On 20 January, the Chinese AI firm DeepSeek announced that it had developed a chatbot matching OpenAI’s performance while requiring significantly fewer computational resources. This revelation sent shockwaves through the market, culminating in a 17% plunge in NVIDIA’s stock price by 27 January. Tech industry leaders, including Microsoft
Building upon the foundational principles of the attention mechanism discussed in the previous section, the Transformer architecture represents a paradigm shift by leveraging attention exclusively, completely replacing the recurrent structures that once dominated sequence modeling. This architectural innovation, first unveiled by Vaswani et al. (2017), has since catalysed a seismic
On 31 January 2025, OpenAI introduced the o3-mini model, marking a significant advancement in the company’s cost-efficient artificial intelligence development. Replacing the previous o1-mini, the new model features enhanced STEM capabilities while delivering a 24% faster response time than its predecessor. The o3-mini delivers outstanding performance in mathematics, coding,
The study "Identification of Social Scientifically Relevant Topics in an Interview Repository: A Natural Language Processing Experiment" presents a pioneering experiment in the automated processing of qualitative data in social sciences. The joint project of the Research Documentation Centre at the Centre for Social Sciences (TK KDK) and
Led by the HUN-REN Centre for Social Sciences, the MORES project (Moral Emotions in Politics: How They Unite, How They Divide) was launched as a collaboration between nine European institutions. The project examines the challenges facing liberal democracy, with a particular focus on the political role of moral emotions and
Addressing artificial intelligence (AI) hallucinations is a critical challenge for ensuring the technology’s reliability. A recent study suggests that multi-level agent systems, combined with natural language processing (NLP)-based frameworks, could significantly mitigate this issue. In the study "Hallucination Mitigation using Agentic AI Natural Language-Based Frameworks," Gosmar
In 2024, the Technology Innovation Institute (TII) of the United Arab Emirates introduced two significant artificial intelligence models in the Falcon series, redefining the development of large language models (LLMs). The Falcon 2, launched in May, and the Falcon 3, released in December, reflect the UAE’s commitment to democratising
Levente, Hungary's latest supercomputer, is being built at the Wigner Research Centre for Physics in Csillebérc as part of the EuroHPC JU program. The project holds strategic importance for domestic research and development and industrial technological progress, particularly since the new machine will be four times more powerful
The attention mechanism has fundamentally reshaped natural language processing (NLP), enabling models to capture complex linguistic relationships with unprecedented accuracy. Introduced prominently in Vaswani et al. (2017), attention allows models to focus on relevant parts of input sequences, enhancing performance in tasks like machine translation and sentiment analysis. This essay
Alibaba has unveiled its latest artificial intelligence model, Qwen 2.5-Max, which the company claims outperforms the current market leaders, including DeepSeek-V3, OpenAI’s GPT-4, and Meta’s Llama-3. A Mixture-of-Experts (MoE) architektúrára épülő modellt több mint 20 billió tokenen tanították, majd felügyelt finomhangolással (SFT) és emberi visszajelzéseken alapuló megerősítéses
In January, the Chinese AI startup DeepSeek launched its R1 language model, delivering performance comparable to OpenAI’s models at a development cost of just $6 million. This poses a significant challenge to Europe’s Mistral AI, which has secured over €1 billion in investment and reached a valuation of
The Chinese company DeepSeek made a surprising entrance into the market of ChatGPT and other leading language models in January 2025, creating a new situation for European AI developers as well. The company, founded by Liang Wenfeng in July 2023, has made its advanced language model freely available, aiming to
