News and Analysis from the Era of Large Language Models
Due to the excessive use of ChatGPT's image generation function, Sam Altman, CEO of OpenAI, on 30 March 2025, urged users to cut back on generating Studio Ghibli-style images, as it is placing an extraordinary strain on the company’s GPU resources and staff. In a message posted
NVIDIA’s annual GPU Technology Conference (GTC) in March 2025 drew significant interest, but investors responded with muted enthusiasm to CEO Jensen Huang’s two-hour keynote speech, resulting in a 3.4% drop in the company’s share price on 18 March. Investors had anticipated announcements of new revenue streams;
On 3 March 2025, Google launched its new Data Science Agent tool, powered by Gemini 2.0 artificial intelligence, on the Colab platform. The new AI agent automates data analysis processes and generates complete Jupyter notebooks based on simple natural language instructions. Integrated into Colab, the tool is primarily designed
On 6 March 2025, Mistral AI unveiled Mistral OCR, an advanced document interpretation API that offers processing of 1,000 pages for $1, surpassing competitors with an impressive accuracy rate of 94.89%. The service is capable of interpreting complex documents, including images, tables, and mathematical formulas, while preserving the
Manus, the general artificial intelligence agent developed by the Wuhan-based Butterfly Effect startup and debuted in early March 2025, has gained global attention with praise from influential tech personalities such as Twitter co-founder Jack Dorsey and Victor Mustar, Head of Product at Hugging Face. The platform, which describes itself as
In March 2025, Sakana AI’s "AI Scientist-v2" system achieved a significant milestone by producing the first fully AI-generated scientific paper to successfully pass peer review at a workshop of the ICLR conference. The research team submitted three AI-generated studies, one of which—titled "Compositional Regularization: Unexpected
In February 2025, Microsoft introduced two new members of the Phi-4 model family, with the Phi-4-multimodal-instruct being particularly noteworthy. Despite having just 5.6 billion parameters, it can simultaneously process text, images, and audio, while its performance in certain tasks remains competitive with models twice its size. The Phi-4-multimodal-instruct was
Hungarian language technology research has reached a significant milestone: a comprehensive study has revealed that a larger corpus size does not necessarily lead to improved performance in morphological analysis. In their study, Andrea Dömötör, Balázs Indig, and Dávid Márk Nemeskey conducted a detailed analysis of three Hungarian-language corpora of varying
Released on 17 March 2025 under the Apache 2.0 licence, Mistral’s Small 3.1 model marks a significant advancement in artificial intelligence technology. This 24-billion-parameter model delivers improved text generation performance, enhanced multimodal capabilities, and an expanded context window of 128,000 tokens, while maintaining an inference speed
Tibor Szécsényi and Nándor Virág, researchers at the University of Szeged, have explored the context sensitivity of the huBERT language model in pragmatic annotation, focusing in particular on the automatic identification of imperative verb functions. Their study, conducted on the MedCollect corpus—a dataset of health-related misinformation—investigates how both
On 28 January 2025, OpenAI introduced ChatGPT Gov, a dedicated AI platform developed specifically for US government agencies. Delivered via Microsoft Azure’s cloud services, the platform provides access to OpenAI’s most advanced language models for public sector use. ChatGPT Gov enables government bodies to adopt AI technologies while
On 16 March 2025, Baidu unveiled its two latest artificial intelligence models: the ERNIE 4.5 multimodal foundation model and the ERNIE X1 reasoning model. These new models represent a significant leap forward in multimodal and reasoning AI while offering solutions that are just a fraction of the cost of
