ABSTRACT. This tutorial delves into the methodologies of time series research, tailored for artificial intelligence applications. Participants will embark on a comprehensive journey beginning with data preprocessing—structuring datasets to meet AI requirements. We will explore classical statistical processing using R, setting the groundwork for advanced analysis. The tutorial then transitions to sophisticated deep learning models, focusing on the detection of hidden events and the prediction of variables within time series data. By blending traditional techniques with cutting-edge AI, attendees will gain a multi-layered skill set for tackling complex temporal data challenges. The objective is to equip researchers and practitioners with the tools and knowledge to enhance predictive accuracy and event identification in various domains.

ABSTRACT. The unprecedented results obtained by Artificial Intelligence research and developments generated enthusiasm but also some worries about the balance among humans and intelligent agents. The huge impact of generative AI, particularly ChatGPT, raised ethics worries about observing IPR regulations, even addressing protection of the human rights. The ecologists also warned about the impact on the nature due to the creation of larger and larger language models.

The history of science taught us that any new and powerful technology comes with risks and opportunities and was met with reservations or denials, growling protests etc. What we see today on the AI canvas is a mixture of acclamations and apocalyptic warnings. AI has the potential to fundamentally transform human experience, offering huge benefits to humanity but also presents risks, which, ignored or mismanaged, can generate imbalances and destructive tensions!

This justifies the intense concern on this issue by international bodies, national or regional authorities, the media or many NGOs around the world, of the majority of people.

ABSTRACT. The novel research institute ICAM focusing on environment studies gathers specialists in biology, geography, physics, chemistry, earth sciences, economy, psychology, mathematics, and computer science. It offers opportunities for multi-disciplinary research. This is also the case for the ROCS project that aims to design a distributed and federated solution integrating cloud-native storage formats and processing tools for Earth observation data (satellite, aerial, in-situ) and to develop a pre-operational version of the platform, integrating all relevant data and validating functionalities through specific case studies.

ABSTRACT. Large Language Models (LLMs) are taking the world by storm and quickly moving up in the list of priorities of most organizations worldwide. Some of the most popular approaches to working with LLMs are Retrieval-Augmented Generation (RAG), Chain of Thought, and Reason/Act (ReAct).

Other important capabilities that fuel the LLM revolution are latent vector spaces and function calling which are at the core of some of the most spectacular applications of LLMs.

This session aims to deliver a grounded (pun intended) walkthrough of the most important LLM-related capabilities and applications, featuring a critical stand to separate reality from hype.

ABSTRACT. We propose a solution for handling abort commands given to robots. The solution is exemplified with a running scenario with household kitchen robots. The robot uses planning to find sequences of actions that must be performed in order to gracefully cancel a previously received command. The Planning Domain Definition Language (PDDL) is used to write a domain to model kitchen activities and behaviours, and this domain is enriched with knowledge from online ontologies and knowledge graphs, like DBPedia.

ABSTRACT. Programmers will always have a dilemma: they must make a choice between Performance, Flexibility and/or Fast Development. In this work, an experimental study is presented regarding the advantage of using assembly language in machine learning algorithms, such as linear regression and k-nearest neighbors. Conclusions as well as future research directions are also included.

ABSTRACT. The proliferation of sophisticated AI generative models like GPT-4 has revolutionized natural language processing (NLP) but also raised critical concerns about content authenticity in academia, media, and digital communications. This paper introduces a dual-approach AI-generated text detector that leverages both traditional machine learning (ML) techniques and advanced fine-tuned large language models (LLMs). Utilizing a comprehensive dataset of over 350,000 samples from five benchmark sources, our approach demonstrated robust performance, with conventional ML methods achieving 91-92% accuracy and fine-tuned LLMs such as BERT and RoBERTa reaching 97-98% accuracy. We developed TruAIText, a practical tool that integrates these models to provide detailed analysis of AI-generated content, including paragraph-level probabilities. Despite its efficacy, the tool requires ongoing updates to counteract adversarial manipulation. Future work will focus on enhancing model resilience and expanding the dataset to ensure adaptability to new AI-generated text models. This research offers a significant contribution to maintaining content integrity and trust in various sectors by providing an effective AI text detection solution.

ABSTRACT. This paper determines the minimum and maximum values of the misbalance deg index and associated graphs for trees and unicyclic graphs. It also discusses some general properties of this index and introduces a related quantity - the edge-averaged misbalance deg index - which is proven to be a more suitable measure for graph "imbalancedess". Furthermore, the paper also establishes the minimum and maximum values of the edge-averaged misbalance deg index and associated graphs for trees and unicyclic graphs. In the case of unicyclic graphs, the second-largest and third-largest values and associated graphs are also obtained for both indices.

ABSTRACT. The power of neural network increases its popularity more and more everyday. However, as for any powerful and popular tool, the security is hard to gain, here it is also the case. In recent years, backdoor attacks have become more and more popular through malicious parties, being risky for both organizations and people. In this paper we discuss the main methods of prevention against this type of attacks. We also provide an experiment, with the help of which we tried to see what part of an image influences its classification. Therefore, with the help of an algorithm for identifying backdoor attacks we made an experiment intended to find the interest zone in X-ray images.

ABSTRACT. As part of the GECCO 2024 competition this paper covers a possible approach to finding a solution for the SpOC 3: Torso Decomposition challenge. We use an evolutionary algorithm and compare it to a fully random approach. Our chosen algorithm could not outperform the random approach when computing 1 million different permutations. We conclude that our algorithm is insufficient and plan to improve on it in future research.