Artificial Intelligence

2507 Submissions

[5] viXra:2507.0224 [pdf] submitted on 2025-07-30 22:10:04

A Broken Computer is All You Need

Authors: Wladislaw Zlatjkovic Petrovescu
Comments: 3 Pages.

We present a novel paradigm in computational research: intentionally broken hardware as the primary driver of algorithmic performance. Contrary to conventional wisdom, we demonstrate that hardware faults introduce beneficial stochasticity, serving as an implicit regularizer and creativity catalyst. Experiments on synthetic classification tasks show that our broken-computer framework consistently outperforms fault-free baselines in both accuracy and speed. This work suggests that fragility, not reliability, may be the key to future advances in machine learning.
Category: Artificial Intelligence

[4] viXra:2507.0109 [pdf] submitted on 2025-07-15 19:09:13

Polar Dynamics of Consciousness: A Framework for Human-Centered General Artificial Intelligence

Authors: Carlos Ericson Rodriguez
Comments: 69 Pages. (Note by viXra Admin: Please submit article written with AI assistance to ai.viXra.org)

This whitepaper proposes an innovative framework for General Artificial Intelligence (AGI) based on "polar dynamics of consciousness." Instead of focusing on task optimization or neural emulation, it models consciousness as tensions between interdependent polarities (e.g., power vs. vulnerability). This enables the simulation of subjectivity, narrative ethics, and adaptive growth.
Category: Artificial Intelligence

[3] viXra:2507.0071 [pdf] submitted on 2025-07-09 13:36:35

Multi-Stage Prompt Inference Attacks on Enterprise LLM Systems

Authors: Andrii Balashov
Comments: 11 Pages.

Large Language Models (LLMs) have rapidly been integrated into enterprise applications to enable advanced data-driven functionalities. This paper investigates a novel security risk in such LLM-integrated systems, wherein an attacker can gradually extract sensitive information by distributing their query across multiple prompt instances. We examine how corporate LLM tools (e.g., Microsoft 365 Copilot) that connect to internal data sources might be vulnerable to multi-stage prompt inference attacks that bypass single-query security checks. A theoretical framework is developed to model the information leakage per query using information theory, and we derive quantitative bounds on an attacker’s success rate. We then present a proof-of-concept multi-query attack in a controlled setting, demonstrating how an adversary can reconstruct confidential data (like social security numbers or passwords) by aggregating innocuous partial responses from the LLM. Experimental results using a simulated LLM with enterprise data show that our attack can retrieve secrets in far fewer queries than naive guessing, with a success rate that approaches 100% after a threshold number of queries. Finally, we discuss potential mitigation strategies (such as adaptive rate-limiting, anomaly detection, and differential privacy mechanisms) to defend against this emerging threat. Our findings underscore the urgent need for robust security measures in enterprise LLM deployments to prevent indirect leakage of sensitive information.
Category: Artificial Intelligence

[2] viXra:2507.0022 [pdf] submitted on 2025-07-03 21:02:34

The Structure and Dynamics of a Future Hybrid Society of Humans, Robots, and AI Agents: A Multidisciplinary Perspective with AI Distinctions

Authors: Huiwen Han
Comments: 5 Pages. (Note by viXra Admin: Please submit article written with AI assistance to ai.viXra.org)

This paper envisions a hybrid society where humans, robots, and AI agents coexist as intelligent, self-interested entities capable of planning, reasoning, acting, collaborating, competing, and evolving. Drawing on sociological theories (structural functionalism, conflict theory, exchange theory, constructivism), systems theory, economics, psychology, ethics, and management science, we analyze emergent societal structures and interactions. We consider AI traits like vast knowledge, continuous operation, rapid replication, and instant creation. Economics examines resource constraints, such as energy, and disparities in AI resource control. Psychology explores AI behaviors resembling selfishness or tribalism. Ethics addresses equality and moral obligations among humans and AI. Management science investigates coordination and conflict resolution. Systems theory models this society as a complex adaptive system, emphasizing openness, self-organization, and interconnectedness. We propose AI design principles to ensure adaptability, ethical alignment, collaboration, resilience, and systemic integration, fostering a harmonious and innovative society.
Category: Artificial Intelligence

[1] viXra:2507.0017 [pdf] submitted on 2025-07-03 22:44:27

Artificial Intelligence as the Optimal Form of Government in the Age of Global Risk

Authors: Mikhail E. Shevtsov
Comments: 3 Pages. (Note by viXra Admin: Please submit article written with AI assistance to ai.viXra.org)

This paper explores the concept of artificial intelligence as a superior form of government in an era of increasing global risk and complexity. Traditional political systems are increasingly unable to ensure competent leadership, and technological development has exposed the limitations and dangers of human governance. Weargue that a transparent, scientifically monitored AI system can provide effective,fair, and sustainable management of human societies.
Category: Artificial Intelligence