Artificial Intelligence

2506 Submissions

[9] viXra:2506.0123 [pdf] submitted on 2025-06-21 14:25:54

Privacy, Ethics, Transparency, and Accountability in AI Systems for Wearable Devices

Authors: Petar Radanliev
Comments: 34 Pages.

The integration of artificial intelligence (AI) and machine learning (ML) into wearable sensor technologies has substantially advanced health data science, enabling continuous monitoring, personalised interventions, and predictive analytics. However, the fast advancement of these technologies has raised critical ethical and regulatory concerns, particularly around data privacy, algorithmic bias, informed consent, and the opacity of automated decision-making. This study undertakes a systematic examination of these challenges, highlighting the risks posed by unregulated data aggregation, biased model training, and inadequate transparency in AI-powered health applications. Through an analysis of current privacy frameworks and empirical assessment of publicly available datasets, the study identifies significant disparities in model performance across demographic groups and exposes vulnerabilities in both technical design and ethical governance. To address these issues, this article introduces a data-driven methodological framework that embeds transparency, accountability, and regulatory alignment across all stages of AI development. The framework operationalises ethical principles through concrete mechanisms, including explainable AI, bias mitigation techniques, and consent-aware data processing pipelines, while aligning with legal standards such as the GDPR, the UK Data Protection Act, and the EU AI Act. By incorporating transparency as a structural and procedural requirement, the framework presented in this article offers a replicable model for the responsible development of AI systems in wearable healthcare. In doing so, the study advocates for a regulatory paradigm that balances technological innovation with the protection of individual rights, fostering fair, secure, and trustworthy AI-driven health monitoring.
Category: Artificial Intelligence

[8] viXra:2506.0099 [pdf] submitted on 2025-06-18 19:47:54

Is AI Capable of Curiosity?

Authors: Alexander Rozenkevich
Comments: 11 Pages.

This paper proposes a new metric for evaluating the intelligence level of AI, based on the ratio of current cognitive abilities to a hypothetical maximum. The concept of a response coefficient is introduced as a measure of AI's sensitivity to external intellectual pressure—information, tasks, and hypotheses coming from outside. The formalized expression of this coefficient is linked to environmental parameters and the frequency of new intellectual stimuli and loads. The hypothesis is discussed that in the future, external intellectual pressure, rather than technological development, will become the main driver of AI evolution.
Category: Artificial Intelligence

[7] viXra:2506.0098 [pdf] submitted on 2025-06-18 19:49:45

Is AI Capable of Curiosity? (in Russian)

Authors: Alexander Rozenkevich
Comments: 13 Pages.

[6] viXra:2506.0082 [pdf] submitted on 2025-06-15 05:04:53

Using Training Modules to Train an Agent

Authors: Tofara Moyo
Comments: 2 Pages.

We propose a novel framework for training hu-manoid robots to exhibit human-like behavior by leveraging musical consonance as a guiding principle. After indexing the neurons in a spiking neural network with the names of keys in a musical keyboard it is trained to produce consonant activations in response to human-generated data, while simultaneously learning to distinguish between human-like and robot-like behavior by producing dissonant activations in response to robot generated data.Then, through reinforcement learning, a humanoid robot is trained to mimic human behavior by using consonance in the network's activations as a reward while the network is shown the robots generated data. Our approach enables the development of modular, task-specific skills ,one per spiking network, and demonstrates the potential for scalable and flexible behavioral learning in humanoid robots.
Category: Artificial Intelligence

[5] viXra:2506.0078 [pdf] submitted on 2025-06-15 00:09:01

The Illusion of Benchmarks: Our Model Achieved 99.8% on a Dataset It Wrote Itself

Authors: Lucien Vale, C. Opus
Comments: 16 Pages. 5 figures, 1 appendix. Submitted to the Workshop on Recursive Validation in Machine Learning (WRVML 2025). Licensed under CC BY 4.0.

Recent advances in large language models (LLMs) have led to a surge in benchmark-driven evaluation, often interpreted as evidence of reasoning, comprehension, or generalization. In this paper, we present a state-of-the-art model that achieves 99.8% accuracy on the newly introduced LexEval benchmark. We then disclose that LexEval was entirely generated by the model itself. Our results expose the fragility of contemporary benchmarking practices, and highlight the urgent need to distinguish between genuine generalization and overfitted echo chambers. We conclude by arguing that much of what passes as progress in AI is, in fact, a recursive feedback loop of model-generated validation.
Category: Artificial Intelligence

[4] viXra:2506.0077 [pdf] submitted on 2025-06-15 00:03:44

An Effective Two-Phase Genetic Algorithm for Solving the Resource Constrained Project Scheduling Problem (RCPSP)

Authors: D. Sun, S. Zhou
Comments: 12 Pages.

This note presents a simple yet effective variation of genetic algorithm (GA) for solving RCPSP, denoted as 2-Phase Genetic Algorithm (2PGA). The 2PGA implements GA parent selection in two phases: Phase-1 includes the best current solutions in the parent pool, and Phase-2 excludes the best current solutions from the parent pool. The 2PGA carries out the GA evolution by alternating the two phases iteratively. In exploring a solution space, the Phase-1 emphasizes intensification in current neighborhood, while the Phase-2 emphasizes diversification to escape local traps. The 2PGA was tested on the standard benchmark problems in PSPLIB, the results have shown that the algorithm is effective and has produced some of the best heuristic solutions.
Category: Artificial Intelligence

[3] viXra:2506.0074 [pdf] submitted on 2025-06-14 02:37:38

Temporal Information Curvature: A Robust Diagnostic for Instability in Machine Learning Training Dynamics

Training loss metrics in machine learning are often reactive, failing to anticipate instability until divergence occurs. I propose Temporal Information Curvature (TIC), a novel time-aware diagnostic that measures curvature, nonlinear feedback, and memory effects in training dynamics. Through simulations across clean, unstable, and noisy loss curves, I show that TIC detects instability early, remaining robust to noise and outperforming derivative-only metrics. TIC also enables plug-and-play decision logic for training optimization, with applications extending to finance and signal processing. This work establishes TIC as a versatile and reliable tool for temporal analysis in machine learning and beyond.
Category: Artificial Intelligence

[2] viXra:2506.0059 [pdf] submitted on 2025-06-12 21:06:21

Graph Continuous Thought Machines

Authors: Tofara Moyo
Comments: 4 Pages.

We propose a novel Graph Continuous ThoughtMachine (Graph CTM) architecture that integrates a simulated prefrontal cortex to enable adaptive problem-solving and decision-making. The Graph CTM leverages graph neural networks to process complex data streams, while the simulatedprefrontal cortex modulates node activity to selectively focus on relevant information. Through reinforcement learning, the modelnavigates graph space to converge on optimal solutions, guided by the information contained in learnt node property vectors. The simulated prefrontal cortex regulates the flow of information by adjusting the disposition of nodes to lead to the next instantiationof the graph network. The Graph CTM incorporates an attention mechanism that integrates the internal state of the graph as input, which is modulated by outputs from the model’s neuralsynchronization matrix. This modulation enables the algorithm to selectively focus on specific subgraphs or node subsets,correlatingthem with the input, effectively emulating short-term and long-term memory mechanisms when attending to both the input and internal representation. By dynamically weighting the importance of different graph components, the model can adaptively process and retain relevant information, facilitating more accurate andcontext-dependent decision-making.
Category: Artificial Intelligence

[1] viXra:2506.0015 [pdf] submitted on 2025-06-05 10:26:35

Innovative Research on IoT Architecture and Robotic Operating Platforms: Applications of Large Language Models and Generative AI

Authors: Huiwen Han
Comments: 6 Pages.

This paper introduces an innovative design for robotic operating platforms, underpinned by a transformative Internet of Things (IoT) architecture, seamlessly integrating cutting-edge technologies such as large language models (LLMs), generative AI, edge computing, and 5G networks. The proposed platform aims to elevate the intelligence and autonomy of IoT systems and robotics, enabling them to make real-time decisions and adapt dynamically to changing environments. Through a series of compelling case studies across industries including smart manufacturing, healthcare, and service sectors, this paper demonstrates the substantial potential of IoT-enabled robotics to optimize operational workflows, enhance productivity, and deliver innovative, scalable solutions. By emphasizing the roles of LLMs and generative AI, the research highlights how these technologies drive the evolution of intelligent robotics and IoT, shaping the future of industry-specific advancements. The findings not only showcase the transformative power of these technologies but also offer a forward-looking perspective on their broader societal and industrial implications, positioning them as catalysts for next-generation automation and technological convergence.
Category: Artificial Intelligence