news

A new study shows that word-initial consonants are systematically lengthened across a diverse sample of languages Speech consists of a continuous stream of acoustic signals, yet humans can segment words from each other with astonishing precision and speed. To find out how this is possible, a team of linguists has analysed durations of consonants at different positions in words and utterances across a diverse sample of languages.

UdeM literature professor Marcello Vitali-Rosati looks at how, for better or worse, computerized large language models are changing how we write - and what we think about it. Computerized large language models (LLMs) are making inroads into the realm of literature. Their ability to generate coherent texts and mimic all manner of writing styles has sparked lively debate among writers, literary theorists and researchers.

Researchers have found that AI large language models, like GPT-4, are better at predicting what comes next than what came before in a sentence. This "Arrow of Time" effect could reshape our understanding of the structure of natural language, and the way these models understand it. Large language models (LLMs) such as GPT-4 have become indispensable for tasks like text generation, coding, operating chatbots, translation and others.

Researchers developed an easy-to-use tool that enables an AI practitioner to find data that suits the purpose of their model, which could improve accuracy and reduce bias. In order to train more powerful large language models, researchers use vast dataset collections that blend diverse data from thousands of web sources.

In controlled experiments, MIT CSAIL researchers discover simulations of reality developing deep within LLMs, indicating an understanding of language beyond simple mimicry. Ask a large language model (LLM) like GPT-4 to smell a rain-soaked campsite, and it'll politely decline. Ask the same system to describe that scent to you, and it'll wax poetic about "an air thick with anticipation" and "a scent that is both fresh and earthy," despite having neither prior experience with rain nor a nose to help it make such observations.

Large language models like ChatGPT cannot learn independently or acquire new skills, meaning they pose no existential threat to humanity. ChatGPT and other large language models (LLMs) cannot learn independently or acquire new skills, meaning they pose no existential threat to humanity, according to new research from the University of Bath and the Technical University of Darmstadt in Germany.

The reason for this might be the feeling of joint social group membership "Holiday" or "vacation", "to start" or "to begin", "my friend's cat" or "the cat of my friend" - in our language, there are different ways of expressing the same things and concepts. But can the choice of a particular variant determine whether we prefer to cooperate with certain people rather than with others? A research team led by Theresa Matzinger from the University of Vienna investigated this and showed that people are more likely to co-operate with others if they make similar linguistic choices in a conversation.

New CSAIL research highlights how LLMs excel in familiar scenarios but struggle in novel ones, questioning their true reasoning abilities versus reliance on memorization. When it comes to artificial intelligence, appearances can be deceiving. The mystery surrounding the inner workings of large language models (LLMs) stems from their vast size, complex training methods, hard-to-predict behaviors, and elusive interpretability.

Combining natural language and programming, the method enables LLMs to solve numerical, analytical, and language-based tasks transparently. Large language models like those that power ChatGPT have shown impressive performance on tasks like drafting legal briefs, analyzing the sentiment of customer reviews, or translating documents into different languages.

Wikipedia is the largest platform for open and freely accessible knowledge online yet, in a new study, researchers have found that around 15% of the content is effectively invisible to readers browsing within Wikipedia. They have developed a new tool to help overcome this. With 60 million articles in more than 300 language versions, Wikipedia's available content grows continuously at a rate of around 200 thousand new articles each month.

For decades, researchers have assumed that people overestimate the risk of dramatic causes of death, such as road traffic accidents. The reason given for this was that such deaths are the subject of far greater media attention than more significant but less spectacular mortality risks. However, a study at the Technical University of Munich (TUM) has now debunked this assumption.

Linguistic analysis provides insight into the vocabularies for body parts in more than a thousand languages Human bodies have similar designs. However, languages differ in the way they divide the body into parts and name them. A team of linguists from the Max Planck Institute for Evolutionary Anthropology in Leipzig and the University of Passau conducted a comparison of body part vocabularies to shed light on the interplay between language, culture, and perception of the human body.

Study finds language-processing difficulties are an indicator - more so than memory loss - of amnestic mild cognitive impairment. Individuals with mild cognitive impairment, especially of the "amnestic subtype" (aMCI), are at increased risk for dementia due to Alzheimer's disease relative to cognitively healthy older adults.

Twitter study may signal shift to greater Ukrainian self-identification Published on Wednesday 10 January 2024 Last updated on Monday 15 January 2024 Russia's invasion of Ukraine in 2022 significantly accelerated the shift from Russian language to Ukrainian on social media in what may signal a move towards greater Ukrainian self-identification, according to new research from the University of Bath, Ludwig-Maximilians-University Munich, and the Technical University of Munich.

Systematic testing of OpenAI's GPT-3 reveals that question format can influence models to agree with misinformation New research into large language models shows that they repeat conspiracy theories, harmful stereotypes, and other forms of misinformation. In a recent study, researchers at the University of Waterloo systematically tested an early version of ChatGPT's understanding of statements in six categories: facts, conspiracies, controversies, misconceptions, stereotypes, and fiction.

Study highlights the positive impact of doctors working over the festive period - may prompt broadcasters to air new Doctor Who episodes each year at Christmas A new Doctor Who episode shown during the festive period, especially on Christmas Day, is associated with lower death rates in the subsequent year across the UK, finds a new study.

Novels and poems often contain descriptions of plants or animals - sometimes more, sometimes less detailed. The extent to which flora and fauna feature in a literary work also depends on who wrote it and under what circumstances. For example, female authors tend to use more species names when they write.

Study on the connection between language and thinking by neuroscientists at Freie Universität Berlin published The influence of language on human thinking may be stronger than previously thought. This is the result of a new study by the language, cognition, and neuroscientist Friedemann Pulvermüller and his team from the Laboratory for Brain and Language Research at Freie Universität Berlin.

Department of Brain and Cognitive Sciences faculty members Ev Fedorenko, Ted Gibson, and Roger Levy believe they can answer a fundamental question: What is the purpose of language? Over a decade ago, the neuroscientist Ev Fedorenko asked 48 English speakers to complete tasks like reading sentences, recalling information, solving math problems, and listening to music.

As in previous years, the speech from the throne was difficult once again: it contained many complicated and unfamiliar words, phrases, and constructions. More even, than the speeches from the throne of the previous four years, analysis by linguists Leo Lentz and Henk Pander Maat shows. Difficult speech from the throne To measure the level of difficulty of the speech from the throne, Lentz and Pander Maat used LiNT, a software tool developed by Utrecht University.