Large language model
This paper was accepted at the EMNLP Workshop on Computational Approaches to Linguistic Code-Switching (CALCS). It explores the challenges of code-switching (mixing different languages in a sentence) in Natural Language Processing (NLP). Previous studies have shown promising results for end-to-end speech translation, but this paper focuses on unexplored areas.
The beef supply chain is complex and requires more visibility than ever to manage inventory and maintain consumer trust. McDonald’s has partnered with Golden State Foods to use RFID technology to track the movement of fresh beef from manufacturer to restaurant in real-time. This “phygital” approach merges technology with physical object identifiers to create efficient…
Xenova’s text-to-speech client utilizes transformer-based neural networks to generate natural-sounding synthetic speech. It offers high-quality synthetic speech that is indistinguishable from human voice, supports various voices and languages, and allows fine-grained control over speech synthesis. The client has applications in e-learning, accessible media, audiobooks, voice assistants, and more. It can be easily installed and tested…
AI web scraping operations that collect online artworks without consent or compensation of the creators have become a major concern for artists. Existing solutions have been limited, but researchers have developed a tool that subtly manipulates image pixels to disrupt AI models’ training process. This tool offers hope for artists and creative entities by safeguarding…
Researchers from the University of Washington and Princeton have developed a benchmark called WIKIMIA and a detection method called MIN-K% PROB to identify problematic training text in large language models (LLMs). The MIN-K% PROB method calculates the average probability of outlier words, allowing researchers to determine if an LLM was trained on a given text.…
A study by Randstad reveals that Indian workers are more concerned about job loss due to artificial intelligence (AI) compared to workers in countries like the US, UK, and Germany. The study found that one in two workers in India are scared of losing their jobs to AI, while the number is one in three…
Joy Buolamwini’s book, “Unmasking AI: My Mission to Protect What Is Human in a World of Machines,” discusses the concept of “x-risk,” the existential risk that AI poses. She argues that existing AI systems that cause harm are more dangerous than hypothetical superintelligent systems. Buolamwini also emphasizes the importance of addressing algorithmic bias and ensuring…
Researchers from Georgia Tech, Mila, Université de Montréal, and McGill University have introduced a scalable framework called POYO-1 for modeling neural population dynamics in diverse large-scale neural recordings. The framework utilizes tokenization, cross-attention, and the PerceiverIO architecture to capture temporal neural activity and achieve few-shot performance in various tasks. It demonstrates rapid adaptation to new…
OpenAI has introduced new features to ChatGPT Plus, affecting AI startups. Users can now access all ChatGPT tools without switching, including Browsing, Advanced Data Analysis, and DALL-E. PDF analysis, previously available through plugins, is now integrated. This move disrupts the business model of startups that developed these plugins. The impact on AI startups is significant…
Researchers from MIT and NVIDIA have developed two techniques that can accelerate the processing of sparse tensors, a type of data structure used for high-performance computing. The techniques, called HighLight and Tailors/Swiftiles, can improve the performance and energy-efficiency of hardware accelerators designed for processing sparse tensors. HighLight can efficiently handle various sparsity patterns, while Tailors/Swiftiles…
MIT researchers have developed a search engine, called SecureLoop, that can identify optimal designs for deep neural network accelerators while maintaining data security. The tool considers the impact of adding encryption and authentication measures on performance and energy usage. It improves accelerator designs by boosting performance and keeping data protected, enabling the improvement of AI…
MIT researchers have found evidence suggesting that the brain may develop an intuitive understanding of the physical world through a process similar to self-supervised learning. Using models known as neural networks, they trained them using self-supervised learning techniques and found that the resulting models generated activity patterns similar to those seen in the brains of…
The researchers from Columbia University and Apple have developed Ferret, a multimodal large language model (MLLM) that combines referencing and grounding for improved image understanding and description. Ferret uses a hybrid region representation and a spatial-aware visual sampler to handle a variety of regional forms and can handle input that combines free-form text and referenced…
Joy Buolamwini, a prominent AI researcher and activist, calls for a radical rethink of AI systems, highlighting the unethical practices of many AI companies. She emphasizes the need for rigorous testing and auditing of AI systems before deployment to avoid harmful consequences. Buolamwini also shares her personal journey of becoming an accidental activist and the…
Microsoft exceeded Wall Street’s Q1 financial projections across all sectors, driven by cloud computing and the Windows operating system. The company’s revenue also surpassed analysts’ expectations, largely due to the anticipation of the release of Microsoft 365 Copilot, a suite of AI tools developed in collaboration with OpenAI. Azure’s revenue grew by 29%, outperforming projections.…
OpenAI has established a team called “Preparedness” to address the potential risks associated with AI. The team will evaluate current and future AI models for risks such as tailored persuasion, cybersecurity threats, autonomous replication, and even existential threats like chemical, biological, and nuclear attacks. OpenAI believes that while advanced AI models can benefit humanity, they…
The author discusses how to succeed in your first data role. They emphasize the importance of becoming comfortable with workflow and data structure, mastering the company’s toolbox, learning the business, sharpening your skills, and becoming self-sufficient. They suggest practicing unused skills, creating personal projects, and managing projects from start to end. In a year or…
GROOT is a new imitation learning technique developed by researchers at The University of Texas at Austin and Sony AI. It addresses the challenge of enabling robots to perform well in real-world settings with changing backgrounds, camera viewpoints, and object instances. GROOT focuses on building object-centric 3D representations and uses a transformer-based strategy to reason…
MLCommons has formed the AI Safety Working Group (AIS) to develop benchmarks for AI safety. Currently, there is no standardized benchmark to compare the safety of different AI models. AIS will build upon the Holistic Evaluation of Language Models (HELM) framework developed by Stanford University to create safety benchmarks for large language models. Several prominent…
AutoMix is an innovative approach to allocating queries to language models (LLMs) based on the correctness of responses. It uses context and self-verification to ensure accuracy, and can switch between different models. AutoMix enhances performance and computational cost in language processing tasks and demonstrates promising capabilities for future research and application.