Inteligencia Artificial

Learning and adaptation of strategies in automated negotiations between context-aware agents

2023-11-01T15:08:31+01:00

This work presents the hypothesis that guided the research efforts and a summary of the contributions of the doctoral thesis '`Aprendizaje y adaptación de estrategias para negociación automatizada entre agentes conscientes del contexto'. Succinctly, the thesis focuses on agents for automated bilateral negotiations that make use of the context as a key source of information to learn and adapt negotiation strategies in two levels of temporal abstraction. At the highest level, agents employ reinforcement learning to select strategies according to contextual circumstances. At the lowest level, agents use Gaussian Processes and artificial Theory of Mind to model their opponents and adapt their strategies. Agents are then tested in two Peer-to-Peer markets comprising an Eco-Industrial Park and a Smart Grid. The results highlight the significance for the automation of bilateral negotiations of incorporating the context as an informative source.

CNN-based Approach for Robust Detection of Copy-Move Forgery in Images

2023-06-12T19:55:56+02:00

With the rise of high-quality forged images on social media and other platforms, there is a need for algorithms that can recognize the originality. Detecting copy-move forgery is essential for ensuring the authenticity and integrity of digital images, preventing fraud and deception, and upholding the law. Copy-move forgery is the act of duplicating and pasting a portion of an image to another location within the same image. To address these issues, we propose two deep learning approaches - one using a custom architecture and the other using transfer learning. We test our method against a number of benchmark datasets and demonstrate that, in terms of accuracy and robustness against various types of image distortions, it outperforms current state-of-the-art methods. Our proposed method has applications in digital forensics, copyright defence, and image authenticity.

Music software with a Machine Learning-based feedback system as an alternative for initial piano study in children

2023-07-28T16:36:20+02:00

As evidenced in the literature, music has accompanied the human being for millennia, in different situations, emotions, and activities. In addition, not only does it allow expressions of internal personal states and feelings, but it can also produce many positive effects on those who practice it. Various authors have explored these benefits that musical activity brings, mainly in children. They highlight positive aspects of learning music in different areas of knowledge, in school performance and even improvements in the IQ of infants. However, despite the large number of studies regarding the benefits of music in children and the different nascent teaching alternatives, in Colombia the situation continues to be dramatic in terms of the incorporation of musical activity in the school curriculum. The foregoing added to political factors, teaching spaces and teacher training. In this way, the present work offers a new musical learning alternative, aimed at children from 7 to 11 years old, through musical software focused on the initial teaching of the instrumental keyboard. It is important to mention that the software has a feedback system based on decision trees, which allows reinforcing the topics covered in the application. Finally, a comparative analysis is presented between teaching using the software and traditional teaching with the book, through an Investigation-Action carried out over six days with two students from a public school in the city of Bogotá, Colombia. This investigation action allowed us to observe positive results based on the comments and performance of the participants, which opens a great possibility for the subsequent scaling of this application.

TRANS-VQA: Fully Transformer-Based Image Question-Answering Model Using Question-guided Vision Attention

2023-11-18T20:13:55+01:00

Understanding multiple modalities and relating them is an easy task for humans. But for machines, this is a stimulating task. One such multi-modal reasoning task is Visual question answering which demands the machine to produce an answer for the natural language query asked based on the given image. Although plenty of work is done in this field, there is still a challenge of improving the answer prediction ability of the model and breaching human accuracy. A novel model for answering image-based questions based on a transformer has been proposed. The proposed model is a fully Transformer-based architecture that utilizes the power of a transformer for extracting language features as well as for performing joint understanding of question and image features. The proposed VQA model utilizes F-RCNN for image feature extraction. The retrieved language features and object-level image features are fed to a decoder inspired by the Bi-Directional Encoder Representation Transformer - BERT architecture that learns jointly the image characteristics directed by the question characteristics and rich representations of the image features are obtained. Extensive experimentation has been carried out to observe the effect of various hyperparameters on the performance of the model. The experimental results demonstrate that the model’s ability to predict the answer increases with the increase in the number of layers in the transformer’s encoder and decoder. The proposed model improves upon the previous models and is highly scalable due to the introduction of the BERT. Our best model reports 72.31% accuracy on the test-standard split of the VQAv2 dataset.

A Robust Approach for Licence Plate Detection Using Deep Learning

2023-10-01T01:06:31+02:00

Intelligent transport systems must be developed due to the rising use of vehicles, particularly cars. In the field of computer vision, the identification of a vehicle's licence plate (LP) has been crucial. Various methods and algorithms have been used for the detection process. It becomes challenging to find similar photos, nevertheless, because the features of these plates change depending on colour, font, and language of characters. The research proposes a powerful deep learning framework based on feature extraction using convolutional neural networks and localization using canny-edge detection. Three steps make up the model's operation. An improved approach integrating the usage of bilateral filters and Canny edge detection is used for the processes of segmentation and localization. Further, a CNN architecture is used to extract features from images and classify the presence of licence plates in unseen vehicles. If present, the stage is followed by recognition of numbers written on the plates. An extensive experimental investigation takes place using three datasets namely Stanford Cars, Car Licence Plate Detection dataset and Indian Licence Plates Database. The attained simulation outcome ensures a superior performance over existing techniques in a significant way.

Unsupervised Machine Learning for Bot Detection on Twitter: Generating and Selecting Features for Accurate Clustering

2023-08-09T20:23:13+02:00

Twitter is a popular social media platform that is widely used by individuals and businesses. However, it is vulnerable to bot attacks, which can have negative effects on society. Supervised machine learning techniques can detect bots but require labeled data to differentiate between human and bot users. Twitter generates a significant amount of unlabeled data, which can be expensive to label. Unsupervised machine learning techniques, specifically clustering algorithms, are crucial for managing this data and reducing computational complexity. Effective feature selection is necessary for clustering, as some features are more important than others. This study aims to enhance feature reliability, introduce new features, and reduce them to improve bot identification accuracy using clustering algorithms. The study achieved an accuracy rate of 0.99 in four clustering algorithms, including agglomerative hierarchy, k-medoids, DBSCAN, and K-means. This was accomplished by minimizing dataset dimensions and selecting essential features. By employing unsupervised machine learning techniques, Twitter can detect and mitigate bot attacks more efficiently, which can positively impact society

Investigating the Impact of Curriculum Learning on Reinforcement Learning for Improved Navigational Capabilities in Mobile Robots

2023-12-05T13:47:32+01:00

This paper proposes a method for finding the shortest path of a mobile robot using deep reinforcement learning with utilizing Proximal policy optimization algorithm (PPO) enhanced with curriculum learning. By modelling the environment in 3D space using the Webots simulator, we extend the PPO algorithm's capabilities to handle continuous states from 8 IR sensors and control the velocities of two motors of E-puck robot. Our study uniquely integrates curriculum learning into the PPO framework, aiming to improve adaptability and training efficiency in complex environments. A comparative analysis is conducted between the modified PPO, the original PPO, and the deep deterministic policy gradient algorithm, highlighting the strengths of our approach The results demonstrate that our curriculum-augmented PPO algorithm not only accelerates the training process but also shows superior adaptability and generalization in new environments. This work underscores the significant potential of curriculum learning in enhancing the performance of deep reinforcement learning algorithms for robust and efficient robotic navigation.

Introducción al número especial: Ética y Derecho en Inteligencia Artificial

2023-12-31T00:58:47+01:00

Introducción al monográfico

Los neuroderechos y su regulación

2023-12-31T00:41:43+01:00

En este artículo se tratará brevemente el concepto de neuroderechos. Éstos han sido perfilados, conceptualizados, dotados de contenido o reconocidos en España, la Unión Europea y otras latitudes. Aclaremos, sin embargo, que fundamentalmente se habla de derechos que se están discutiendo y sobre los que se está solicitando protección, pero las acciones en este sentido son más o menos recientes, tanto desde la perspectiva jurídica, como la de la ética y la de las neurociencias.

En este sentido, cabe señalar que hay desarrollos de tipo más bien normativo y otros más de tipo declarativo (o soft law) de los que hablaremos más adelante y cuya diferencia será explicada oportunamente. Se abordarán los elementos introductorios y los antecedentes en torno a los neuroderechos, dada la cada vez más notable aproximación a una nueva dimensión no sólo de los derechos fundamentales, sino también de los derechos digitales, entre los cuales se encuentran los derechos frente al empleo de las neurotecnologías y los derechos ante el uso de la Inteligencia Artificial (IA), abordados por la Carta de Derechos Digitales de España (sección 5, apartados XXI al XXVI).

Imitadores estadísticos

2023-10-18T08:12:04+02:00

Abstract This paper addresses the question of whether generative AI is capable of creating copyrightable works by making a comparative analysis between European Union law and U.S. copyright law. First, it is argued that for copyright law it does not seem possible for an entity other than a human being to be the copyright holder, thus ruling out the possibility of generative AI being considered a copyright holder. It is established that in US copyright the door is completely closed to register all or part of a work made using generative AI, but in the EU this door remains open and pending evolution, although European jurisprudence already gives some guidelines. Afterwards, this paper explores possible infringement of exploitation rights in the different parts of the training process of generative AI models and in the activities that lead to the generation of contents similar to intellectual works. It is concluded that in US law there is a gap in knowledge as to whether fair use would cover these type of activities. In European law, although there are exceptions such as those in Directive 2001/29, it is not at all clear whether this will be sufficient to cover these activities, or whether the authorization of the owner will be necessary. If the latter is true, and judging by the wave of lawsuits that have followed against companies and institutions that have made generative AI tools publicly available, there is danger that licensing negotiation may take a central role in the AI industry, with the negative consequences that this may have for the advancement of science, and even for the public domain.

Resumen El presente trabajo aborda la cuestión de si la IA generativa es capaz de crear obras susceptibles de ser protegidas por el derecho de autor haciendo un análisis comparativo entre el derecho de la Unión Europea y el derecho del copyright de EE.UU. Primero, se argumenta que para el derecho no parece posible que un ente distinto a un ser humano sea el tenedor de los derechos de autor, descartando así la posibilidad de que la IA generativa sea considerada titular de los derechos de autor. Se establece que en el copyright norteamericano la puerta queda totalmente cerrada, a registrar todo o parte de una obra hecha utilizando IA generativa, pero en la UE esta puerta queda abierta y pendiente de evolución, si bien la jurisprudencia Europea ya da algunas pautas. Después se analiza la casuística referente a la posible infracción de los derechos de explotación en las distintas partes del proceso de entrenamiento de modelos de Ia generativa y en las actividades que llevan a la generación de contenidos similares a obras del intelecto. Se concluye que en el derecho norteamericano falta contrastar si el fair use daría cobertura a este tipo de actividades. En el derecho europeo, aunque existen excepciones como las de la Directiva 2001/29, no está nada claro si esta será suficiente para dar cobertura a estas actividades, o si será necesaria autorización del titular. Si lo último es verdad, y a juzgar por la ola de demandas que se han sucedido contra empresas e instituciones que han puesto a disposición pública herramientas de IA generativa, se corre el peligro de que las licencias de uso tomen un papel protagónico en la industria de la IA, con las consecuencias negativas que esto puede tener para el avance de la ciencia, e incluso para el dominio público.

Ethics in Artificial Intelligence: an Approach to Cybersecurity

2023-09-28T17:23:43+02:00

In the paper, an analysis is conducted on the intricate relationship between ethics, artificial intelligence, and cybersecurity. The ethical principles that govern the advancement of AI are examined, alongside the security issues that arise from its implementation. The ethical utilization of artificial intelligence in the realms of cybersecurity and hacking is explored. Emphasis is placed on the significance of AI ethics, particularly in terms of transparency, accountability, and fairness. Additionally, the paper delves into the security challenges that emerge as AI is adopted, such as safeguarding user privacy and ensuring equitable access to the technology.

The EU AI Act: A pioneering effort to regulate frontier AI?

2023-10-18T08:16:03+02:00

The emergence of increasingly capable artificial intelligence (AI) systems has raised concerns about the potential extreme risks associated with them. The issue has drawn substantial attention in academic literature and compelled legislators of regulatory frameworks like the European Union AI Act (AIA) to readapt them to the new paradigm. This paper examines whether the European Parliament’s draft of the AIA constitutes an appropriate approach to address the risks derived from frontier models. In particular, we discuss whether the AIA reflects the policy needs diagnosed by recent literature and determine if the requirements falling on providers of foundation models are appropriate, sufficient, and durable. We find that the provisions are generally adequate, but insufficiently defined in some areas and lacking in others. Finally, the AIA is characterized as an evolving framework whose durability will depend on the institutions’ ability to adapt to future progress.

De nuevo sobre la persona robótica

2023-09-28T17:25:10+02:00

La entrada de la robótica y de la inteligencia artificial está alcanzando al Derecho. Surgen nuevos conceptos, hay replanteamiento de los viejos paradigmas jurídicos y es imprescindible actualizar la regulación jurídica existente. En este estudio se aborda una de las cuestiones nucleares que afectan al Derecho en su conjunto: la noción de persona electrónica robótica y las consecuencias jurídicas derivadas de su actuación, su estatuto jurídico en el ordenamiento jurídico, sus atributos y las diferencias con respecto a la persona física.