Sirvi Märksõna "Artificial intelligence" järgi

Nüüd näidatakse 1 - 8 8

listelement.badge.access-status Avatud juurdepääs ,
Alignment and Safety Challenges in a Superintelligent AI Landscape
(Tartu Ülikool, 2025) Jõemaa, Evelin; Eden, Grace, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Artificial intelligence (AI) technologies are developing at a rapid pace, and with that, the discussions on AI safety have gained importance. This thesis aims to explore the concerns that members of online AI alignment communities have in regards to AI safety and alignment and compare them to the opinions of the key figures in the AI safety and alignment field. Five interviews were conducted via online conferencing platforms with participants recruited from online AI alignment communities. The interviews were on the topics of AI safety, AI alignment, and ethical concerns of AI development. Additionally, five interviews with key figures in the AI safety and alignment field were analysed from interviews available on YouTube for comparison with participant interviews. Interviews were transcribed, and thematic analysis was conducted to identify key themes. The findings show a significant concern that the participants have regarding current safety measures. The concerns were often related to the rapid speed of the advancements in AI technology and the shortcomings they saw with the current safety measures in being able to handle the developments. Participants saw the unintended consequences of AI development being a bigger risk in the future, but also highlighted the already present risks of current AI models, such as the ability for people to create deep-fakes, etc. The community forum participants highlighted the need for more collaboration between private companies and governments to have better measures put in place internationally for developing safe AI models.
listelement.badge.access-status Avatud juurdepääs ,
Elections in digital times: a guide for electoral practitioners
(2022) Krimmer, Robert; Rabitsch, Armin; Kužel, Rast’o; Achler, Marta; Licht, Nathan
Strengthening democracy and electoral processes in the era of social media and Artificial Intelligence Democracy requires free, periodic, transparent, and inclusive elections. Freedom of expression, freedom of the press, and the right to political participation are also critical to societies ruled by the respect of human rights. In today’s rapidly evolving digital environment, opportunities for communication between citizens, politicians and political parties are unprecedented –– with information related to elections flowing faster and easier than ever, coupled with expanded opportunities for its verification and correction by a growing number of stakeholders. However, with billions of human beings connected, and disinformation and misinformation circulating unhinged around the networks, democratic processes and access to reliable information are at risk. With an estimated 56.8% of the world’s population active on social media and an estimate of 4 billion eligible voters, the ubiquity of social networks and the impact of Artificial Intelligence can intentionally or unintentionally undermine electoral processes, thereby delegitimizing democracies worldwide. In this context, all actors involved in electoral processes have an essential role to play. Electoral management bodies, electoral practitioners, the media, voters, political parties, and civil society organizations must understand the scope and impact of social media and Artificial Intelligence in the electoral cycle. They also need to have access to the tools to identify who instigates and spreads disinformation and misinformation, and the tools and strategies to combat it. This handbook aims to be a toolbox that helps better understand the current scenario and share experiences of good practices in different electoral settings and equip electoral practitioners and other key actors from all over the world to ensure the credibility of the democratic system in times of profound transformations.
listelement.badge.access-status Avatud juurdepääs ,
Exploring Data Quality Management Challenges and the Emerging Role of AI Solutions
(Tartu Ülikool, 2025) Käosaar, Kaisa; Nikiforova, Anastasija, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
High-quality data is essential for reliable decision-making and efficient operations across organizations. However, managing data quality (DQ) remains a complex and resource-intensive challenge. In response, artificial intelligence (AI) has been increasingly integrated into data quality tools. Yet, there is limited understanding of whether these AI-powered tools meet the practical needs of data professionals. This study addresses this gap by investigating how current AI-enabled data quality tools address the practical needs and challenges faced by data professionals. Using a mixed-methods approach, the study combines semi-structured expert interviews with a structured analysis of 28 AI-enabled data quality tools. Interview findings reveal persistent challenges such as limited support for unstructured data, low explainability, fragmented workflows, and minimal involvement of business users. While many tools perform well in data profiling, rule-based validation, and structured data integration, fewer support collaboration, domain-specific customization, or transparent AI behaviour. Despite progress, most tools fall short of meeting the complex and context-driven demands of enterprise-level data quality management (DQM).
listelement.badge.access-status Avatud juurdepääs ,
Modelling Creativity Using Artificial Neural Networks
(Tartu Ülikool, 2025) Kirikal, Johan; Aru, Jaan, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
The mental processes and algorithms that lead to creativity are still largely unknown. Psychological theories suggest that creativity arises from a balance between associative memory structure and the executive processes that retrieve and recombine information. In this thesis, I explore whether artificial neural networks can exhibit creative-like behaviour through controlled entropy modulation. Using the Continuous Generative Flow Network (C-GFN), a biologically inspired architecture, I simulate the impact of increased stochasticity on creative output. By varying the standard deviation in the model’s inner dynamics, I test the hypothesis that higher internal entropy leads to more creative outputs, as the entropy modulation theory of creativity predicted. As a result, models with elevated entropy not only generated more diverse symbolic representations but also did so more quickly and covered a greater distance in their latent space, emulating the faster-and-further phenomenon observed in human creativity research. These findings provide computational support for the associative and executive theories of creativity and highlight the role of entropy modulation in creative behaviour.
listelement.badge.access-status Avatud juurdepääs ,
Representation Learning on Free Text Medical Data
(Tartu Ülikool, 2021) Perli, Meelis; Kolde, Raivo, juhendaja; Laur, Sven, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Over 99% of the clinical records in Estonia are digitized. This is a great resource for clinical research, however, much of this data cannot be easily used, because a lot of information is in the free text format. In recent years deep learning models have revolutionized the natural language processing field, enabling faster and more accurate ways to perform various tasks including named entity recognition and text classifications. To facilitate the use of such methods on Estonian medical records, this thesis explores the methods for pre-training the BERT models on the notes from “Digilugu”. Three BERT models were pre-trained on these notes. Two of the models were pre-trained from scratch. One on only the clinical notes, the other also used the texts from the Estonian National Corpus 2017. The third model is an optimized version of the EstBERT, which is a previously pre-trained model. To show the utility of such models and compare the performance, all four models were fine-tuned and evaluated on three classification and one named entity recognition downstream tasks. The best performance was achieved with the model trained only on notes. The transfer learning approach used to optimize the EstBERT model on the clinical notes improved the pre-training speed and performance, but still had slightly worse performance than the best model pre-trained in this thesis.
listelement.badge.access-status Avatud juurdepääs ,
Süstemaatiline analüüs GPT-4o, DALL·E 3 ja Stable Diffusion 3.5 põhjal. Kas pildigeneraatorid integreerivad mudeleid maailma kohta?
(Tartu Ülikool, 2025) Lindström, Helena; Aru, Jaan, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
Text-to-image models intend to create correct images based on given prompts. This research aims to determine whether studied text-to-image models have integrated world models. Text-to-image models GPT-4o, DALL·E 3 and Stable Diffusion 3.5 generate images based on prompts. For this study, 25 objects that follow a specific logic have been selected, and based on these objects, 25 prompts have been constructed. Based on this study, GPT-4o appeared to be the best in depicting selected objects as 31% of the generated images were correct. Only 12% of images created by DALL·E 3 and 11% of images created by Stable Diffusion 3.5 were correct. In conclusion, due to poor results, it can be stated that GPT-4o, DALL·E 3 and Stable Diffusion 3.5 have not incorporated world models.
listelement.badge.access-status Avatud juurdepääs ,
Tehisintellekti kasutamine gümnaasiumiõpilaste õppetöös
(Tartu Ülikool, 2025-06-04) Laak Kaarel; Laius, Anne; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Loodusteadusliku hariduse keskus
listelement.badge.access-status Avatud juurdepääs ,
Tehisintellektil põhinev valgu kolmemõõtmelise struktuuri ennustamise tarkvara AlphaFold2
(Tartu Ülikool, 2025) Lehes, Caroline; Remm, Maido, juhendaja; Tartu Ülikool. Loodus- ja täppisteaduste valdkond; Tartu Ülikool. Arvutiteaduse instituut
The aim of this thesis is to investigate which algorithmic innovations have enabled the artificial intelligence–based protein structure prediction model AlphaFold2 to achieve significantly higher accuracy compared to earlier approaches. First, an overview is given of the hierarchical levels of protein structure and the experimental methods used for structure determination. The following section focuses on the algorithmic architecture of AlphaFold2 and its deep learning–based working principles. The practical part aims to demonstrate a potential use case in research: the three-dimensional structure of a hypothetical bacteriophage protein is predicted using the ColabFold (Mirdita et al., 2022) environment, and structurally similar proteins are searched with FoldSeek (van Kempen et al., 2023) to infer the protein’s potential function.