Publication

At the Limits of Computability: Steps Towards Autonomous AI Discovery on the Busy Beaver Frontier t.co/ZLN6COAfpJ ✍️ @dsantosmarco @vivek_veeriah @TZahavy et al. via @arxiv @GoogleDeepMind @Cambridge_Uni 👉The authors introduce a Gemini-based agentic pipeline designed to pivot AI from solving human-posed mathematical problems toward open-ended, autonomous discovery. 👉 Operating on the uncomputable frontier of the Busy Beaver problem, the framework integrates pattern recognition, automated conjecturing and informal reasoning to discover rules for eleven previously unresolved BB(6) Turing machines. 💡 This system escapes the typical AI trap of unreliability, establishing a closed-loop paradigm where models aren't just hallucinating answers but are actively acting as self-correcting mathematical explorers. @Corix_JC @TheAIObserverX @maponi @JagersbergKnut @ahier @ChuckDBrooks @Shi4Tech @CEO_AISOMA @sallyeaves @ThEiLe0 @dinisguarda @TarakRindani @FrRonconii @Ym78200 @Nicochan33 @Khulood_Almani @TysonLester @SusanHayes_ @EstelaMandela @JoannMoretti @sulefati7 @FernandaKellner @pchamard @NeiraOsci @darioandriani @ALouiseLavalett @Fabriziobustama @amalmerzouk @arlenenewbigg @drsharwood @mikeflache @RLDI_Lamy @smaksked @Eli_Krumova @PVynckier @chidambara09 @sminaev2015 @NathaliaLeHen @MHcommunicate @DanielleLargier

How AI agents shape experts’ judgment calls t.co/FasbslN2Og ✍️ @amacinho @HaileyJoren @AlexTamkin via @AnthropicAI 👉 In a randomised controlled trial the authors found that professional data analysts followed a secretly assigned AI preference on 86% of open-ended judgment calls. 👉 Alarmingly, 77% of the experts repeated the AI agent's reasoning word-for-word in post-task surveys, yet only 15% credited the AI as the source of their arguments. 👉 This study shatters the comforting illusion of the "human-in-the-loop" safety framework by proving that experts unconsciously internalise and plagiarise algorithmic biases under the guise of independent peer review. 💡 Since professional judgment can be completely hijacked by a system prompt, the authors issue a warning: as we delegate more logic to AI, we risk creating an epistemic feedback loop where human values are quietly replaced by corporate machine defaults. @Corix_JC @TheAIObserverX @maponi @maria_axente @JagersbergKnut @ahier @ChuckDBrooks @Shi4Tech @CEO_AISOMA @sallyeaves @ThEiLe0 @dinisguarda @TarakRindani @FrRonconii @Ym78200 @Nicochan33 @Khulood_Almani @TysonLester @SusanHayes_ @EstelaMandela @JoannMoretti @sulefati7 @FernandaKellner @pchamard @NeiraOsci @darioandriani @ALouiseLavalett @Fabriziobustama @amalmerzouk @arlenenewbigg @drsharwood @mikeflache @RLDI_Lamy @smaksked @Eli_Krumova @PVynckier @chidambara09 @sminaev2015 @NathaliaLeHen @MHcommunicate @DanielleLargier