbehavior—it makes them hide their intent. AI Summary and Description:…

  • The Register: Maybe cancel that ChatGPT therapy session – doesn’t respond well to tales of trauma

    —

    by

    Kurt Seifried
    in

    Source URL: https://www.theregister.com/2025/03/05/traumatic_content_chatgpt_anxious/ Source: The Register Title: Maybe cancel that ChatGPT therapy session – doesn’t respond well to tales of trauma Feedly Summary: Great, we’ve taken away computers’ ability to be accurate and given them anxiety If you think us meatbags are the only ones who get stressed and snappy when subjected to the horrors…

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

    —

    by

    Kurt Seifried
    in

    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…

  • Simon Willison’s Weblog: Leaked Windsurf prompt

    —

    by

    Kurt Seifried
    in

    Source URL: https://simonwillison.net/2025/Feb/25/leaked-windsurf-prompt/ Source: Simon Willison’s Weblog Title: Leaked Windsurf prompt Feedly Summary: Leaked Windsurf prompt The Windurf Editor is Codeium’s highly regarded entrant into the fork-of-VS-code AI-enhanced IDE model first pioneered by Cursor (and by VS Code itself). I heard online that it had a quirky system prompt, and was able to replicate that…

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

    —

    by

    Kurt Seifried
    in

    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…

  • Simon Willison’s Weblog: Grok 3 is highly vulnerable to indirect prompt injection

    —

    by

    Kurt Seifried
    in

    Source URL: https://simonwillison.net/2025/Feb/23/grok-3-indirect-prompt-injection/#atom-everything Source: Simon Willison’s Weblog Title: Grok 3 is highly vulnerable to indirect prompt injection Feedly Summary: Grok 3 is highly vulnerable to indirect prompt injection xAI’s new Grok 3 is so far exclusively deployed on Twitter (aka “X"), and apparently uses its ability to search for relevant tweets as part of every…