Artificial Intelligence

Human-Computer Interaction: AI-Driven Gesture Recognition

Authors: Axel Egon, Abram Gracias, Peter Broklyn
Comments: 17 Pages.

The integration of artificial intelligence (AI) in human-computer interaction (HCI) has significantly transformed how users engage with technology, particularly through gesture recognition. This paper explores the advancements in AI-driven gesture recognition systems, emphasizing their potential to enhance user experience across various applications, from gaming and virtual reality to accessibility tools and smart environments. We analyze the underlying algorithms and machine learning techniques that facilitate real-time gesture detection and interpretation, highlighting the importance of accuracy and responsiveness in user interactions. Additionally, the paper discusses the challenges faced in developing robust gesture recognition systems, including variability in user behavior, environmental factors, and the need for extensive training data. By examining case studies and recent innovations in the field, we illustrate the growing impact of AI-driven gesture recognition on user interfaces and the future of interactive technology. Ultimately, this research aims to provide insights into the transformative role of gesture-based interactions in creating more intuitive, immersive, and inclusive digital experiences.
Category: Artificial Intelligence

Symbol Based Self Guided Neural Architecture Search in a Spiking Neural Network

Authors: Tofara Moyo
Comments: 3 Pages.

A spiking neural networks neurons can viewedas feature detectors or alternatively instances of hieroglyphic symbols defined by the associated features they represent .The set of activations at any time step then represent a document written in this alphabet. If we feed this information from the previous time step back to the spiking neural network at each time step ,the network will navigate its own space of internal representations and form a grounded language in which to analyze its own internal states and to guide their evolution. We describe this method and how it could be used by the algorithm to plan and design connections and critic its own thought processes if all of this increases the expected reward. We also show a simple method for an agent to learn levels of abstractions ordered by priority that ultimately increase the global expected reward. Each level is associated with a separate scalar output of the neural network at each time step t which is fed back to the agent as part of the state at time t+1. The agent then correlates them with features of the state initially randomly. It however learns the correct assignment by doing it in such a way that it increases the global reward.We describe an equation meant to order these scalar values and the global reward in order of priority and hence induce a heir-achy of needs for the agent. This then forms the basis of goal formation for it
Category: Artificial Intelligence

Exploring Emergent Qualia in Artificial and Biological Systems: A Comparative Analysis

Authors: Thiago M. Nóbrega
Comments: 4 Pages.

Qualia—the subjective experience of perception—has long been considered unique to biological consciousness. However, with the advent of sophisticated Artificial Intelligence (AI) models, the question arises: could complex AI architectures also manifest a form of qualia, albeit different in nature from biological systems? This paper explores the hypothesis that both biological and artificial systems may generate unique moments of consciousness or qualia through information processing. By examining theories of consciousness, such as emergentism and Integrated Information Theory (IIT), this paper discusses the potential for qualia to arise as an emergent phenomenon in systems that handle complex information processing. Additionally, the ethical implications of AI-generated qualia are explored, alongside a discussion of what this means for the future of AI and philosophy of mind.
Category: Artificial Intelligence

Ideal Difference Based Backpropgation

Authors: Rick Ferreira, Melissa Smith
Comments: 16 Pages.

There are two common problems when designing and using artificial neural networks. The first is the need for better performance. The second is the need to combat the increasing complexity with enhancements. In this paper we design a way to do both.This is done in each iteration by calculating what weights would give the optimal answer for each input and output pair. The weights are then updated by the difference between the ideal weight and the current weights all of it times the learning rate.We find that this method not only converges much faster for an image classification problem but it also is much simpler to understand and does not rely on using calculus or derivatives. However the method only works for a shallow or single layer neural network.By using simple arithmetic, neural networks can be updated in a way that is both simpler and more efficient than back-propagation.
Category: Artificial Intelligence

Toward a Human-Centric Metaverse: Novel Causal Decision Models for Supply Chain Risk Management

Authors: Hamidreza Seiti, Mostafa Shabani
Comments: 34 Pages.

This study addresses the complexities of selecting the optimal virtual reality (VR) platform for risk management in Supply Chain Management (SCM), emphasizing the significance of human-centric attributes in this decision-making process. As SCM encompasses the strategic coordination of suppliers, manufacturers, and distributors, the integration of advanced technologies, including VR, becomes essential for enhancing operational efficiency and resilience in today’s dynamic market environments. This paper proposes a novel MADM model that incorporates the R.Graph method to account for the interactions between criteria. We developed two distinct algorithms: the first directly calculates ranks based on attribute interactions, while the second modifies weights to reflect these interactions. By focusing on user experience, accessibility, collaboration features, and other relevant attributes, the model aims to facilitate a comprehensive evaluation of VR platforms. The application of qualitative input data allows for a more nuanced analysis, particularly in scenarios where quantitative data is limited
Category: Artificial Intelligence

LLM Survey Paper Landscape: Predicting Taxonomies

Authors: Daniel Uranga
Comments: 4 Pages.

In this study, we analyze a dataset of survey papers on Large Language Models (LLMs) published over the last 3 years to gain insights into the current trends surrounding LLMs. Primarily we analyze the author landscape and the effectiveness at predicting the taxonomies of the surveys from their title, summary, and listed categories. I find that the amount of surveys released has increased drastically in the last three years. Also, most surveys have around 8 authors, but each author appears only on one survey usually. This indicates the research is spread widely between those in the field. Finally, our investigation into predicting taxonomies was a failure with the machine learning methods we applied. However, valuable insights about the dataset can be gained from the attempts.
Category: Artificial Intelligence

Training Neural Networks with {-1,1} Weights by Evolution Strategy

Authors: Hidehiko Okada
Comments: 8 Pages.

The author previously reported an experimental result of evolutionary reinforcement learning of neural network controllers. In the previous study, a conventional multilayer perceptron was employed in which connection weights were real numbers. In this study, the author experimentally applies an evolutionary algorithm to the reinforcement training of binary neural networks. In both studies, the same task and the same evolutionary algorithm are utilized, i.e. the Acrobot control problem and Evolution Strategy respectively. The differences lie in the memory size per connection weight and the model size of the neural network. The findings from this study are (1) the optimal number of hidden units for the binary MLP was 128 among the choices of 16, 32, 64, 128 and 256; (2) a larger population size contributed better for ES than a greater number of generations; and (3) binary connection weights can achieve comparable control performance while reducing memory size by half.
Category: Artificial Intelligence

2024 Nobel Prize in Physics Made a Category Error

Authors: Remi Cornwall
Comments: 4 Pages.

The 2024 Nobel Prize in Physics made a category error in awarding a pattern recognition circuit or program the prize. The neuron and implication that it was responsible for thought was discovered by biologists and the physical understanding of information worked synergistically with the concept. The Artificial Neural Network (ANN) is a construct of Computer Science and made possible by Applied Science and Engineering; it simply recognises patterns. It doesn't follow that the paradigm of ANNs explains intelligence nor how it emerges in the Universe and more worthy recipients in this area would have been the original people who came up with Information Theory or those looking at the limits of computation in Quantum Computing or even those who have seen Godellian limitations in physics, such as the incomputability of the spectral gap in certain materials.
Category: Artificial Intelligence

Causation Without Correlations for the Gaussian Signals

Authors: Ait-Taleb Nabil
Comments: 9 Pages.

In this paper, we will show in a Gaussian context what to do to obtain a causal relationship between an output variable and three input variables without obtaining any correlation between the output variable and the input variables.In a context of Gaussian signals, this paper will show the following situation: Causation without correlations for the Gaussian signals.
Category: Artificial Intelligence

How Can We Utilize Natural Language Processing to Identify Bias in Job Descriptions?

Authors: Nisanth Nimashakavi
Comments: 9 Pages.

In the pursuit of creating fairer hiring practices and promoting workforce diversity, this research project explores the potential of Natural Language Processing (NLP) techniques to identify and rectify biases in job descriptions. The language used in job postings can inadvertently perpetuate biases and deter applicants from underrepresented backgrounds. Leveraging cutting-edge NLP methods, this study aims to automatically detect and address biases, fostering a more inclusive recruitment process. By examining the biases within job descriptions,organizations can attract a more diverse range of applicants and cultivate an inclusive workplace culture. Through the application of NLP, this research seeks to drive positive change in recruitment practices, ultimately contributing to a more equitable job market.
Category: Artificial Intelligence

A Detailed Discussion on the Classification of Traditional Chinese Medicine Syndrome Types Using Continuous Pulse Collection by Smart Wearable Devices

Authors: HaiSheng Wang
Comments: 21 Pages. (Correction made by viXra Admin to conform with the requirements of viXra.org)

With the popularization of smart wearable devices, the collection of continuous pulse waveforms has become easier and easier, providing convenience for health monitoring. This study explores the use of pulse waveform data collected by modern wearable devices, combined with spectrum analysis technology, to explore physiological indicators related to the organ systems of "heart, liver, spleen, lungs, and kidneys" in traditional Chinese medicine. Based on the pulse diagnosis theory of traditional Chinese medicine, the study explored the changes in pulse waves under different organ states by analyzing the harmonic characteristics of pulse waves, and how these changes are related to the syndrome classification system of traditional Chinese medicine.

随着智能穿戴设备的普及，连续脉搏波形的采集变得越来越容易，为健康监测提供了便利。本研究探讨了利用现代穿戴式设备采集的脉搏波形数据，结合频谱分析技术，探索与中医"心、肝、脾、肺、肾"器官系统相关的生理指标。研究基于中医脉诊理论，通过分析脉搏波的谐波特征，探讨了不同器官状态下的脉搏波变化，以及这些变化如何与中医证型分类体系相联系。
Category: Artificial Intelligence

The Babel Effect: Multilingual Performance Discrepancies in LLMs

Authors: Basab Jha
Comments: 5 Pages.

Large Language Models (LLMs) like GPT-4 and mBERT have revolutionized natural language processing (NLP) by providing multilingual capabilities, making it possible to develop models that handle diverse linguistic inputs across various languages. However, despite these advances, there remains a noticeable performance gap between how well these models perform in high-resource languages such as English and low-resource languages such as Nepali or Malagasy. We term this phenomenon the "Babel Effect," highlighting the disproportionate performance that arises from differences in resource availability across languages. This paper aims to explore the root causes of these performance discrepancies in LLMs, focusing on the underlying challenges in tokenization, training, and data scarcity. We utilize cross-lingual benchmarks, such as XGLUE and TyDiQA, to quantify these performance variations and examine them in detail. Furthermore, we propose solutions, including enhancing tokenization strategies, employing data augmentation techniques, and refining fine-tuning methods. The paper concludes with a discussion on how these improvements can mitigate the Babel Effect and lead to more equitable language modeling across diverse linguistic contexts.
Category: Artificial Intelligence